War Predictions 2026–2027: A 12-Month Global Conflict Risk Forecast | The Purple Report

September 2026 monthly edition. Assessment date: 23 September 2026. Forecast window: 23 September 2026 to 23 September 2027.

War predictions for 2026–2027 need to distinguish where existing wars may continue, where fighting may spread, and where a new interstate campaign could begin. This twelve-month global conflict risk forecast examines the Gulf and Red Sea, Ukraine, Sudan, Myanmar, the Sahel, eastern Democratic Republic of the Congo, South Asia, the Horn of Africa, Taiwan and the South China Sea. It also explains how distant conflict could affect Singapore through shipping, energy, technology, travel and essential services.

The Purple Report’s central forecast is that continued wars and recurring regional escalation are more plausible than a single, coordinated world war during this window. Its provisional numerical assessments place further sustained direct combat involving Iran and the United States or Israel above even odds, while a large-scale invasion of Taiwan’s main island remains a lower-probability, very-high-consequence scenario. These are subjective analytical judgments, not measured frequencies, official warnings or predictions of an inevitable future.

This report applies the public framework in How War Works with StrategizeOS, WarOS, Weather, Geography, and Environment. That framework connects political aims to geography, time, sustainment, restraint and civilian consequences. The present report has a different job: to state bounded forecasts, attach them to dated evidence, identify what would change them, and preserve the original assessment for later review.

Reading routes: 1. Daily Civilisation Pulse · 2. Today’s Top Signals · 3. System Map · 4. Pressure vs Repair · 5. Shadow Signals · 6. What To Watch Tomorrow · 7. Final Daily Reading · 8. ID / Almost-Code

Quick read: what this report actually predicts

The most important distinction is between an event’s likelihood and its consequences. A war already under way can be very likely to continue without being a new war. An encounter can become lethal without turning into a sustained campaign. A maritime inspection regime can obstruct trade without an amphibious invasion. A lower-probability event can still deserve extensive civilian preparedness because there is little time to improvise once it begins.

The numerical register contains eight different, explicitly defined events. Further Iran-related direct combat is assessed at 60–80%; renewed sustained Pakistan–Afghanistan combat at 45–65%; sustained Ethiopia–Eritrea or India–Pakistan combat at 20–35% each; a specified lethal China–Philippines exchange at 15–30%; a sustained enforced Taiwan interdiction regime at 10–20%; a main-island Taiwan invasion at 5–10%; and sustained Russia–NATO combat at 5–15%. The events differ, overlap and are not independent. The percentages must not be added to obtain a probability of world war.

The continuation watch is equally important. Ukraine, Sudan, Myanmar, the central Sahel and eastern DRC remain in the report because the civilian consequences of persistent organised violence matter even when there is no new declaration, surprise invasion or dramatic change of front. Their entries are qualitative continuation assessments, not additional numerical forecasts disguised as facts.

The strongest reason for caution about the numbers is the absence of a calibrated forecasting record behind this first war register. The ranges make the judgment inspectable; they do not make it statistically validated. Each has a defined window, an event threshold, sources, alternative explanations and revision conditions. A future monthly edition must show changes openly rather than quietly replacing an inconvenient baseline.

For Singapore, the base case is external disruption rather than direct conventional attack. That is an analytical assessment, not a security guarantee. The practical questions concern continuity: whether essential supplies arrive, whether firms can finance delayed shipments, whether households receive reliable information, whether travellers can change plans, and whether schools can maintain learning without turning every alarming headline into an emergency.

Reader routes

Read Daily Civilisation Pulse for the definitions and forecast register. Read Today’s Top Signals for the regional reasoning. Read System Map for the connection between war, weather, trade, health and Singapore. Read Pressure vs Repair for evidence of restraint. Read Shadow Signals for claims that remain below the threshold for publication as facts. Read What To Watch Tomorrow for the monthly review method and civilian decision examples. The final sections preserve the synthesis and public record identifiers.

This is a dated research edition. Sources range from current September reporting to explicitly dated institutional background. A source’s publication date is not automatically the date of the event it describes. No earlier daily Purple Report is used as independent proof that a news claim is true, and no count of completed scans or execution of a calibrated prediction engine is asserted here.

1. Daily Civilisation Pulse

1.1 The monthly pulse: continuation, escalation, restraint and exposure

The opening question is not simply whether the world is dangerous. It is which transitions are plausible during the next twelve months, what evidence supports that view, and what is still unknown. A useful pulse therefore follows four different tracks: the persistence of existing combat, the entry of additional belligerents, the implementation of restraint, and the condition of civilian functions exposed to the fighting.

Those tracks can move in different directions. A ceasefire may reduce direct fighting while leaving water networks damaged and displaced households unable to return. A market may recover while shipping remains expensive. A government may open talks while retaining the capability to resume attacks. Conversely, an isolated violent incident can occur while broader relations remain contained. One headline cannot stand in for all four tracks.

The current evidence illustrates the distinction. A senior Iranian official’s conditional reopening proposal on 22 September is evidence of a bargaining position, not proof that commercial passage has normalised. UNFPA’s 21 September Yemen update documents disrupted access to reproductive care amid displacement. These are related regional observations, but they measure different realities: diplomatic intent in one case and civilian access in the other. Reuters, 22 September; UNFPA, 21 September.

A monthly report should not conceal those differences inside one sweeping verdict about a country or leader. The forecast concerns events. The consequence analysis concerns people and services. The evidence review concerns what can be known. Keeping these separate makes it possible to say that a scenario is plausible while refusing to claim that a particular government has secretly decided to execute it.

1.2 What “war” means in this report

International law, conflict datasets and ordinary language use different classifications. The ICRC explains that an international armed conflict can arise when states resort to armed force against one another; its legal classification is not dependent on this report’s seven-day convention. UCDP’s statistical definitions use calendar-year battle-death thresholds, distinguishing armed conflict from the higher-intensity category it calls war. Those are different analytical purposes. ICRC: international armed conflict; UCDP definitions.

For several numerical questions here, sustained combat means direct combat on at least seven distinct days within any thirty-day period wholly inside the forecast window. This is an editorial resolution rule. It is not a claim that a shorter campaign is harmless, that international humanitarian law begins on day seven, or that a four-day conflict is not a war in ordinary speech. A separate severity note will record consequential events that do not meet the numerical question.

That rule creates an important limitation. The report could assign a low probability to sustained combat while a shorter, destructive exchange remains quite plausible. Readers should therefore look at the event definition before interpreting a percentage. A forecast of a lethal encounter answers a different question from a forecast of a campaign. An invasion forecast answers a different question again.

Direct state combat is distinguished from violence by a non-state partner. Assistance, financing, training or political alignment may matter to escalation, but they do not automatically make every act by a partner count as direct combat between the sponsor and another state. Attribution must be corroborated. For the Iran question, state armed forces include attributable Iranian state military branches; the question does not count every attack by an Iran-aligned organisation as an Iranian regular-force attack.

Similarly, uniforms are not a shortcut to reliable attribution. In the China–Philippines question, the specified event involves an exchange of weapons fire between uniformed state forces and a corroborated death. Collision-only incidents and civilian fishing encounters are excluded from that numerical resolution. They still belong in the risk narrative. The choice of threshold is a way to keep the forecast honest, not a judgment about the importance of injured people.

1.3 The eight-event baseline register

Event	Defined outcome within 23 September 2026–23 September 2027	Provisional probability	Confidence in assessment
Iran–United States or Israel	Further direct combat between attributable state armed forces on at least seven distinct days within thirty days	60–80%	Medium
Pakistan–Afghanistan	Further direct combat between Pakistani forces and Afghan de facto state forces meeting the same duration threshold	45–65%	Medium
Ethiopia–Eritrea	Direct regular-force combat meeting the duration threshold; proxy-only fighting excluded	20–35%	Low–medium
India–Pakistan	Direct state-force combat meeting the duration threshold; an isolated collision or militant attack alone excluded	20–35%	Low–medium
China–Philippines	Reciprocal weapons fire between uniformed state forces causing at least one corroborated death	15–30%	Low–medium
Taiwan commercial interdiction	A broad, actually enforced inspection, quarantine or blockade regime materially obstructing access to Taiwan proper for seven consecutive days	10–20%	Low–medium
Taiwan main-island invasion	A large-scale landing campaign intended to seize and hold substantial territory on Taiwan’s main island	5–10%	Low
Russia–NATO	Sustained reciprocal direct combat between Russian regular forces and a NATO member’s forces meeting the duration threshold	5–15%	Low

These are the retained ranges from the initial 23 September research register. They are structured analyst judgments. They are not published probabilities from the cited institutions, outputs of an automated simulation, or statistical confidence intervals. The regional chapters explain the evidence and competing mechanisms rather than inventing a numerical calculation that did not occur.

The ranges should be read in both directions. A 20–35% assessment means the event deserves attention, but non-occurrence remains more likely in this judgment. A 60–80% assessment leaves a substantial possibility that restraint, bargaining or changing capacity prevents the defined outcome. Neither range licenses certainty. In particular, a low probability of invasion is not a prediction that there will be no intimidation, disruption, military activity or political pressure.

The apparent precision of the endpoints should not be mistaken for a measured margin of error. They communicate the breadth of the initial judgment. There is no demonstrated basis here for treating 35% as scientifically distinct from 34%. Monthly review should therefore explain material changes in evidence before changing a number, rather than moving percentages to create an appearance of constant analytical activity.

1.4 Probability, confidence and consequence are three different fields

Probability asks how likely the defined event is. Confidence asks how firmly the assessment rests on evidence, definitions and a defensible model of the situation. Consequence asks what follows if the event occurs. Combining them into a single label hides useful information. A high-consequence scenario can have a low event probability and low confidence because decisive information is not observable.

Consider a hypothetical port. An analyst may think a prolonged closure is unlikely, yet a hospital importing temperature-sensitive supplies through that port may still need an alternative. That decision need not prove the analyst’s probability wrong. It reflects the hospital’s exposure and the cost of late preparation. Another organisation with flexible suppliers may rationally choose a different response to exactly the same forecast.

The report therefore does not prescribe a universal reaction to any percentage. It supplies a structured description of possible events and their pathways. A traveller needs current official advice and itinerary-specific information. A business needs contractual and logistical information. A public agency needs broader assessments and authority to act. A school needs age-appropriate explanations and a continuity plan. Those are different decisions, not competing interpretations of one alarm level.

Confidence is especially constrained where official intentions are opaque, reporting is restricted, or source families depend on the same original statement. More articles can improve awareness without improving independence. Conversely, an authoritative source can be highly reliable about a dated measurement while remaining unable to reveal a future decision. Reliability about the present must not be promoted into certainty about the future.

1.5 What changed from the earlier Purple series

The daily Purple Reports developed a sequence from repair and memory towards prevention, redundancy, resilience cost, allocation and opportunity cost. This monthly war edition adds warning quality and decision thresholds. The question is no longer only where pressure exists. It is what observation should justify a change in judgment, how much confidence that observation deserves, and which response remains useful if the forecast is wrong.

This is the first frozen numerical baseline for these eight war questions, not evidence that earlier daily headlines predicted them. The report does not award itself credit for events before 23 September. The September airstrikes, collisions, negotiations and humanitarian releases are inputs available at the baseline. Only qualifying events inside the declared window can resolve the prospective numerical questions.

There is a second separation from the September natural-hazard forecasting report. That companion addresses physical hazards over a different planning horizon. War involves strategic actors who react to one another and can change behaviour after receiving warnings. The same evidence discipline can connect the two reports, but earthquake probabilities, seasonal forecasts and decisions to use military force cannot be merged into one universal numerical model.

1.6 Why the current assessment is not a countdown to world war

A worldwide war is not simply the sum of several regional conflicts. A defensible world-war question would need its own definition of participants, direct combat, duration and geographical extent. This report does not claim to have calculated that outcome from the eight regional ranges. The events share drivers, possible restraints and potential escalation pathways, so elementary independent-event arithmetic would be misleading.

The March 2026 U.S. intelligence assessment described cooperation among China, Russia, Iran and North Korea as selective and predominantly bilateral, with differing interests and constraints. That is a dated institutional assessment, not a guarantee of future behaviour. It nevertheless provides a reason to reject the unsupported premise that every regional event is already directed by one integrated alliance plan. ODNI release and prepared testimony, 18 March 2026.

The base case in this report is a patchwork: existing wars continue in several places; some crises are contained; some bargaining attempts fail; some commercial systems adapt; and particular incidents create opportunities for wider confrontation. This can produce grave human consequences without satisfying a world-war definition. The absence of a general war is not equivalent to peace for people already facing displacement, bereavement, hunger or prolonged loss of schooling.

1.7 How to use the long report without drowning in detail

Readers should begin with one bounded question. Are they trying to understand whether an existing war will continue, whether a new front could open, whether a particular route may be disrupted, or whether a claim circulating online is credible? The report’s regional chapters then provide a mechanism and its counterargument. Its system map shows how consequences could travel beyond the immediate theatre.

The most useful reading habit is to identify the sentence that would make the analysis wrong. If a chapter says a verified implementation mechanism would lower concern, a later report must recognise that mechanism when it appears. If it says a collision is not a shooting, the definitions must remain distinct even after a frightening video. If it predicts persistence, it must allow a genuine settlement to count against that prediction.

A forecast earns value by being correctable. Its purpose is not to remove uncertainty from human affairs, but to make uncertainty visible enough that people can ask better questions, protect essential functions and recognise evidence that changes the situation. That is the standard applied in the chapters that follow.

2. Today’s Top Signals

2.1 Iran, Israel, the United States and the Gulf: continuation is not new onset

Forecast: a 60–80% provisional probability of further direct combat between attributable Iranian state forces and U.S. or Israeli forces on at least seven distinct days within thirty days during the forecast window. Assessment confidence: medium. The event is continuation or re-escalation. It is not a claim that a conflict which already exists will begin for the first time.

Confirmed and attributed baseline: Reuters reported on 22 September that a senior Iranian official offered a conditional reopening of Hormuz if U.S. military pressure and the blockade were reduced. The proposal was a reported negotiating position. It did not establish that the conditions had been accepted or that passage had resumed at a normal commercial level. Reuters: conditional Hormuz proposal.

The central analytical problem is that several bargains may be running at once. A shipping arrangement can concern passage and security. A military pause can concern immediate retaliation. A broader negotiation can concern sanctions, nuclear activities or regional security. Agreement on one issue does not necessarily settle the others. The report’s higher probability rests on this possibility of incomplete settlements: a pause can reduce immediate pressure while leaving enough unresolved disputes for renewed direct combat later in the year.

This is a mechanism, not knowledge of a secret decision. A state may prefer to bargain and still retain coercive options. Another state may interpret continued preparation as evidence that a pause is temporary. Commercial operators may remain cautious even when political statements improve. Each actor is responding to a different exposure, and their clocks are not identical. A government can announce a policy in an hour; restoring predictable commercial confidence may take longer.

There are at least three plausible branches. In the first, reciprocal restraint is implemented and survives an early test. Commercial passage becomes more regular, technical talks continue, and later disputes are handled without sustained direct fighting. In the second, limited agreements hold temporarily but a new incident produces another campaign. In the third, direct fighting widens geographically or draws additional participants into combat. The numerical question can resolve in the second branch without the third occurring. That distinction prevents a prediction of further fighting from being inflated into a prediction of unlimited regional war.

The strongest counterargument to the elevated range is that costs can create incentives for accommodation. Damage, commercial uncertainty, resource demands and concern about uncontrolled escalation can make an implemented bargain more attractive than further combat. A credible assessment must allow this counterargument to change the number. It would be inconsistent to say economic pressure matters when it encourages confrontation but disregard it when it encourages restraint.

The most informative downward evidence would be reciprocal commitments that are implemented and independently observable. That might include a sustained reduction in direct attacks, reliable commercial access, an operating mechanism for disputed incidents, and evidence that the same arrangement survives political criticism. Announcing a meeting is a weaker signal. Announcing an agreement without demonstrating implementation is also weaker. The forecast should fall most when behaviour changes and remains changed.

Upward evidence would require a similarly concrete combination: renewed attributable direct attacks, disruption of the negotiating channel, and sustained reciprocal activity rather than a single exchange. No one indicator should carry the whole conclusion. A speech may be intended for bargaining. A market movement may reflect several causes. An anonymous account may be incomplete. Convergence matters, and convergence is not obtained by repeating one source across many outlets.

The civilian analysis must also remain separate from the combat threshold. Households can face higher transport costs, disrupted employment, medicine delays or travel restrictions even if the seven-day event does not occur. Commercial harm can persist after a military pause. Conversely, further direct combat does not establish that every regional port, airport or supply route is unusable. Country-level labels often conceal large differences in actual exposure.

For Singapore readers, the relevant connection is therefore conditional and functional. A Gulf escalation could affect the price, timing and certainty of imported energy and goods, but it does not follow that all Singapore gas physically passes through Hormuz. The energy chapter below distinguishes pipeline supply, global LNG markets, physical availability and price transmission. A strong warning explains the chain rather than substituting a map of the Gulf for a diagnosis of Singapore’s power system.

The monthly entry will be revised if these mechanisms change. It will not rise simply because the same confrontation remains prominent in the news, and it will not fall merely because attention moves elsewhere. The test is whether the circumstances that could sustain direct combat are strengthening or weakening relative to arrangements that can contain it.

2.2 Yemen and the Red Sea: one regional connection, several independent conflicts

Assessment: continued organised violence and commercial-security disruption remain important qualitative watches. No separate numerical Yemen probability is added in this edition. Houthi activity is not automatically counted as direct Iranian state combat for the Gulf forecast.

The current humanitarian evidence is specific. UNHCR reported on 15 September that renewed hostilities along Yemen’s western coast had displaced more than 100,000 people internally and sent others to Djibouti. WFP’s 17 September release described an expanding response under constrained funding. These releases establish civilian pressure and response activity; they do not establish that every Yemeni conflict actor has the same objectives or command structure. UNHCR, 15 September; WFP, 17 September.

Yemen illustrates the danger of calling an entire theatre a proxy war and stopping the analysis there. External support can affect capability, but local organisations may also possess their own political aims, territorial interests, sources of income and relationships with communities. A settlement involving an external sponsor may influence those actors without automatically controlling every decision they make. The forecast must leave room for that incomplete alignment.

The Red Sea adds another layer because civilian shipping can be affected by an armed conflict even when the ship, cargo owner, crew and destination are not belligerents. Commercial operators respond to uncertainty as well as damage. An insurer may change terms before a government announces a new policy. A carrier may prefer a longer route even after the immediate incident rate falls. An importing business may then experience delay without knowing which specific security decision caused it.

For analysis, the sea and land tracks should be recorded separately. One monthly entry should concern fighting, displacement and humanitarian access within Yemen. Another should concern actual attacks on commercial shipping, route availability and operators’ behaviour. A third should track diplomacy and its implementation. Progress on one track can be real even while another deteriorates. Collapsing them into a single red or green indicator makes the report less informative.

A useful restrained scenario would involve reduced fighting, accessible humanitarian routes, and a shipping-security arrangement that commercial operators can rely upon. It would not require every political disagreement to disappear immediately. An adverse scenario would involve continuing displacement, restricted aid access and recurring maritime incidents that prevent confidence from returning. A further adverse branch could connect local combat more tightly to interstate retaliation. These are scenarios, not claims that a particular sequence has already been chosen.

The civilian consequences are broader than the number of displaced people. A household that leaves its home may lose documents, income, transport, school access and routine healthcare at the same time. The journey to safety can remove it from the service network on which it depended. Receiving communities then need additional water, shelter, sanitation, learning space and medical support. Displacement is therefore both an outcome of conflict and a new load on institutions elsewhere.

This is not an argument that displaced people are themselves a security threat. They are people exposed to the conflict. A careful report distinguishes the burden on services from insinuations about the people who need them. The same distinction matters when discussing movement across the Red Sea: a humanitarian crossing and a military deployment are not interchangeable forms of mobility.

Repair evidence should focus on delivery rather than announcements. Are relief organisations reaching communities repeatedly? Can health facilities keep operating? Are transport permissions predictable? Is funding sufficient to sustain the promised programme? A successful delivery is meaningful, but it does not prove that the whole access problem is solved. Conversely, an interruption does not erase every functioning response mechanism.

The connection to the twelve-month outlook is persistence. A ceasefire that changes immediate violence but leaves the civilian economy and access system impaired may be difficult to sustain. A more durable arrangement would have to create practical reasons for people and institutions to keep using it. That proposition is not unique to Yemen, but Yemen makes the distinction between a signed pause and a functioning civilian environment particularly important to this report.

2.3 Russia–Ukraine: the continuation forecast and the settlement test

Assessment: substantial organised fighting through a significant part of the next year remains the qualitative base case. This is not a prediction that negotiations cannot succeed, and it is not a forecast of which side will obtain a particular territorial outcome.

September reporting documents continuing reciprocal attacks. Reuters reported Russian strikes and electricity disruption in Ukraine on 22 September, attributing local damage information to Ukrainian authorities. AP also reported attacks in both countries alongside diplomatic engagement. Those reports support the current-combat baseline, not a precise prediction of the next campaign. Reuters, syndicated by Investing.com; AP, 22 September.

The analytical reason to distinguish talks from settlement is the implementation problem. Parties can agree that a war is costly while disagreeing over what an acceptable cessation requires. They may also disagree over the reliability of future commitments. A promise to stop fighting is more consequential when there is a workable method of observing compliance, reporting violations and resolving disputes before retaliation begins again.

A monthly forecast should therefore follow the architecture of any proposal rather than only the tone of leaders’ statements. What activity would stop? Where would the arrangement apply? Which organisations would observe it? How would disputed incidents be handled? Would a local violation suspend the whole arrangement or activate a contained review? These questions concern implementation, not endorsement of any particular political settlement.

The counterargument to persistence is that a negotiated interruption can arrive faster than an analyst expects. An opening may become workable when parties reassess the costs of continuing, outside support changes, or a limited arrangement demonstrates that some commitments can hold. The report should not insist that only a comprehensive final treaty counts as improvement. A verified reduction in particular forms of violence can protect people even while larger disagreements remain unresolved.

At the same time, a partial arrangement should be described as partial. If attacks on one category of infrastructure decline but ground combat continues, that is a genuine change with a limited scope. It is neither meaningless nor the end of the war. The forecast ledger should contain fields precise enough to preserve this result without forcing it into a binary narrative of either total failure or peace.

Time is central to the civilian consequences. A damaged power network affects a household differently depending on season, housing, backup arrangements, medical needs and the duration of interruption. A family may tolerate a short outage while a clinic, refrigeration facility or water-pumping system reaches its limit much earlier. This is why the How War Works framework treats environment and sustainment as part of the system rather than scenery around it.

The report will not turn that observation into guidance on where to attack infrastructure. Its purpose is the opposite: to identify how civilian function can be lost and why preserving service continuity matters. The useful question is whether repair teams, spare parts, safe access and finance remain available, not which individual asset would be most vulnerable to an adversary.

A second clock concerns the economy. A business may keep operating despite physical damage, but repeated uncertainty can delay investment, reduce insurance availability or make customers seek alternatives. Recovery therefore does not begin only after the final shot. Decisions about equipment, staffing, credit and relocation are already shaping the future civilian environment during the conflict.

The strongest evidence against the persistence assessment would be a verified and durable reduction in organised fighting supported by implementation arrangements that survive incidents. The strongest evidence for persistence would be continued campaigns and failed implementation despite repeated negotiations. Simply scheduling another meeting is not enough to move the forecast substantially. Neither is a single day of reduced activity, particularly when reporting coverage or weather could explain the change.

For Singapore, the connection is not confined to distance from the battlefield. Energy markets, food trade, finance, maritime access and technology controls can transmit consequences beyond Europe. The strength of that transmission depends on specific contracts and dependencies. This report therefore does not predict a universal economic outcome from the continuation of war; it identifies channels that a Singapore-focused analysis should measure separately.

2.4 Sudan: why persistence includes the civilian economy

Assessment: continued organised violence and a prolonged civilian emergency remain the qualitative base case. A particular city’s change of control would not, by itself, resolve this continuation watch or establish national recovery.

Reuters’ 22 September reporting describes a sharp currency deterioration and pressure on ordinary households in divided Sudan. A July UN human-rights report, summarised by the UN country office, examined how control of territory, commodities and trade contributes to sustaining the conflict. The July source is background; the September report supplies a more recent economic observation. Reuters, 22 September; UN Sudan: dated reports and releases.

Sudan’s chapter needs a different lens from a maritime crisis. A war can persist because the institutions and livelihoods that would support a stable alternative have been damaged. If households cannot rely on salaries, banking, transport or markets, nominal access to a town does not restore normal life. If an armed organisation can obtain revenue within a conflict economy, military pressure alone may not remove every incentive for continued control.

This is a conditional mechanism, not a claim that all participants have identical financial motives. People may fight for political aims, security, coercion, survival or multiple reasons at once. The report should not invent private motivations from a revenue map. It can instead ask which material relationships allow organised violence to continue and whether proposed arrangements change those relationships.

A durable improvement would involve more than lower visible fighting around one location. Civilian movement would become safer and more predictable. Access to food and healthcare would improve. Workers would be able to receive payment. Humanitarian organisations would have repeated access rather than exceptional permissions. Local institutions would regain the ability to maintain services. Those are observable conditions that can support a reduced-conflict assessment without requiring speculation about personal intentions.

The alternative is fragmented stabilisation. Some places could become calmer while other areas experience renewed fighting or tighter restrictions. The countrywide narrative might then appear to improve even though particular populations become more exposed. A national average is poorly suited to that problem. The monthly ledger should preserve geographical differences and specify which communities and services a reported improvement actually reaches.

The civilian economy also changes how a forecast should interpret prices. A rising food price may reflect supply loss, transport danger, currency weakness, lack of competition or several of these together. A fall in the quoted local price could even coexist with worsening access if household income collapses faster. The analyst therefore needs both market availability and purchasing power. Food physically present in a market is not equivalent to food accessible to a family.

Recovery itself can produce new risks. People returning to an area need information about safe routes, buildings, water and remaining explosive hazards. A report of return movements is not proof that conditions are safe or that the movement was fully voluntary. The forecast should distinguish observed return from durable reintegration, just as it distinguishes a ceasefire announcement from verified implementation.

The strongest downward evidence for the continuation assessment would be sustained reductions in violence across relevant areas, reliable humanitarian access and practical arrangements that restore civilian livelihoods. The strongest upward evidence would be renewed campaigns, deeper fragmentation of control and repeated obstruction of essential services. Statements by outside governments deserve attention, but they are not substitutes for changes experienced by civilians on the ground.

Sudan should not be relegated to a footnote because its effects on international financial markets are less immediately visible than a major shipping crisis. This report uses a civilian-consequence frame as well as an international-system frame. A conflict can be globally under-covered and still impose enormous, lasting loss on people whose lives are not captured by the price of a traded commodity.

2.5 Myanmar: conflict, access and uneven local arrangements

Assessment: substantial organised armed conflict remains the qualitative base case, with the possibility of uneven local reductions rather than one uniform national trajectory. This chapter does not forecast an election outcome or claim that a single local arrangement represents the preferences of the whole population.

UNFPA’s 11 September release described humanitarian needs arising from conflict, displacement and recurring natural disasters in Myanmar, alongside an air-bridge delivery of medical and nutrition supplies. The release demonstrates a compound civilian emergency and an operating aid response. It does not establish uniform conditions across the country. UNFPA via UN Myanmar, 11 September.

The key analytical issue is the mismatch between a national label and local realities. A country may contain areas of active fighting, areas under negotiated arrangements, areas with restricted movement and areas where services continue unevenly. Treating the whole map as one battlefield can conceal both danger and repair. Treating a few calmer areas as proof of national stabilisation makes the opposite error.

The framework therefore asks how local agreements connect. An arrangement that stops fighting between particular organisations in one area may be important, but it can leave neighbouring communities exposed. It may also depend on conditions that are difficult to reproduce elsewhere. A forecast of national persistence can coexist with genuine local improvement. The report should record that improvement rather than dismissing it because it does not resolve the entire conflict.

For civilians, access is often the decisive intermediate variable. A clinic may exist but be unreachable. Supplies may arrive nationally without reaching a particular community. A school may have teachers but lack safe transport, learning materials or predictable attendance. A household may have relatives elsewhere who could help, but communications or movement restrictions can interrupt that support. These conditional pathways explain why a service map needs more than a list of surviving buildings.

Weather and disaster exposure can intensify these access problems without causing the conflict. A flooded road may obstruct relief regardless of which organisation controls the area. A storm can increase demand on medical services at the same time that conflict restricts movement. This is a compound-risk mechanism, not a claim that climate makes war inevitable. The report keeps environmental conditions, political decisions and organised violence as distinct fields.

The most useful repair indicators would be repeated access to essential services, reduced violence in specified locations, reliable local communication and arrangements that survive changes in conditions. A single aid flight is meaningful but cannot establish sustained last-mile access. A large funding announcement is also insufficient without evidence of delivery. Conversely, the persistence of conflict does not mean no humanitarian work succeeds.

A narrower adverse scenario would involve increased local fighting or restrictions that disrupt an already fragile service network. A broader adverse scenario would involve simultaneous deterioration across several areas, stretching humanitarian capacity and increasing displacement. A restrained scenario would involve multiple local arrangements that endure and allow civilian institutions to function more predictably. The report does not assume that these arrangements automatically combine into a national settlement.

For Southeast Asian readers, Myanmar is part of the regional civilian and economic environment. Cross-border movement, family connections, trade and humanitarian operations can transmit consequences. That observation must not turn refugees or migrant communities into proxies for armed organisations. People seeking safety and people directing military operations belong to different categories, even when a superficial geographical description places them in the same story.

Monthly review will focus on the distribution of violence and access rather than a dramatic but incomplete national narrative. Where evidence is missing, the correct state is an observation gap. It is not automatically low risk, and it is not permission to invent a worst-case account. The report must remain precise enough to recognise both a newly inaccessible community and a local service that has genuinely begun working again.

2.6 The central Sahel: persistence can spread without a declared interstate war

Assessment: continued severe insurgent conflict in the central Sahel remains a qualitative base case. The watch concerns organised violence, civilian access and possible geographical extension, not a prediction that every country in West Africa will enter the same condition.

Reuters reported on 21 September that sources described a major attack on Malian forces earlier in the month. Other reporting carried claims from the armed organisation that were not independently confirmed in full. The event belongs in the conflict baseline; disputed casualty totals should not be treated as a settled numerical measure of the whole regional situation. Reuters via SABC, 21 September; AP via Yahoo: claims and verification limits.

The central analytical question is how a conflict affects the ordinary routes through which a population lives. A road can remain physically intact while becoming too dangerous or unpredictable for regular commerce. A market can remain open while suppliers reduce deliveries. A school can stand undamaged while teachers or pupils cannot reach it. This is a form of functional loss that may not appear in a map showing only destroyed infrastructure.

The forecast should therefore track access and continuity as well as attacks. An increase in violence near a transport route may matter beyond the immediate location if it changes the movement of food, fuel, medical supplies or public services. However, the report must not infer control of an entire region from one incident. Temporary disruption, repeated intimidation, durable territorial control and administrative governance are different observations.

Another analytical trap is to turn regional similarity into a single causal explanation. Several countries can face related organisations or comparable pressures while differing in local institutions, community relationships, terrain and economic structure. A trend in one location may justify checking another, but it does not automatically transfer a probability. The warning system needs comparison cases, including places where similar pressures did not produce the same outcome.

A restrained scenario would involve improved civilian access, reduced attacks and functioning local arrangements that interrupt the persistence mechanism. An adverse scenario would involve the extension of disruption into additional routes or population centres. A different adverse branch would involve continued violence at a similar geographical scale but deeper economic isolation. That third possibility can be missed if the report treats expansion of territory as the only sign of deterioration.

The civilian consequence of persistence is cumulative. Repeated missed school terms are not equivalent to a single interruption followed by full recovery. Repeated market closures can deplete household savings and change livelihoods. Repeated displacement can weaken support networks. These are conditional mechanisms of long-duration conflict, not forecasts of a particular community’s behaviour or claims about any ethnic or religious population.

The strongest evidence of repair would be sustained changes experienced by civilians: reliable transport, repeated service delivery, reduced coercion and improved safety over time. A temporary military success may contribute to those changes, but it is not their equivalent. The report should ask what function has returned and to whom, rather than equating an official announcement with a completed civilian outcome.

The monthly watch also needs restraint in attribution. An armed group may claim an attack to advertise reach. An official account may be incomplete during an ongoing operation. A video may show a genuine event without proving its date or location. The appropriate response is to preserve the uncertainty, not choose the account that fits the most dramatic story. This is particularly important where independent access is limited.

2.7 Eastern DRC: verification is a repair mechanism, not a completed peace

Assessment: significant armed violence remains a qualitative continuation concern, while active ceasefire-verification work is a genuine counter-signal. The report does not assign a new numerical interstate-war probability for the Great Lakes region in this edition.

In mid-September, MONUSCO reaffirmed support for ceasefire monitoring and verification associated with the peace process in eastern DRC. The reported discussions concerned technical, logistical and security support for implementation. This establishes work on a mechanism; it does not establish that all relevant actors have complied or that civilian conditions have normalised. ACP, 17 September.

The mechanism matters because armed conflicts often suffer from disputed observation. One party reports a violation, another denies it, and retaliation can begin before a shared account exists. A verification arrangement can create a place where information is checked and disputes are handled. Its contribution is not merely symbolic. But it needs access, personnel, communication, authority and acceptance by those whose behaviour it is intended to observe.

The distinction between a mechanism on paper and a mechanism in operation should structure the monthly review. Has the team deployed? Can it reach relevant areas? Are reports produced? Do the parties engage with findings? Does a disputed incident lead to clarification rather than immediate escalation? These are practical questions. The existence of a commission name is not a substitute for answers.

At the same time, the report should avoid the opposite error of dismissing verification because it cannot instantly settle every issue. An arrangement that reduces uncertainty around particular incidents can protect people even while broader political disagreements continue. Partial improvement should be recorded at its actual scale. A well-functioning local mechanism is not national peace, but neither is it nothing.

Civilian return is another area requiring careful language. A movement of people back towards their homes may indicate improved access, exhaustion of displacement options, economic necessity or several factors together. It should not automatically be reported as proof that the area is safe. A stronger assessment would consider whether water, shelter, health services, documentation and livelihoods are available and whether people can remain without renewed displacement.

The analytical adverse scenario is not necessarily one spectacular collapse of negotiations. It may be gradual non-implementation: delays in deployment, inconsistent access, disputes over findings, or local violence that erodes confidence. The restrained scenario would involve a mechanism that survives these difficulties and produces repeatable improvements. The report should look for that operating history rather than demand a perfect agreement before acknowledging progress.

External relationships also matter, but the forecast must preserve attribution boundaries. Allegations of state support for armed organisations require their own evidence. The existence of cross-border interests does not prove command over every local action. The presence of regional actors does not remove the agency of communities and organisations within the country. A useful model can represent connections without pretending all decisions originate from one centre.

For the twelve-month outlook, the decisive question is whether peace arrangements begin changing the conditions that permit violence to recur. Verification can help, but it cannot substitute for every other component. Civilian security, local administration, economic access and credible implementation are connected but distinct. The monthly ledger will therefore record verified restraint alongside continuing violence rather than allowing either to erase the other.

2.8 Pakistan–Afghanistan: why a recurrent campaign is a different prediction from occupation

Forecast: a 45–65% provisional probability of further direct combat between Pakistani forces and Afghan de facto state forces on at least seven distinct days within thirty days. Assessment confidence: medium. This is a recurrence question. It is not a forecast of a successful occupation or the replacement of either government.

Reuters reported renewed Pakistani airstrikes on 21 September. Pakistan described militant targets; Afghan authorities condemned the strikes, and the UN reported civilian deaths. The competing accounts of targets and consequences must remain attributed. The incident demonstrates renewed cross-border military activity after relative calm, not proof that a sustained campaign meeting this register’s threshold has already begun inside the future window. Reuters, 21 September.

The analytical recurrence mechanism is a dispute about security responsibility. An attack within one country can be attributed to actors across the border. Retaliation may then be described as limited or necessary by the attacking side, while the other side treats it as a violation requiring a response. If the original security dispute remains unresolved, a reduction in fighting can become an interval rather than a settlement.

This mechanism does not require the analyst to decide every contested allegation in advance. It requires the report to distinguish the attack, the attribution claim, the response and the implementation of any restraint. If a particular allegation is not independently corroborated, it can still matter as a stated reason for action, but it must not be converted into a confirmed fact simply because a government repeated it.

The strongest reason not to place the probability higher is that both sides can face substantial costs from sustained confrontation. Border commerce, civilian movement, diplomatic relationships and the demands of other security problems can create incentives for containment. Mediation can also provide a route for commitments that are difficult to arrange directly. These restraints are not guarantees, but they are part of the forecast rather than afterthoughts.

The broad range straddles even odds because the report is more confident in the recurrence mechanism than in the exact duration of a future episode. A limited exchange could occur and remain below seven distinct combat days. A sequence of incidents could spread across the year without meeting the specified thirty-day threshold. Both outcomes would be consequential, but they would not automatically resolve the numerical question as a success.

A credible downward revision would follow sustained cessation, functioning arrangements for border incidents, and observable civilian normalisation such as predictable crossings. The report would need to check whether these changes reflect implementation rather than temporary absence of reporting. A credible upward revision would follow repeated reciprocal state-force activity and deterioration of the mechanisms that previously interrupted it.

Geography belongs in the explanation but not as destiny. A difficult border environment can complicate observation, movement and enforcement. It can also make extended campaigns costly. The same physical condition can therefore support opposite branches of the forecast: it may enable recurring insecurity while restraining ambitions for prolonged occupation. This is why a geographical description alone cannot determine the outcome.

The civilian questions concern more than the immediate locations of strikes. Crossings can connect families, labour markets, clinics and traders. Interruption may affect people who have no connection to armed activity. A report that focuses only on combatants can miss those dependencies. Equally, it should not treat everyone crossing a border as a security actor. The categories must remain separate.

For the monthly ledger, every new episode will be dated and coded according to what is actually corroborated. A militant attack is not direct state-to-state combat. A statement threatening retaliation is not an executed operation. A government claim of casualties is not an independently reconciled toll. Maintaining these distinctions makes the recurrence forecast testable instead of allowing it to absorb every event into its own justification.

2.9 Ethiopia–Eritrea: an interstate watch with strong competing explanations

Forecast: a 20–35% provisional probability of direct combat between Ethiopian and Eritrean regular forces meeting the seven-day threshold. Assessment confidence: low–medium. Proxy-only fighting does not resolve this question.

The current baseline combines a September development with an older restraint statement. Reuters reported on 21 September that seven Ethiopian armed groups announced an alliance against the federal government, while noting unresolved differences among participants. In March, Ethiopia’s prime minister publicly said the country would not wage war against neighbours over sea access. The first observation is current; the second is a dated statement that must be tested against subsequent conduct. Reuters via Yahoo, 21 September; Africanews, 4 March.

The forecast is concerned with the interaction between internal conflicts, interstate mistrust and disputes over access. None of these alone establishes a decision for direct war. An internal coalition can increase pressure on a government without becoming a unified military organisation. A commercial access dispute can generate bargaining without producing an invasion. An allegation of external support can affect decisions even while its factual basis remains contested.

The reason to include this front is that these issues can interact. A government might interpret internal violence partly through an external-security lens. A neighbouring state might interpret that interpretation as a threat. Public statements may then narrow the room for compromise. But the report does not claim to know that this chain will occur, or that any leader privately intends it. The probability reflects a plausible pathway, not a disclosed plan.

The strongest counterargument is that the same internal pressures can make interstate war less feasible. Resources committed elsewhere are not freely available for a new campaign. Economic costs can make negotiated access more attractive. Regional diplomacy can offer arrangements that separate commercial interests from sovereignty. A serious forecast must represent those brakes rather than assume pressure always converts into aggression.

This is also where language has exceptional importance. Access to a port, commercial transit rights, use of facilities and sovereignty over territory are different claims. A report should not silently replace one with another. Doing so could make an ordinary commercial negotiation look like an announced territorial demand, or make a territorial dispute look easier to settle than it is. Exact wording and actual commitments need to be preserved.

An upward revision would require corroborated evidence of direct interstate military confrontation or sustained changes in posture that fit a campaign more strongly than bargaining or routine activity. It would also depend on whether diplomatic restraint mechanisms were weakening. A downward revision would follow practical access arrangements, credible border restraint and evidence that internal conflicts were not being translated into interstate combat.

The forecast’s relatively broad uncertainty also reflects observability. Public reporting may be thin, delayed or based on interested sources. A confident-looking map cannot fill that gap. The report should label what it does not know, including the degree of coordination among armed organisations and the effectiveness of commitments made publicly. Unknowns should widen uncertainty, not automatically push the probability either up or down.

The civilian consequences of a new interstate front could be severe even if the campaign were geographically bounded. Displacement, interruption of commerce, pressure on already stretched services and delayed recovery from earlier fighting could overlap. Those are conditional consequences, not evidence that war has been chosen. Their importance explains why a below-even-odds event belongs on the watchlist without presenting it as inevitable.

The monthly question is therefore precise: are the conditions for direct regular-force combat strengthening, or are disputes being channelled into implemented arrangements that keep them below that threshold? A coalition announcement alone cannot answer it. Neither can a reassuring statement alone. Behaviour over time is the decisive return.

2.10 India–Pakistan: limited intent does not guarantee limited consequences

Forecast: a 20–35% provisional probability of direct state-force combat on at least seven distinct days within thirty days. Assessment confidence: low–medium. This threshold excludes a lone maritime collision, a militant attack by itself, and a shorter campaign that never meets the duration rule.

A mid-September collision between naval vessels produced protests from both countries. Pakistan and India offered different accounts of responsibility and location. The incident is therefore a verified point of friction with contested attribution, not an independently established order to begin a new campaign. Associated Press: naval collision.

The key forecast mechanism is not a shared intention to fight an unlimited war. It is the possibility that a triggering event leads one side to undertake what it considers a bounded response, while the other side interprets that response as requiring retaliation. Each step can be described internally as limited even as the combined sequence becomes harder to contain.

The words limited and proportional do not settle the recipient’s interpretation. An action intended to send a signal may be read as evidence of a broader campaign. A public commitment can make a compromise harder to communicate. An ambiguous incident can create pressure to act before attribution is complete. These are general escalation mechanisms; the report does not claim that any named official will necessarily respond in this way.

The strongest restraint is the potential cost of a widening conflict, including the risk associated with nuclear-armed states. Diplomatic intervention, direct communication and established incident arrangements can also interrupt the sequence. Their effectiveness must be evaluated through use, not mere existence. A communication channel that remains available during a crisis is more informative than a dormant agreement mentioned only after an incident.

The probability is below even odds because recurrence is not automatic. A serious incident may be investigated, contained or addressed through diplomacy. A limited exchange may end before becoming a sustained campaign. A year can contain considerable hostility without crossing this specific threshold. The report should not convert every deterioration in relations into a numerical upward revision without examining the restraining evidence.

Water and trade disputes require particular care. They can intensify political friction and civilian concern, but their presence does not prove that military action is planned. A treaty disagreement, a technical operating dispute, a supply interruption and an armed attack are distinct events. The analysis should examine the practical mechanism linking them rather than treating the word water as a prediction of war.

A stronger warning would involve convergent evidence: repeated direct encounters, worsening incident-management arrangements, and corroborated activity consistent with a sustained campaign. A weaker warning would be an isolated statement, a routine exercise without other changes, or a viral claim whose origin cannot be established. The report should assign evidential weight according to the observation’s ability to distinguish preparation from signalling.

The civilian consequence map should include disrupted crossings, transport, trade, education, family contact and health access. These can deteriorate before a sustained campaign begins. They can also remain impaired after fighting stops. An analytical forecast should not wait for the seven-day threshold before acknowledging their importance. The threshold governs one scored question, not the recognition of human harm.

A useful monthly review would ask whether any new episode stayed bounded and why. If crisis communication helped contain it, that is evidence relevant to future forecasts. If attribution remained unresolved but restraint held, the report should record that rather than assuming the absence of escalation was accidental. A warning system learns from successful containment as well as from conflict that actually occurs.

2.11 China–Philippines: a lethal encounter is not the same forecast as a war

Forecast: a 15–30% provisional probability of reciprocal weapons fire between uniformed Chinese and Philippine state forces causing at least one corroborated death. Assessment confidence: low–medium. This is a lethal-confrontation question, not a prediction of sustained interstate war.

The September baseline is a collision, not a shooting. Reuters reported damage to a Philippine fisheries vessel during an encounter on 18 September, with no injuries reported. Philippine authorities accused the Chinese coast guard of ramming; China’s coast guard blamed the Philippine vessel’s manoeuvre. Both accounts belong in the record. Reuters via Internazionale, 19 September.

The analytical concern is repeated physical interaction under contested rules. Close encounters can create opportunities for an incident even when neither government has decided to start a sustained war. Different interpretations of lawful presence and acceptable behaviour can make the same manoeuvre appear routine to one side and threatening to the other. A forecast must examine that interaction without assuming that a collision proves an intention to kill.

The numerical question is deliberately narrower than a general crisis. It requires weapons fire in an exchange and a corroborated death. A future collision with fatalities would be extremely serious but would not automatically resolve this particular forecast. Nor would an unrelated shooting by private individuals. Those events would be recorded in the severity and interpretation ledger with their own definitions. The report does not get to broaden the question after the fact to claim success.

The base case remains coercion and confrontation below a sustained-war threshold. That does not make the situation comfortable for crews or fishing communities. It means the report judges that the incentives and mechanisms for containing encounters are still significant. The danger is that an event produces casualties, the public meaning of the incident becomes difficult to reverse, and subsequent responses create a larger confrontation than initially intended.

The strongest warning would be evidence that encounter-management practices were failing repeatedly and that actual weapons use, rather than nonlethal coercion alone, was becoming part of the interaction. A single dramatic video is insufficient without verification of date, location, sequence and context. Footage can show damage while leaving the preceding manoeuvres unclear. An analyst should not claim omniscience from one camera angle.

The strongest downward evidence would be implemented encounter protocols, reduced frequency of dangerous interactions and a record of disputes being handled without escalation. Public statements supporting restraint matter more when behaviour reflects them. A diplomatic agreement that crews understand and use is more consequential than an agreement that remains remote from actual encounters.

External involvement is a separate branch. A lethal incident could lead to consultation, investigation, additional deployments, diplomatic pressure or direct military involvement. None of these should be assumed automatically. Treating an alliance relationship as a mechanical switch erases the decisions, legal interpretations and crisis-management steps that would determine the response. The forecast should preserve those branches rather than jumping from one casualty to a general Pacific war.

The civilian consequence map includes livelihoods and access before any wider campaign. Fishing communities may face uncertainty about where they can work. Government service vessels may experience disruption. Commercial operators may adjust risk assessments if the situation broadens. However, a localised encounter does not establish that the entire South China Sea is closed to trade. Geographical scale must remain explicit.

For Singapore, this is a close regional watch rather than a forecast of automatic military participation. The relevant questions concern actual route disruption, diplomatic developments, travel decisions and commercial exposure. A careful monthly report will distinguish a local incident from a regionwide interruption and will show the evidence required to move from one assessment to the other.

2.12 Taiwan: separate pressure, commercial interdiction and main-island invasion

Forecasts: a 10–20% provisional probability of a broad, actually enforced inspection, quarantine or blockade regime materially obstructing commercial access to Taiwan proper for at least seven consecutive days; and a separate 5–10% probability of a large-scale landing campaign intended to seize and hold substantial territory on Taiwan’s main island. Confidence: low–medium for interdiction and low for invasion.

A crucial dated source is the March 2026 U.S. intelligence assessment, reported by Reuters and USNI News. It assessed at that time that Chinese leaders did not have a fixed unification timetable or a current plan to invade in 2027. Taiwan’s representatives stressed continued vigilance. This is evidence against treating a capability milestone as a scheduled attack, not an assurance that intentions cannot change. Reuters, 18 March; USNI News, 19 March.

The report’s base case is continued pressure short of a main-island invasion. This is a comparative scenario judgment. It does not imply that every form of pressure is harmless, lawful or economically minor. A coercive commercial restriction could create substantial disruption without a landing campaign. Political influence, military signalling, cyber activity and commercial measures can also occur without meeting either numerical threshold.

The first distinction is between an exercise and enforced interdiction. An exercise notice may create temporary uncertainty or cause operators to reroute. An enforced regime requires actual implementation that materially obstructs access over the specified period. The report will seek evidence from multiple commercial operators, relevant authorities and independent reporting. It will not resolve the question from a map of declared zones alone.

The second distinction is between interdiction and invasion. A landing campaign involves a different combination of objectives, commitment and sustained activity. It cannot be inferred simply because a military possesses relevant equipment or conducts training. Capability is an important part of the conditions for an operation, but it is not the decision itself. The report must avoid making the year 2027 do the work that evidence should do.

The third distinction is between an attempt and a successful outcome. The invasion question concerns a qualifying campaign being launched, not whether it succeeds or how much territory is ultimately controlled. Forecasting success would require a different model and carries additional uncertainty. This report deliberately does not provide operational plans, preferred routes, target recommendations or a tactical assessment of how to conduct such a campaign.

Why is interdiction assessed above invasion? The analytical judgment is that a coercive access regime could be attempted without the same immediate commitment to a large-scale territorial campaign. That does not make it easy to sustain or safe from escalation. Commercial resistance, legal disputes, international responses and incidents at sea could rapidly change its costs. The report treats it as a distinct possible transition, not a guaranteed intermediate step on a ladder to invasion.

A ladder metaphor can itself mislead. Events do not have to pass through neat stages. Pressure can intensify and then recede. A crisis can produce a negotiated pause. An incident can create an unexpected branch. A large campaign could be attempted without a prolonged, publicly visible interdiction phase. The two probabilities are therefore not simply consecutive steps whose likelihoods can be multiplied.

The strongest reason for keeping the invasion forecast low is the scale and uncertainty of the commitment, together with the existence of alternatives to immediate war. That is an analytical restraint argument, not an assertion that costs always deter. Decision-makers can interpret costs differently or believe that future conditions will be less favourable. The report cannot read those private calculations; it can only test public evidence against competing explanations.

The strongest reason not to set the probability at zero is that capability, political disagreement and the possibility of a crisis remain relevant. An event with a low probability can still occur. The purpose of the range is to acknowledge uncertainty without converting a high-consequence possibility into the base case. Calling the outcome inevitable would be unsupported; calling it impossible would be equally careless.

A material upward revision would require several independent indicators of sustained campaign-scale preparation and political commitment that fit an imminent operation better than routine training or bargaining. A material downward revision would follow durable reductions in coercive activity and implemented arrangements that reduce the likelihood of a dangerous incident. A reassuring speech alone is not enough, just as an alarming speech alone is not enough.

The economic consequence analysis has a separate evidence base. Taiwan’s economy ministry reported record August export orders driven by AI demand, according to Reuters on 22 September. This is a current indication of commercial importance, not evidence that a war decision has become more likely. Reuters: Taiwan export orders.

A disruption affecting such a connected economy could travel through manufacturing schedules, inventories, finance and customer expectations. The consequence would depend on the duration and scope of interruption, substitutes, existing stocks and the ability of firms to adapt. It is therefore wrong to assume either that a short crisis has no global effect or that any incident stops all technology production. The report should follow actual output and delivery data rather than treating a headline as a complete supply-chain model.

For families and schools, the relevant lesson is also linguistic. The words pressure, drill, quarantine, blockade, invasion and war should not be used interchangeably. A reader who can keep those categories distinct is less vulnerable to sensational headlines and more capable of recognising a real change. That is a practical contribution of the vocabulary discipline behind this forecast.

2.13 Russia–NATO: incidents, attribution and the threshold for sustained direct combat

Forecast: a 5–15% provisional probability of sustained reciprocal direct combat between Russian regular forces and the forces of at least one NATO member, meeting the seven-day threshold. Assessment confidence: low. Support for Ukraine, an unattributed sabotage incident or a single intercepted drone does not resolve the question.

In September, an Italian NATO aircraft intercepted a drone in Lithuanian airspace. Officials described a likely origin while examination and attribution remained part of the investigation. NATO’s public eastern-flank material also records an expanded vigilance posture. The incident and the posture are observable; neither alone proves a decision to begin sustained direct war. Reuters via Defense News, 15 September; NATO: eastern flank.

The analytical risk is an escalation sequence in which an incident, attribution and response become tightly coupled under time pressure. The same event may be interpreted as an accident, spillover, a deliberate probe or an attack. Those interpretations can imply different responses. A forecasting system should therefore track the evidence for attribution rather than adopting the most consequential interpretation immediately.

The report’s base case is continued dangerous incidents and heightened readiness without sustained direct Russia–NATO combat. That judgment leaves room for serious harm below the numerical threshold. A single incident involving deaths could be consequential even if it is contained. Repeated nonlethal incidents could impose costs and create anxiety without becoming the defined campaign. These outcomes need their own descriptions rather than being forced into the binary war question.

The strongest restraint is the potentially extreme cost of uncontrolled widening, together with mechanisms intended to clarify incidents and maintain deterrence. However, deterrence is not a measured guarantee. It depends on how intentions and capabilities are interpreted. A posture designed to reassure one population can be perceived differently elsewhere. The report should describe those competing interpretations without pretending to know which private belief dominates every decision.

An upward revision would require clearer evidence of intentional, reciprocal state-force combat and deterioration of the channels that could contain it. A downward revision would follow a sustained record of incidents being resolved without escalation and practical improvements in communication or restraint. The absence of an incident during one quiet week is weaker evidence than the successful handling of a serious test.

A further distinction concerns military participation. A citizen of a NATO country fighting under another organisation’s command does not automatically make that person’s home state a direct belligerent for this register. Nor does supplying equipment by itself satisfy the specified direct-combat event. Attribution must follow command and action, not nationality alone. This avoids expanding the forecast definition whenever the news supplies a convenient association.

The civilian consequence map is wider than a battlefield. Energy security, transport, communications, insurance and public confidence can be affected by persistent tension. Defensive investment can improve resilience while also consuming resources. Those consequences belong in the system map, but they are not independent proof that a sustained war will occur. Preparation for a scenario is evidence that institutions consider it consequential, not a disclosed probability of the scenario.

The range remains broad because the decisive transition could depend on a small number of future incidents and decisions. That is a reason for careful monitoring, not a licence for dramatic certainty. The monthly report will retain the distinction between a warning about escalation and a claim that escalation has already become inevitable.

2.14 The Korean Peninsula: capability is not a launch decision

Assessment: continued weapons development, coercive signalling and possible limited confrontation are more plausible in this judgment than a general peninsula war. No numerical probability for a full-scale Korean war is issued in this edition.

Reuters reported on 18 September that North Korea rejected an IAEA resolution and reaffirmed its nuclear status. This establishes a public position and continued strategic emphasis. It does not establish a decision to initiate a general war. Reuters, 18 September.

The analytical distinction is between accumulating options and choosing to use them. A weapons programme can support deterrence, coercion, bargaining or preparation for conflict. Its existence is relevant to the consequences of a crisis, but it cannot by itself determine which future choice will be made. A test can demonstrate a capability while leaving its reliability and intended use uncertain.

The more informative warning would be a combination of actual incidents, changes in sustained readiness and weakening crisis-management mechanisms. A statement, parade or test should be examined in context rather than automatically interpreted as a countdown. Conversely, long familiarity with threatening language should not make the analyst dismiss a genuinely different operational state when corroborated evidence appears.

A restrained branch could involve continued disagreement and capability development without direct war. An adverse branch could involve a local encounter that becomes difficult to contain. A much more severe branch would involve sustained general combat. These branches differ in probability and consequence. The report does not collapse them into one headline about the peninsula being either safe or on the brink.

Civilian preparedness should remain grounded in official local guidance and continuity needs rather than this report’s qualitative assessment alone. The monthly ledger will track concrete changes and preserve uncertainty. It will not treat leadership psychology, rumours about personal health or unsupported claims of irrationality as an analytical shortcut.

2.15 South Sudan: a political decision window, not an election prediction

Assessment: renewed internal violence remains a qualitative watch where security arrangements and political bargaining are under strain. This report does not predict electoral winners, assess electability or recommend a political choice.

Reuters reported on 22 September that changes to election law removed requirements linking elections to a permanent constitution and census. The government described the amendments as facilitating preparations; the opposition coalition objected to the process. These are documented positions concerning a political timetable, not evidence that violence is predetermined. Reuters via Polity, 22 September.

The conflict-analysis question is whether disagreements remain within functioning arrangements or become connected to armed mobilisation and violence. An election date can concentrate decisions, expectations and disputes, but it does not mechanically cause war. The report should examine observable security behaviour, civilian access and the operation of dispute-resolution channels rather than treat the calendar as an explanation.

A useful downward signal would be implemented agreements that allow disputes to be handled without armed coercion. A useful upward signal would be corroborated security fragmentation or organised violence, not simply criticism by a political opponent. Political disagreement is not itself a conflict event. Preserving that distinction protects both analytical accuracy and the ordinary right to disagree.

The civilian consequence map includes movement, markets, schooling and access to services during any period of uncertainty. Temporary interruption can become damaging when households have little capacity to absorb it. But the report will not infer a nationwide emergency from isolated incidents or assume that all communities experience the same conditions. The monthly watch remains geographical and evidence-specific.

2.16 Thailand–Cambodia: a concrete diplomatic counter-signal

Assessment: renewed localised clashes remain a watch, but sustained broad war is not the base case of this report. No numerical probability is issued here because the current pass does not establish a sufficiently developed, independently reviewed estimate for that event.

The Permanent Court of Arbitration confirmed that the first meeting of the Cambodia–Thailand conciliation process took place in Singapore from 14 to 16 September. The proceedings concern the maritime dispute under UNCLOS. That is an implemented diplomatic step, not merely a proposed meeting. It does not automatically resolve separate land-border disagreements. PCA, 16 September.

This case matters because a warning system can become biased towards deterioration. If every clash raises concern but a functioning conciliation process is dismissed as irrelevant, the system has no fair way to recognise restraint. The appropriate response is more precise: the maritime dispute has an operating process, while other issues need separate observation. Improvement is real at its actual scope.

The monthly questions concern implementation. Do meetings continue? Are procedural commitments honoured? Does the process create clearer information about disagreements? Are unrelated incidents kept from overwhelming it? None of these requires the report to choose a winner in the territorial dispute. They concern the capacity to manage disagreement without force.

A regional peace process can also have civilian value before a final settlement. Predictable communication may reduce uncertainty for commercial and local actors. The extent of that benefit would need evidence; it should not be assumed from ceremony alone. The report therefore records both the verified procedural step and the remaining uncertainty about consequences.

2.17 Gaza and the civilian meaning of a partial return

This chapter is a recovery and relapse watch, not a new numerical onset forecast. Reuters reported in September that children in Gaza had returned to in-person learning in damaged or makeshift settings after a prolonged interruption. The return of teaching is a concrete civilian function; it does not establish full reconstruction, safety everywhere or the end of the political conflict. Reuters via TBS, 21 September.

The analytical lesson is that peace cannot be measured only by the absence of a new battle. A child needs a place to learn, materials, a teacher, food, safe movement and continuity. A classroom can reopen while several of these remain precarious. Recording the reopened classroom matters, but so does recording whether learning can continue next week and next month.

The adverse branch is renewed violence or restrictions that interrupt emerging civilian routines. The restrained branch is a sequence of practical improvements that makes those routines more dependable. The report should observe both without assuming that the first improvement guarantees all later steps. A partial return is neither a complete settlement nor a meaningless gesture.

For an educational publication, this is central rather than decorative. War destroys time as well as buildings. Missed learning can affect later choices, work and recovery. The exact consequences differ among children and cannot be reduced to a universal number, but the mechanism is clear: repeated interruption removes opportunities that a normal calendar would otherwise provide. The monthly report should keep those opportunities visible when attention returns to diplomatic statements or military maps.

2.18 Secondary fronts, omitted probabilities and the limits of the board

A global report is not complete merely because it contains many country names. Some potential fronts require additional evidence before a numerical forecast would be defensible. The Americas, the South Caucasus, other African conflicts and additional maritime disputes remain part of the wider observation field. Their absence from the eight-event table means no numerical estimate is issued here, not that the places are risk-free.

The correct way to expand the board is to define a new event before gathering a convenient set of headlines. A possible external military intervention, a renewed civil conflict and a territorial incursion require different questions. Adding them to a list without a clear actor, geography, window and resolution rule would create the appearance of coverage without a useful forecast.

This boundary also protects against hindsight. If an event occurs in a place not included, the report must record it as outside the original register. It should not claim that a broad phrase such as regional instability predicted the outcome. A new entry can be created for subsequent developments, but it must carry its own date and acknowledge the miss or prior lack of coverage.

The same rule applies to severity. An event outside the numerical table can be more harmful than one inside it. The table is a set of testable questions, not a ranking of whose lives matter. The broader monthly report must continue following civilian harm and repair even when those observations do not resolve one of the eight forecast events.

3. System Map

3.1 Geopolitics and security: connected does not mean centrally controlled

The regional forecasts are connected through diplomacy, commerce, military relationships and perceptions of what other actors will do. But a connection is not the same thing as a single chain of command. A state can cooperate with another on one issue, disagree on another, and avoid direct participation in its war. The March ODNI assessment’s description of selective bilateral cooperation is useful precisely because it resists a simple two-bloc picture. Its date remains part of the evidence boundary.

For the forecast, connections can amplify or restrain. An actor may believe that another crisis distracts a potential opponent, increasing perceived opportunity. The same crisis may also make governments more cautious because resources are committed elsewhere or because commercial costs are already high. The report cannot assume only the first mechanism operates. Both belong in the scenario tree.

A hypothetical example makes the distinction clearer. Suppose two unrelated maritime crises occur in the same month. Shipping costs might rise across both regions because operators reassess risk and vessel availability. That economic coupling does not prove that the military actors coordinated their actions. It establishes a shared consequence, not shared intent. A good system map labels the edge correctly: commercial transmission rather than joint command.

The same discipline applies to alliances. A security relationship can shape expectations without determining every future response. Consultation, diplomacy, additional readiness and direct combat are distinct possible outcomes. The numerical register counts the last only when its defined conditions are met. It does not assume that every external relationship converts automatically into belligerency.

The useful geopolitical map therefore contains different kinds of links: command, material support, commercial dependence, legal commitment, diplomatic influence and shared exposure. Treating them as one generic connection would make the diagram visually impressive and analytically weak. The report’s purpose is to identify which link could transmit which consequence, and what observation would show that transmission has actually begun.

3.2 Economy and markets: the route from a distant event to a household budget

War can affect an economy through physical damage, transport disruption, uncertainty, public spending and the cost of finance. These channels do not always move together. An asset price can recover while a delivery route remains constrained. A currency can weaken for reasons unrelated to the conflict. A rise in insurance cost can precede a physical interruption. The analyst needs a chain of evidence, not a convenient story attached to every market movement.

Consider an explicitly hypothetical Singapore importer. It normally receives a component every thirty days and holds ten days of usable stock. A route disruption adds twelve days to the next delivery. The business does not immediately become insolvent, and the entire national economy does not stop. Its first problem is a two-day gap if no alternative stock exists. That gap may lead to delayed production, expedited freight, a substitute component or a changed customer schedule.

Now add a second constraint. The business must pay its supplier before it receives payment from customers. A longer voyage ties up cash for longer. Even if the goods eventually arrive intact, the firm may need additional working capital. If credit is expensive or unavailable, the logistical interruption becomes a financial problem. This example is not a prediction of actual delivery times. It demonstrates why a physical route, an inventory policy and a financing arrangement belong in the same consequence map.

A larger firm with several suppliers may absorb the interruption more easily. A smaller firm with one specialist input may face a sharper effect. A household connected to either firm through employment can therefore experience the same international event differently. A national average may conceal that distribution. The report should not claim a uniform consequence simply because a global price moved.

The counter-signal is adaptation. Substitution, stock management, contract changes and alternative routes can reduce the effect. They also have costs. A replacement supplier may be less efficient; extra inventory ties up funds; expedited transport may increase expense. The correct conclusion is not that adaptation eliminates harm or that every adjustment proves crisis. It is that the economy changes the route through which the shock is absorbed.

For monthly review, the relevant evidence would include actual delivery performance, freight and insurance terms, production interruptions and financing conditions. Market prices remain useful, but they should be treated as one observation among several. A price is a representation of expectations and transactions, not a direct sensor of every physical system on which the economy depends.

3.3 Energy: availability, price and redundancy must be separated

EMA states that about 95% of Singapore’s electricity is generated using imported natural gas. Its consumer explanation distinguishes pipeline supplies from Malaysia and Indonesia from LNG obtained through global markets. It also explains that prolonged global disruption can affect prices and uncertainty. This is a structural exposure statement, not evidence that all Singapore gas travels through the Gulf. EMA: how Singapore’s energy supply affects consumers.

Three separate questions follow. Is the required fuel physically available? What price must be paid for it? Can the electricity system convert that fuel into reliable power? A favourable answer to one does not guarantee favourable answers to the others. Fuel can remain available at a higher price. Prices can stabilise while a particular delivery is delayed. A local technical fault can interrupt electricity without a global fuel shortage.

A war forecast should therefore avoid the shortcut from higher oil prices to an impending Singapore blackout. That conclusion requires evidence about the actual generation system, fuel arrangements, reserves, equipment and operating conditions. Equally, continued electricity supply does not prove that households and firms are unaffected by global energy costs. Physical continuity and affordability are different outcomes.

Redundancy also has several forms. A system may diversify suppliers, maintain reserves, improve efficiency or add alternative generation. These measures operate on different timescales and cannot all substitute instantly for one another. A long-term investment can improve future resilience without solving a delivery problem this week. A reserve can bridge an interruption but may not remove exposure to prolonged high prices.

The analytical question is the time available before a constraint becomes consequential. A short disruption may be absorbed by existing buffers. A longer one may require changed contracts or consumption. A still longer one may alter investment decisions. The report should track duration, not merely whether a disruption has been announced. Persistence changes the type of response needed.

The energy chapter is consequently a conditional exposure analysis. It does not predict a particular electricity tariff, advise a financial trade, or claim knowledge of confidential inventories. For Singapore readers, the most useful return is a clearer distinction between global market pressure, physical supply and local service reliability.

3.4 Food and agriculture: a harvest is only one part of food security

Food-system exposure can arise from farming conditions, labour availability, fertiliser, fuel, transport, storage, trade rules and household income. War can interact with several of these, but a credible report must identify the particular pathway. A failed harvest, a blocked road, a currency shock and a loss of wages are not the same event even if all can reduce access to food.

An explicitly hypothetical grain-importing country illustrates the difference. Suppose its supplier produces a normal harvest, but shipping becomes slower and more expensive. The physical crop is available; the problem lies in transport and cost. If the importing country has stocks and financing, it may absorb the delay. If households have little purchasing power, a smaller retail increase may still have serious consequences. The harvest alone does not settle the outcome.

Now reverse the example. A local harvest is poor, but imports, stocks and assistance compensate. There is real agricultural loss without a national food shortage. Farmers may still suffer income damage even when urban consumers can buy food. A report that records only aggregate supply would miss the livelihood loss. A report that calls the entire country food-insecure from one damaged region would overstate the inference.

Timing matters because agriculture has biological and logistical clocks. A delayed input can affect a planting decision long before the harvest. A missed transport window can affect storage quality. A household may sell productive assets to manage a temporary shortage, making recovery harder later. These are conditional mechanisms of persistence; they are not claims that every conflict will produce the same sequence.

The repair field includes functioning markets, safe transport, storage, substitute suppliers and support targeted to people who cannot afford available food. Each has its own evidence requirement. Announcing an import agreement is not the same as delivered cargo. Delivered cargo is not the same as affordable food at household level. A subsidy may help access without resolving a transport bottleneck.

The monthly report should therefore ask two parallel questions: what is happening to food availability, and what is happening to people’s ability to obtain it? The first is a supply question. The second includes income, prices, access and distribution. Neither should be used as a proxy for the other.

3.5 Water, climate and weather: conditions change options, not destiny

The IPCC’s Sixth Assessment Report describes a conditional and comparatively weak relationship between climate variability and armed conflict relative to other socioeconomic drivers, with stronger evidence for lower-intensity violence than for major war. It does not support a simple claim that climate change determines a particular conflict. This is a research background finding, not a current forecast for any one country. IPCC AR6 Working Group II, Chapter 16.

This distinction is central to the connection with How War Works: weather, geography and environment. Environment can affect movement, supply, civilian need, repair access and the cost of sustaining activity. It does not make political decisions by itself. A difficult season can restrain an operation, intensify civilian suffering, complicate diplomacy or create several effects at once.

Imagine two hypothetical regions experiencing similar drought. In one, reliable trade and water distribution reduce household pressure. In the other, conflict interrupts transport and public services. The same rainfall anomaly can produce different outcomes because institutions, access and resources differ. It would be inaccurate to attribute the difference to weather alone. It would also be inaccurate to ignore weather because political conditions matter more.

Weather forecasts and war forecasts operate on different evidential foundations. A forecast of rainfall concerns a physical process with its own models and uncertainty. A forecast of military action concerns decisions by actors who can adapt. The two can inform one another, but their probabilities cannot simply be multiplied without a defensible model linking them. A seasonal outlook is not an invasion calendar.

Geography introduces another kind of constraint. Distance, terrain and access can affect the cost of moving people and goods. The same constraint can limit an attacker, complicate civilian escape and delay relief. That shared exposure is why geography belongs inside the consequence map. The report does not translate it into operational advice about routes or targets. It uses it to explain why a damaged connection may be difficult to replace quickly.

Water systems are especially important because they connect physical infrastructure, energy, chemicals, maintenance, labour and public health. A treatment plant can survive while lacking a necessary input. A distribution network can operate while failing to reach a displaced community. A reservoir level can improve while water quality remains unsafe. The system needs several observations, not one reassuring photograph.

The monthly warning should therefore distinguish the physical condition from the political and institutional response. Is there a verified environmental stress? Which functions are exposed? What buffers exist? Are they actually reaching the affected population? Is conflict obstructing their operation? Only after those questions are answered should the report make a claim about compound risk.

3.6 Health: the service chain matters as much as the hospital building

The ICRC’s work on urban services during protracted conflict emphasises the cumulative damage to electricity, water and sanitation, not only the immediate destruction visible after an attack. Its 2015 report is historical institutional research and is used here as a mechanism reference, not as a current count of people affected. ICRC: urban services during protracted armed conflict.

A hypothetical hospital shows the practical connection. Its walls, wards and staff may remain in place, but unreliable electricity can affect refrigeration and equipment. Interrupted water supply can affect hygiene. Unsafe transport can delay patients and staff. A finance problem can interrupt procurement. The hospital’s nominal existence therefore cannot stand in for the availability of care.

The relevant unit is a service chain: a person recognises a need, reaches care, receives assessment, obtains treatment and can continue necessary follow-up. War can interrupt any link. The result may be delayed care rather than a dramatic building collapse. A forecast concerned only with physical destruction will underestimate this kind of loss.

Health consequences also vary by vulnerability and duration. A brief interruption may be manageable for one person and dangerous for another who needs continuous treatment. Displacement can separate people from records, medicines and familiar support. The report should describe these mechanisms without offering personalised clinical advice or assuming that every exposed population experiences the same outcome.

Repair evidence needs the same discipline as escalation evidence. A delivery of supplies is a useful input. Whether those supplies reach functioning facilities and improve access is a separate outcome. A new plan can strengthen coordination, but it does not itself demonstrate reduced disease or improved survival. The monthly report should retain both fields so that the response is neither dismissed nor overstated.

This is also why humanitarian data should not be treated as a military-intent sensor. Rising medical need can indicate worsening civilian exposure without identifying who ordered a particular action or predicting the next campaign. The data are important in their own right, not merely as clues for strategic analysis.

3.7 Infrastructure and logistics: longer routes consume capacity

UNCTAD’s 2025 maritime review documented how rerouting and longer voyages increased transport work and costs. That report covers an earlier period and is used here to explain the mechanism, not to assert the current volume moving through any strait. The physical lesson is that a route can remain available while requiring more time and resources to carry the same cargo. UNCTAD, 24 September 2025.

Consider a simplified hypothetical fleet whose ships each complete a round trip in thirty days. If rerouting increases the round trip to forty days, the same fleet completes three quarters as many trips per unit of time, assuming everything else is unchanged. That is a 25% reduction in trip capacity. No ship has been destroyed. The constraint arises because time has been consumed.

Real shipping is more complex: port waiting, cargo type, speed, contracts and repositioning matter. The example is not a forecast of freight rates. It demonstrates why a distant security problem can create a capacity effect even when alternative routes exist. Redundancy preserves a function but may not preserve its original cost, speed or volume.

A similar mechanism applies on land. A damaged bridge may be bypassed, but the detour can reduce the number of deliveries a vehicle completes. An alternative airport can remain open while lacking sufficient handling capacity. A replacement warehouse can exist but be too far from customers for the previous schedule. Calling a system resilient requires asking which function survives and at what performance level.

Logistics also has a knowledge layer. Operators need reliable information about access, restrictions and timing. Conflicting reports can cause delays even when physical movement is possible. Better information does not remove the hazard, but it can reduce avoidable uncertainty. The report should therefore track authoritative notices and actual operator behaviour separately from commentary.

The monthly signal is not simply whether an alternative route has been announced. It is whether the alternative carries the required function repeatedly. That means delivered goods, predictable timing and usable access for the people who depend on it. A map showing two lines is not proof that the second line can absorb the first line’s load.

3.8 Technology, AI and cyber: speed can improve response and accelerate error

The use of digital systems introduces both capability and dependence. Faster information processing can help people identify patterns, compare reports and coordinate repair. It can also distribute an error quickly or encourage decisions before verification is complete. The report does not assume that technology makes conflict either inevitable or controllable by itself.

An AI-generated summary is a representation of sources, not an independent witness. If several summaries derive from one report, they remain one evidence family. If an automated system assigns a label such as imminent, the label needs an inspectable basis. The presence of sophisticated software does not remove the need to distinguish observation from inference.

Cyber incidents require their own vocabulary. Attempted access, successful intrusion, data theft, service interruption and physical consequences are different outcomes. A suspected intrusion is not automatically an act of war. A service outage is not automatically caused by a hostile state. Attribution and legal classification require evidence beyond the fact that a system stopped working.

For the twelve-month outlook, the most important technological question is how dependencies alter the consequences of a crisis. If an organisation relies on one communications service, a disruption may impede coordination. If it maintains tested alternatives and usable records, some functions may continue. The same principle applies to suppliers, software, expertise and specialised equipment.

The report does not provide attack methods or vulnerability maps. Its civilian focus is continuity and verification. Can people recognise an unreliable message? Can an organisation operate essential functions during a digital interruption? Can it distinguish a temporary technical fault from a confirmed security event? Can a correction reach the same audience that received the original error?

Those questions also govern this publication. Its forecasts are human-readable judgments with sources and limitations. They are not made more reliable by presenting them as the output of an unseen machine. A future revision must be justified by evidence and reasoning, not by an unexplained change in an automated score.

3.9 Governance and trust: uncertainty needs a place to go

A crisis produces claims before it produces a settled account. Institutions need ways to receive those claims, investigate them, communicate what is known and correct errors. A source can be authoritative about its own action while incomplete about an adversary’s intention. A government statement is therefore valuable evidence, but its scope must be kept clear.

Verification mechanisms can provide a place for disputed observations to be handled without immediately becoming retaliation. Public communication can also reduce confusion by separating confirmed facts from information still under review. These are general functions, not a ranking of political systems or an endorsement of a particular government.

Trust should not be treated as blind acceptance. A credible institution may say that a report remains unconfirmed, explain why, and later revise its account. The correction is part of the evidence history. An analyst who treats every correction as proof that all information is worthless makes careful reporting harder. An analyst who ignores corrections makes the ledger inaccurate.

The monthly report needs the same discipline. If an early account of an incident changes, the source record should show the revision. If a forecast probability changes, the reason should be stated. If evidence remains inaccessible, the report should not replace it with confident prose. Uncertainty is a field to preserve, not a defect to hide.

The governance consequence of conflict also includes ordinary administration: records, payments, permits, public services and access to information. These can deteriorate without a formal collapse. Conversely, some functions can recover while political disagreement remains unresolved. A useful system map is detailed enough to record those mixed outcomes without turning them into a simple verdict about who governs well.

3.10 Education, work and society: the loss of ordinary time

UNICEF reported on 16 September that renewed fighting in Yemen had disrupted health facilities and suspended learning in hundreds of schools. The release connects current hostilities to children’s access to essential services. It is a dated observation, not a prediction that every school in the country will remain closed for the next year. UNICEF Yemen, 16 September.

Education belongs in a war forecast because conflict changes the conditions under which learning occurs. Attendance depends on transport, safety, household income, health and available adults. A child may be physically uninjured yet lose access to a stable learning routine. A teacher may be present but unable to obtain materials or communicate with displaced families.

The consequence is not identical for every child. Some have support, continuity and opportunities to recover. Others face repeated interruption. The report should not invent a universal learning-loss number from a conflict headline. It can explain the mechanisms and identify the services that help preserve continuity: safe learning space, reachable teachers, appropriate materials and a record of what the child has already learned.

For an eduKate reader, an appropriate classroom exercise is to examine claims and evidence, not to role-play military planning. Students can compare two headlines about the same incident, identify what each actually establishes, and rewrite an overstated sentence. They can distinguish a forecast from a fact, a probability from a certainty, and a country’s population from its government or armed organisations.

A practical example is the sentence: “There was a naval collision, so war will begin next month.” The first clause may be verifiable. The second requires a separate causal argument and evidence. A stronger version would say that the collision increases the importance of watching incident-management arrangements, while the outcome remains uncertain. This is not merely cautious language; it is a more accurate structure of reasoning.

Work is affected through similar dependencies. A person may be willing and able to work but unable to reach the workplace, obtain an input or receive payment. A business may preserve employment while reducing hours or changing tasks. The report should therefore avoid using one company’s announcement as a proxy for an entire labour market. It should ask what function changed, for whom, and over what period.

3.11 Singapore and Southeast Asia: exposure without automatic belligerency

The report’s Singapore base case is external disruption rather than direct conventional attack. This is an analytical judgment with limits, not an assurance of complete safety. Singapore’s exposure depends on actual trade, energy, financial, digital and human connections, not simply the physical distance from a conflict.

EMA’s explanation of imported gas provides one concrete example. Global fuel conditions can affect Singapore even when some supply arrives through regional pipelines. UNCTAD’s shipping research provides another: longer routes can affect capacity and cost without a local port being attacked. These mechanisms connect distant conflict to civilian life while avoiding the unsupported conclusion that Singapore automatically becomes a battlefield.

Terrorism is a separate category. ISD’s 10 September 2026 assessment describes a high terrorism threat and the relevance of overseas conflicts to radicalisation. That is an official terrorism assessment, not an interstate-invasion probability. The report does not merge those threats or infer any individual’s beliefs from nationality, religion or interest in international news. ISD: Singapore Terrorism Threat Assessment Report 2026.

For travellers, the appropriate decision source is current official advice combined with itinerary-specific information. MFA’s general guidance states that no overseas travel is entirely risk-free and that its advice has limits; it encourages preparation and informed choices. This report is not a replacement for those advisories, airline information, local instructions or insurance terms. MFA: general travel advice.

A household does not need to convert every geopolitical forecast into urgent purchasing. More useful questions concern accessible documents, reliable contact arrangements, essential medical needs, and knowledge of official information channels. Decisions should be proportionate to actual circumstances. A broad twelve-month risk range cannot tell a particular family whether a specific journey is suitable on a particular day.

Businesses need a different level of detail. A supplier in a region may have multiple production sites, different transport routes and contractual alternatives. A country label is not enough to assess exposure. The relevant chain is the actual product, supplier, route, inventory, payment schedule and substitution time. A report can identify the questions, but the business must establish its own facts before acting.

Schools need continuity and calm interpretation. They should distinguish a verified local disruption from a distant scenario, maintain clear communication and avoid spreading unverified claims. Students should not be encouraged to treat a forecast table as a list of enemies. Civilian populations are not interchangeable with governments, military forces or armed groups, and education should preserve that distinction.

Southeast Asia also contains repair mechanisms. The Cambodia–Thailand conciliation meeting in Singapore is one example of a dispute entering an operating process. It does not solve every regional tension, but it demonstrates why the monthly report must include diplomacy and implementation rather than only military activity. The ability to handle disagreement is part of the regional system being observed.

3.12 Science and earth systems: measure the right thing at the right scale

A useful observation has an object, a place, a time and a method. A satellite image may show physical change, but interpretation requires comparison and context. A weather model may describe a regional probability, but not the exact conditions at every location. A casualty report may describe confirmed deaths in one area, but not the total impact of a national conflict.

The forecasting error often begins when an observation is promoted across scale. One damaged vessel becomes a closed sea. One local ceasefire becomes national peace. One military exercise becomes a scheduled invasion. One quiet reporting period becomes evidence that a conflict has ended. The source may be accurate while the inference is wrong.

An explicitly hypothetical comparison shows the value of scale discipline. Two regions each report ten incidents. In one, reporting coverage has doubled; in the other, coverage has collapsed. The same count may represent different underlying trends. Without information about observation, a numerical comparison can be misleading. This does not mean data are useless. It means their production process belongs in the analysis.

Science also contributes negative evidence. If an expected indicator is absent where it would probably have been observed, that absence can matter. If the area is inaccessible or the sensor is poorly suited to the phenomenon, absence is much less informative. The report should not treat every failure to find evidence as evidence of safety, nor every gap as evidence of a hidden threat.

The monthly ledger is therefore an observation system as well as a prediction system. It records what was checked, what the source could establish, and what remained unknown. It allows later reviewers to distinguish a wrong forecast from an unavailable observation, while still holding the analysis accountable for how it handled uncertainty.

4. Pressure vs Repair

4.1 What would make the next twelve months less violent?

A war forecast that lists only escalation mechanisms is incomplete. It may notice a hostile speech while overlooking a verified withdrawal, count a weapons delivery while ignoring an implemented inspection arrangement, or interpret every negotiation as deception. That method cannot learn because all possible observations are forced into the same conclusion. This report instead asks what evidence would make its forecast less concerning.

The answer must be more specific than goodwill. A functioning restraint mechanism changes the options or incentives available during a crisis. It can provide a channel for correcting an incorrect accusation, establish who investigates an incident, make a reciprocal step observable, or reduce the immediate exposure of civilians. Its effect is conditional. A hotline that exists on paper but is not answered during an incident is not equivalent to a channel that actually resolves an urgent misunderstanding.

The table below contains analytical tests, not assertions that every listed mechanism is already operating. It also does not recommend a particular government’s policy. It explains what kinds of observable developments would count as evidence against escalation in this forecasting register.

Pressure being assessed	Potential restraint or repair	Observable return that would matter
Direct reciprocal strikes	A reciprocal cessation with a verification channel	A sustained reduction in attributable attacks, not only an announcement
Obstructed commercial navigation	Implemented arrangements for predictable passage	Multiple operators resume normal services under usable insurance conditions
Repeated dangerous maritime encounters	An incident-management arrangement used by the parties	Documented resolution of encounters without repeated hazardous escalation
Cross-border militant allegations	A monitored process for handling specific security complaints	Complaints produce investigations or agreed remedies rather than automatic retaliation
Unresolved territorial disagreement	A continuing legal or diplomatic process	Meetings, evidence exchange and interim restraint occur as agreed
Fragmented armed coalitions	Negotiations that include actors capable of implementation	Local units comply and excluded actors do not systematically defeat the arrangement
Disrupted civilian services	Access for qualified repair staff and supplies	Water, healthcare and transport function repeatedly rather than briefly
Displacement and uncertain return	Independent assessment of conditions in areas of return	People can return voluntarily with access to services and a means of reporting problems
Unclear attack attribution	A credible investigation with preserved evidence	Claims narrow or change when the evidence requires it
Escalation caused by misinformation	Timely correction through channels the parties use	A false account is corrected before it becomes the accepted basis for retaliation

The value of this table is symmetry. If commercial passage improves, the report must record it. If talks fail, it must record that too. The assessment should not require impossibly perfect evidence for improvement while accepting a single dramatic headline as proof of deterioration.

4.2 Repair has stages, and each stage has a different evidential value

A useful repair sequence is proposal, agreement, preparation, implementation, verification and persistence. These stages should not be compressed into the word peace. A proposal is evidence that a possible bargain has entered public discussion. An agreement identifies commitments. Preparation may involve staffing or access arrangements. Implementation is actual conduct. Verification establishes what occurred. Persistence tests whether the new conduct survives pressure.

Imagine a fictional border dispute in which two states agree to investigate future incidents jointly. The announcement alone changes very little about physical exposure. A later appointment of investigators is more concrete. An actual investigation that produces a shared account is stronger evidence. If another incident occurs and the parties still use the mechanism rather than immediately abandoning it, confidence in the restraint increases further. This example illustrates an evidence ladder, not a claim about a particular current conflict.

The same logic applies to commercial routes. A public declaration that a strait is open may be followed by limited passage, partial insurance cover and continued operator hesitation. Full commercial normalisation requires more than a few ships appearing on a tracking service. It involves repeated journeys, acceptable commercial terms, destination access and the absence of interruptions that make planning unreliable.

A monthly review should therefore ask which stage changed. Moving from proposal to preparation is a real delta even if it does not yet justify a large reduction in the war forecast. Moving backward from verified implementation to repeated violations is also a delta, even if the original agreement remains formally in force. The ledger can preserve these intermediate states without exaggerating either progress or failure.

4.3 Constraints can restrain war without resolving a dispute

A conflict does not require a fully resolved political disagreement in order to become less violent. Actors may continue to disagree while deciding that direct fighting is too costly, too uncertain or incompatible with other requirements. This is an analytical possibility across the theatres in this report, not a claim to know the private calculations of particular leaders.

Conversely, a reduction in available resources does not always produce peace. It may reduce the scale of fighting, shift violence towards different methods, create incentives to obtain external support, or make a limited agreement more attractive. The direction depends on the mechanisms connecting the constraint to a decision. It is not sound to infer either inevitable aggression or inevitable restraint from economic difficulty alone.

This is why the report separates capability, intention and choice. A force may possess capability that it does not intend to use. An actor may express an intention it cannot readily implement. It may also choose a limited action despite having wider capability. Forecasting requires attention to the relationship among these fields, rather than treating one as a complete substitute for the others.

For readers, the practical consequence is that a forecaster should be able to explain both directions of a key variable. If financial pressure is said to increase war risk, what is the proposed mechanism? If it is said to reduce risk, what alternatives become more attractive? If neither mechanism can be stated clearly, the observation may belong in contextual background rather than in a numerical probability revision.

4.4 Civilian continuity is a repair outcome in its own right

Repair should not be measured only by whether the numerical war threshold is avoided. A service can improve while fighting continues, and civilians can remain in danger after fighting declines. A more useful account tracks access to healthcare, safe water, food, education, communication and lawful movement alongside the combat record.

The ICRC’s work on water-related services explains why repairs, skilled staff, supplies and access matter together. Interruptions to one essential service can affect others; a hospital may suffer consequences from damaged water or electricity even when the hospital building itself is not struck. This is a general humanitarian mechanism documented before this forecast window, not a new casualty finding about any particular theatre. Source: ICRC on protecting water-related essential services during armed conflict.

An explicitly fictional recovery example makes the distinction concrete. A town’s water pump is repaired on Monday, but the electricity supply fails on Tuesday and replacement treatment chemicals cannot arrive by Friday. Counting Monday as complete recovery would misrepresent the service experienced by residents. The relevant measure is whether safe water continues to reach people, not simply whether one component was repaired once.

That distinction matters to the war outlook because damaged services can generate further displacement and longer recovery needs. It also prevents the forecast from becoming a contest to identify military escalation while ignoring the systems that allow people to survive it. The continuing delivery of essential services can be a meaningful positive return even when the broader political dispute remains unresolved.

4.5 Reversibility changes the cost of acting under uncertainty

A reversible civilian preparation can be useful at a lower level of certainty than an irreversible commitment. Reviewing contact details, testing a backup communication process or understanding a supplier’s contractual terms may be worthwhile across several possible futures. Those steps differ from abandoning a business, cancelling an essential journey or making a major financial decision solely because a forecast gives a concerning range.

This report does not supply personal investment, medical or legal instructions. It offers a structure for asking how costly a response is, how reversible it is and which evidence is needed before taking it. The same probability can justify different responses for different people because exposure and consequences differ. A shipping operator, a family with no travel plans and a student writing an essay do not face the same decision.

A hypothetical school excursion illustrates the point. Early in planning, organisers can identify official advisory sources, document communication responsibilities and clarify cancellation terms. These actions do not assume that war will happen. Closer to departure, the decision requires current destination-specific advice and the institution’s own procedures. An annual global risk range cannot replace that local, time-sensitive assessment.

Reversibility also belongs in the monthly ledger. If a forecast caused readers to incur unnecessary costs, the review should consider that consequence rather than celebrate only the absence of a missed warning. A responsible forecasting series should aim to improve decisions under uncertainty, not maximise fear or the number of precautionary actions attributed to it.

4.6 Neither agreement nor escalation should erase the civilian record

An implemented agreement can deserve recognition without implying that all damage has been repaired. A renewed conflict can deserve close attention without making previous successful restraint meaningless. The monthly record should hold both propositions at once.

Consider a fictional ceasefire that permits a school and clinic to reopen for three months before renewed fighting closes them. The eventual relapse does not prove that the three months of access were imaginary. Equally, the temporary reopening does not establish a durable settlement. Recording duration, service continuity and the reason for interruption produces a more useful account than classifying the entire period as either success or failure.

This report therefore maintains two connected but separate records: the defined conflict events and their civilian consequences. The first supports forecast adjudication. The second supports public understanding. Neither should be allowed to erase the other simply because a clean headline would be easier to write.

5. Shadow Signals

A Shadow Sensor is a question under observation, not a licence to publish an allegation. Its purpose is to keep potentially important uncertainty visible while preventing it from entering the confirmed record prematurely. The entries below identify interpretations that the available evidence does not establish. Listing an interpretation here is not evidence that it is secretly true.

Some weak signals deserve additional research; others should lose attention when they repeatedly fail verification. A responsible watchlist has an exit as well as an entrance. It should record when an explanation becomes less plausible, when an apparent indicator was a reporting error, or when a concern has not received enough evidence to justify continued prominence. Otherwise the shadow section becomes a permanent store of suspicion.

5.1 Is there a fixed date for a Taiwan invasion?

Not established. The dated March intelligence assessment discussed earlier weighs against presenting 2027 as a publicly established invasion timetable. It cannot rule out changed decisions later. The responsible uncertainty lies between those propositions: capability development is relevant, but a capability milestone does not identify a chosen operation or launch date.

The question would deserve promotion only with convergent, independently corroborated evidence that distinguishes a campaign commitment from exercises, bargaining and routine readiness. A widely circulated countdown does not supply that evidence. Neither does a map showing where an operation could theoretically occur. A possible operation and an intended operation are different objects of analysis.

Confidence in any exact-date claim: low. A future revision must identify what changed in evidence, not simply announce that the calendar has moved closer to a symbolic year. This entry remains a warning against false precision, not an alternative prediction of a secretly known date.

5.2 Has a conditional Hormuz offer restored normal shipping?

Not established by the offer. The reported proposal is evidence about a negotiating position. A claim of restored navigation would need a different evidence package: repeated completed passages, relevant operational notices and actual commercial participation. Announced willingness is neither accepted terms nor sustained performance.

Observed vessels provide stronger physical evidence than a price movement, but the data still require a defined sample. Vessel type, observation window, tracking coverage and comparison period matter. Ships not captured by a particular service are not automatically absent. A temporary increase from a depressed baseline is not necessarily normalisation.

Confidence in normalisation from diplomatic language alone: insufficient. The signal can move when implementation is observed. The report must also recognise genuine improvement if it appears; maintaining a warning after its physical basis has weakened would be another form of error.

5.3 Does an internal alliance prove interstate war?

No. The Ethiopian alliance announcement concerns organisations and their stated intentions. It does not, by itself, demonstrate unified operational control, agreement on all territorial disputes or an Eritrean decision to enter direct regular-force combat. Those are separate propositions requiring separate evidence.

A coalition can remain a political declaration, become partly effective, fragment, or develop more substantial coordination. The significance of an external relationship can also vary. Treating the most consequential possibility as confirmed would erase these alternatives. A forecast should identify them before a later event tempts the author to rewrite the announcement as an obvious warning.

Confidence in the announcement: higher than confidence in its future effects. Promotion would require observable coordination and independently established links to the particular interstate event being forecast. Continued disagreement within the coalition, or implemented regional restraint, should count against a simple escalation story.

5.4 Does a maritime death automatically trigger a wider war?

No automatic transition is established. A lethal incident would be a serious change, but attribution, consultation, investigation and political decisions would still shape what follows. Fatality, reciprocal weapons fire, a sustained campaign and external military entry are distinct events. The report’s China–Philippines question deliberately stops short of predicting all subsequent branches.

In a hypothetical crisis, prompt agreement on the facts might permit restraint. In another, disputed evidence and retaliatory commitments might narrow the available choices. Both examples are possible pathways, not observations of decisions already taken. The evidence needed to assess the next branch would arrive through conduct, not through an assumption that tragedy activates one inevitable sequence.

Confidence in an automatic-war claim: low. A qualifying lethal exchange would resolve the narrow forecast and open a separate assessment of escalation. It would not retrospectively transform the original question into a successful prediction of a regional war.

5.5 Is an airspace incident proof of an intentional attack?

Not without attribution and context. An object’s detection, origin, controller, intended destination and cause of deviation are different questions. Investigators may be more certain about some than others. A photograph of debris can support identification without establishing a leadership decision to widen a war.

The same discipline applies to suspected sabotage. Physical damage can be verified while responsibility remains unclear. A state connection can be alleged without being proven. The absence of immediate attribution should not be presented as either proof of innocence or proof of a sophisticated conspiracy.

Confidence in broad attribution without an adequate evidential chain: low. The relevant promotion trigger is corroboration that connects the act to the actor and supports the claimed intent. An investigation that narrows or revises its account is useful evidence, not an inconvenience to be omitted from the forecast.

5.6 Is a peace process already a durable settlement?

Not necessarily. A meeting or verification process is a real institutional action. Its existence can justify cautious improvement in an assessment, especially when it provides access or communication previously unavailable. Durability requires an operating history that an initial announcement cannot supply.

There is also a geographical boundary. A maritime conciliation procedure does not automatically settle land-border issues. A local ceasefire does not establish national cessation. A reduction in one type of attack does not prove that every form of violence has declined. Each improvement should be recognised at the scope for which evidence exists.

Confidence in the process can be high while confidence in the outcome remains limited. The report will promote durable restraint when implementation persists and survives disputes. It will not demand the disappearance of all disagreement before acknowledging progress, nor claim complete peace from one successful meeting.

5.7 Can climate or food prices predict a specific war?

Not on their own. Environmental stress and rising costs can affect livelihoods and the choices facing institutions. Whether those pressures contribute to armed conflict depends on the political, social and economic mechanisms through which they are handled. The physical stress is not the military decision.

A convincing argument would need evidence linking the observed stress to the relevant actors and decisions while considering alternatives. The same price increase could produce imports, assistance, substitution, protests, negotiations or violence under different conditions. Simply drawing an arrow from drought to war leaves the explanatory work undone.

Confidence in a deterministic climate-to-war claim: low. This report uses environmental information to examine exposure and operating constraints. It does not convert a seasonal forecast into the date or location of a military campaign.

5.8 Are many agreeing reports independent confirmation?

Only if their evidence is sufficiently independent. Multiple outlets can reproduce one wire report, quote one spokesperson or analyse one satellite image. Their agreement may show broad distribution, not additional observation. A source-family record should identify the original material and distinguish it from copies, translations and commentary.

Independent sources can also share a common limitation. Two accounts may agree on a location while both lack information about intent. Two images may show movement while neither establishes its purpose. Corroboration is specific to a proposition; it does not certify every interpretation attached to the incident.

Confidence in headline-counting as a method: insufficient. The report should gain confidence when an additional source tests a previously uncertain proposition. Repetition alone must not be treated as a new reason to raise a war probability.

5.9 Does quiet reporting mean the danger has passed?

Only where observation remains adequate. An absence of reports is more informative when the event would probably have been detected and reported. It is less informative where access has deteriorated, communications are interrupted or the available sources cover only a small part of the relevant area.

Nevertheless, a data gap should not be used to keep every alarming hypothesis alive indefinitely. The analyst should state what observation is missing and how it could change the assessment. Without that discipline, uncertainty becomes an unfalsifiable argument that a threat must exist because it cannot be seen.

Confidence depends on observability. A genuinely quieter period with continuing coverage can support a downward revision. A loss of coverage should normally widen uncertainty rather than automatically lower or raise the event probability.

5.10 Is Singapore automatically a belligerent in a regional war?

Not established. Commercial exposure, diplomatic engagement, defence cooperation and direct participation in combat are different relationships. A country can face serious economic consequences without becoming a battlefield. A naval exercise does not establish a decision about a future conflict. Neither does a country’s dependence on international trade.

The appropriate civilian response is to identify actual exposure and current official guidance, not infer a national war decision from association. A business should examine its supplier and route. A traveller should examine the itinerary and advisories. A school should examine continuity and information quality. These actions address real questions without declaring an outcome that the evidence does not support.

Confidence in automatic-entry claims: low. Any material change would require its own attributed evidence and a clearly stated scope. The report must not turn Singapore’s connectedness into a deterministic prediction of military participation.

6. What To Watch Tomorrow

In this monthly edition, tomorrow means the next meaningful observation as well as the next calendar day. The first formal review point is 23 October 2026; the original twelve-month horizon remains 23 September 2026 to 23 September 2027. A major verified change can justify an interim note. This is the editorial review plan, not a claim that automatic monitoring or future publication has been scheduled.

6.1 The next-observation board

Forecast track	Next evidence that could matter	What would not be enough
Iran and the Gulf	Implemented reciprocal restraint, attributable direct combat, and actual commercial passage	A reopening offer or price movement alone
Pakistan–Afghanistan	Dated reciprocal state-force encounters and the operation of mediation or border arrangements	A militant attack automatically counted as state combat
Ethiopia–Eritrea	Corroborated direct interstate activity or practical access and border-restraint arrangements	A coalition announcement treated as unified command
India–Pakistan	Whether a new incident is contained, investigated or followed by repeated direct combat	Assuming a trigger necessarily produces a campaign
China–Philippines	Verified encounter details, weapons use, casualties and functioning incident-management channels	A collision relabelled as a shooting
Taiwan interdiction	Actual sustained enforcement that obstructs commercial access	Exercise maps without enforcement
Taiwan invasion	Convergent campaign-scale evidence and a qualifying landing campaign if one begins	A symbolic year or possession of equipment
Russia–NATO	Attribution, reciprocal conduct and evidence that containment succeeds or fails	A single intercepted object treated as sustained war
Existing civil wars	Geographical persistence of violence, service access, verified restraint and durable return	One location’s improvement applied to the whole country
Singapore exposure	Actual delivery, service, travel or security changes through identified connections	A distant crisis assumed to have the same effect on every household

This board is a research priority list, not an instruction to gather sensitive military information or approach dangerous locations. The report uses public institutional releases, responsible reporting and civilian operating information. It does not need individual troop coordinates, exploitable infrastructure details or unauthorised access to private systems to improve its public-interest analysis.

6.2 Freeze the question before observing the answer

Every scored event needs an identifier, actors, geography, time window and resolution rule. The rule should specify what counts and what does not. Otherwise a forecaster can claim success after almost any development by changing the meaning of war, escalation or crisis. The eight numerical questions above retain their original ranges and distinguish campaign, encounter, interdiction and invasion.

Temporal boundaries matter as much as geographical ones. Fighting before the baseline is evidence used to form the forecast, not a successful prediction. A later source may report an earlier event; its publication date does not move that event into the forecast window. Where timing is uncertain enough to affect resolution, the correct status is unresolved until stronger evidence is available.

For the seven-day questions, a qualifying record should identify seven distinct combat days within the specified thirty-day period, the participating state forces and the evidence supporting each relevant claim. Multiple reports of the same encounter should not become multiple combat days. A summary saying tensions lasted a week is not necessarily evidence that direct fighting occurred on seven days.

The qualitative phrases in the Taiwan questions also require disciplined judgment. Broad commercial obstruction must be supported by evidence of actual enforcement and material effects, not merely stated intentions. A main-island invasion requires evidence of a large-scale campaign to seize and hold territory, not a training exercise or an offshore raid. If an unusual event leaves the original wording genuinely ambiguous, the report should disclose that weakness instead of silently expanding the definition.

6.3 Keep an append-only revision history

The baseline is retained even when it proves wrong. A later assessment should be added with its own date, evidence and reasoning. It should identify whether the revision arose from a new observation, a corrected source, an improved definition, or a changed interpretation of evidence already available. Those are different reasons for change and different opportunities to learn.

There are also two legitimate future forecast horizons. One asks about the remaining time until 23 September 2027. Another asks about the next twelve months from a new assessment date. They must have separate entries. If an October report quietly shifts the deadline to October 2027, readers lose the ability to evaluate the September forecast. A rolling outlook should never overwrite a fixed-cohort forecast.

The monthly ledger should therefore show the original probability range, any revised range for the same event and horizon, the new evidence, and the present outcome status. A fresh rolling-year forecast may sit alongside it, clearly labelled. This creates continuity without pretending that a shorter remaining exposure window and a new full year are the same question.

Corrections deserve the same visibility. If a supposed independent source was actually a republication, its evidential weight should be adjusted. If an event was incorrectly dated, the resolution should be reconsidered. If a forecast definition excluded an important form of harm, the report can create a better future question while acknowledging the limitation of the original. Improvement does not require erasing the record.

6.4 Score probabilities without claiming a calibration record

For later scoring, the initial register retains the midpoint of each range: 70%, 55%, 27.5%, 27.5%, 22.5%, 15%, 7.5% and 10%, in the order shown in the baseline table. These midpoints are bookkeeping choices. They do not imply a precision greater than the surrounding ranges, and the ranges themselves are not confidence intervals.

A transparent probability-error rule is to square the difference between the forecast probability and the observed binary outcome. Use one for a qualifying event and zero for verified non-occurrence at the end of the window. A 70% forecast scores 0.09 if the event occurs and 0.49 if it does not: respectively, (0.70 − 1)² and (0.70 − 0)². Lower values are better under this scoring rule. The example is arithmetic, not a forecast result.

This avoids treating every unlikely event that occurs as proof of an incompetent forecast, or every likely event that fails to occur as proof of dishonesty. Probabilities allow both outcomes. What matters over a sufficiently large and suitable record is whether the probabilities discriminate and align with realised frequencies. Eight heterogeneous, dependent questions over one year cannot establish that record.

An average score across these eight entries would be descriptive only. A strong score driven by easy continuation questions would not prove expertise about rare new wars. Questions with overlapping causes or outcomes should not be treated as eight independent experiments. Numerical scoring must be accompanied by review of event definitions, source quality, timing, missed fronts and the decisions the report encouraged.

Unresolved events should not be forced into zero because the evidence is inconvenient. Equally, a qualifying occurrence must not be ignored because its scale differed from the author’s imagined scenario. Outcome adjudication answers the frozen question first; a separate review explains whether the mechanism, severity and consequence analysis were useful.

6.5 Distinguish a miss, a false alarm and successful prevention

A missed event is a meaningful outcome outside the forecast’s field or below its anticipated level of concern. A false alarm concerns a warning that did not justify its cost or did not lead to the specified event under an explicitly stated decision rule. Neither label should be assigned casually to a probabilistic statement. At 20–35%, non-occurrence is the more likely outcome of the forecast itself.

Successful prevention raises a different question. If an event does not occur after a diplomatic intervention, the event forecast still resolves according to the original rule. Whether the intervention prevented it is a causal claim requiring evidence. The report cannot automatically award itself credit by saying its warning helped avert the war. It should record the intervention and analyse its plausible effect without changing the observed outcome.

Timing errors also matter. A forecast made after decisive evidence had become public may be correct but offer little warning value. A very early warning may create attention costs without identifying a useful decision point. Lead time should therefore be assessed alongside the probability score. The goal is not merely to describe an event eventually, but to improve understanding while meaningful choices remain available.

Mechanism errors should be retained even when the event occurs. If a forecast expected escalation through one pathway but the event arrived through another, the numerical outcome alone cannot validate the explanation. A useful review asks which assumptions survived, which failed and what new question should be framed before the next cycle.

6.6 Choose observations for information value, not drama

The next article to read should ideally answer an unresolved question. If ten reports already establish that a meeting occurred, an eleventh account of the opening ceremony adds little. A document describing the agreement’s implementation may add much more. If the dispute concerns whether commercial passage is normal, a further political speech may be less useful than consistent operating information from several carriers.

A practical test is to write down the competing explanations before searching. In a hypothetical mobilisation story, the alternatives might include routine training, coercive bargaining and preparation for a sustained campaign. The useful observation is one that changes their relative plausibility. Merely collecting material compatible with all three can enlarge a dossier without improving the forecast.

This method also limits overconfidence from sophisticated tools. A clearer image may improve identification while leaving intent unknown. An automated classification may save time while preserving the original source’s uncertainty. An additional database can broaden coverage without solving dependence among reports. The relevant question is always what uncertainty was actually reduced.

6.7 A worked monthly-update example

Consider a completely fictional October review of a maritime forecast. In the first week, a dramatic video is circulated under three headlines. Verification establishes that all three derive from the same original account. The event is real, but the evidence count is one family rather than three. Its classification remains a nonlethal collision and it does not resolve a forecast requiring reciprocal weapons fire and a death.

During the second week, the two parties use an existing communication mechanism and publish compatible accounts of how another encounter was handled. That supplies evidence of functioning restraint. During the third week, an unrelated commercial operator reports a delay, but the cause is weather rather than enforcement. The delay belongs in the logistics record and should not be used to corroborate a blockade story.

The monthly analyst now has more than a tally of alarming events. There is one verified collision, one successful containment episode and one commercial delay with a different cause. A justified update might leave the probability unchanged or lower it modestly, depending on the wider context. The method does not prescribe a numerical amount; it requires the reason for any change to match the actual evidence.

The published update should retain the original range, state that the event remains open, describe the evidence and identify the next uncertainty. It should not say the forecast was vindicated because there was a collision, nor claim peace because a hotline worked once. This example shows how an evidence ledger can make a monthly series more informative without producing a new crisis narrative every month.

6.8 Reader questions before sharing a war prediction

Before sharing a prediction, identify the event it actually forecasts. Is it war, a clash, a blockade or simply higher tension? Check the horizon and the date of the evidence. Identify who owns the probability: an institution, a survey, a model or the article’s author. Look for the strongest contrary evidence and the stated conditions for lowering concern.

Then ask whether the headline is broader than the evidence. A country name can conceal a local event. A percentage can conceal an ambiguous definition. A description of capability can conceal the absence of evidence about intention. A scenario can be presented so vividly that readers forget it has not occurred. Sharing the qualification with the claim is part of accurate communication.

For students, a useful exercise is to rewrite one overconfident sentence into three parts: observed fact, interpretation and unresolved question. The exercise should not require taking sides in a conflict or treating populations as adversaries. Its purpose is to strengthen reading and reasoning so that evidence can support concern without being distorted into certainty.

7. Final Daily Reading

The conclusion of this monthly outlook is a judgment about transitions, not a prophecy about a country. Between 23 September 2026 and 23 September 2027, this report expects existing wars and recurring regional campaigns to remain more prominent than a single coordinated world war. That assessment could be wrong. Its value depends on the clarity of the questions, the quality of the supporting evidence and the willingness to record contrary outcomes.

The Gulf is the strongest escalation concern in the numerical register because the forecast concerns further direct combat within an already confrontational system. Pakistan–Afghanistan is a recurrence concern because a reduction in fighting does not by itself settle the underlying security dispute. Ethiopia–Eritrea and India–Pakistan remain below even odds for the defined sustained campaigns, but their potential consequences justify close attention. These are not eight versions of the same event.

In East Asia, the distinctions are particularly important. A China–Philippines exchange causing a death would be grave, but it would not automatically establish a sustained regional war. An enforced Taiwan access regime would be different from an exercise and different again from a landing campaign. The main-island invasion scenario remains a lower-probability, very-high-consequence possibility in this assessment. The report rejects both a confident invasion countdown and a confident assurance that such a campaign is impossible.

The Russia–NATO assessment follows the same logic. A dangerous incident can occur without a decision for sustained direct combat. Attribution and response matter. So do the mechanisms that permit governments to correct an interpretation without treating every correction as weakness. The forecast concerns whether the defined transition occurs, not whether every day of the coming year will be free of risk.

Meanwhile, continued conflict in Ukraine, Sudan, Myanmar, the central Sahel and eastern DRC deserves attention in its own right. A report that spends all its space on the next possible great-power war can overlook people already experiencing the conditions it warns about. Existing violence can deepen a humanitarian emergency without a new declaration or a spectacular change on a map. A family’s loss of safe water, income or learning time is not made less important by the familiarity of the conflict.

7.1 The important unit is the function that people need

The public framework in How War Works connects military action to the systems that sustain a society. The forecasting extension asks a corresponding civilian question: what function could disappear, and how would we know soon enough to preserve it? A hospital is more than a building. A port is more than a location. A school is more than an enrolment list. Each depends on people, information, supplies, access and time.

This is why the report separates physical survival from service continuity. A bridge may stand while its route is unusable. Food may exist while a household cannot afford it. A classroom may reopen while attendance remains impossible for displaced children. A ceasefire may hold while the basic requirements for recovery remain missing. Conversely, a repair can return a valuable function even when the larger dispute has not ended.

A useful war forecast should therefore remain interested in outcomes after the military event. Did the alternative transport arrangement actually work? Did civilians regain access to care? Could displaced families remain safely where they returned? Did a verification team resolve an incident before retaliation? These questions are not secondary to an account of civilisation health. They explain whether the pressure was contained or continued spreading through everyday life.

They also prevent a forecast from becoming a detached exercise in predicting violence. Correctly anticipating a campaign does not mean the event is desirable. Failing to anticipate a settlement does not make the settlement unwelcome. The purpose of the ledger is to improve understanding and preserve choices, not to give the publication a reason to prefer the future in which its most alarming prediction succeeds.

7.2 What the September baseline commits us to

The baseline commits the report to a fixed period and specified event definitions. It commits it to retaining the original probability ranges when later estimates change. It commits it to acknowledging that a shorter campaign can be devastating without satisfying a seven-day question. It commits it to recognising actual restraint even when the result is less dramatic than another warning.

It does not commit the report to defending an original interpretation forever. If evidence reveals that two supposedly separate sources came from one account, the evidential weight must change. If a current claim was wrong, the text should be corrected with an explanation. If a new agreement proves effective, the forecast should be reconsidered. Persistence of the record is not persistence of a belief regardless of evidence.

The coming year may also expose omissions. An event outside the eight questions must be recorded as outside the original numerical register. A broad reference to global uncertainty cannot be used to claim it was predicted. An additional forecast may be issued prospectively, with its own date and horizon. That is how a limited first register can improve without pretending it began as a complete map of the future.

For Singapore readers, this produces a practical form of calm rather than reassurance without evidence. The task is to identify actual dependencies, follow relevant official information and match preparation to the decision at hand. A student reading international news, a family planning travel and a firm organising imports need different information. None should be asked to make a major decision from an annual percentage alone.

7.3 Warning quality is part of resilience

The previous Purple Report progression moved from repair towards prevention, redundancy, allocation and opportunity cost. This monthly edition adds warning quality because poor information can consume the very resources that resilience is meant to preserve. An exaggerated warning can divert attention and create avoidable costs. A late warning can leave too little time for a useful response. A vague warning can be impossible to evaluate afterwards.

The better question is not whether the report sounds sufficiently certain. It is whether the reader can identify the event, understand the evidence, distinguish the alternatives and recognise the next meaningful observation. Those are qualities that a publication can work to improve before knowing whether its prediction will be right.

Decision thresholds then connect warning to action. A low-cost, reversible check may be reasonable under broad uncertainty. A high-cost or irreversible action needs a much more specific evidential basis and the appropriate expertise. There is no single percentage that tells every household, institution and business to respond in the same way. Consequences, exposure and the time needed to prepare are part of the decision.

The final September reading is therefore neither a declaration that a world war is coming nor a promise that deterrence will prevent one. It is a set of bounded judgments about where conflict may persist, recur or cross a more dangerous threshold, together with an obligation to preserve evidence of restraint and repair. The most important future update may be a downward revision that explains why a feared pathway has become less plausible.

A useful forecast leaves room for a better future—and states clearly what evidence would show that the world is moving towards it.

8. ID / Almost-Code

Public report identifier: PR.WAR.12M.2026-09-23
Edition: September 2026 monthly war outlook
Forecast window: 23 September 2026–23 September 2027
First formal review point: 23 October 2026
Method: structured analytical judgment, with AI-assisted research and synthesis; not an independently calibrated war-prediction model.

The compact record below is a reader-facing summary of the forecast contract. It is not executable military software, an official threat advisory or a claim that a private analytical runtime was executed. The prose event definitions above govern interpretation. All eight outcomes are open at the baseline.

PURPLE_WAR_OUTLOOK_2026_09_23 {
    public_id: PR.WAR.12M.2026-09-23
    edition: SEPTEMBER_2026_BASELINE
    forecast_window: 2026-09-23_to_2027-09-23
    first_formal_review: 2026-10-23
    automatic_monitoring_scheduled: false

    method: STRUCTURED_ANALYTICAL_JUDGMENT
    probability_ranges: SUBJECTIVE_NOT_CALIBRATED
    confidence_intervals: false
    empirical_calibration_record: NOT_ESTABLISHED
    official_intelligence_product: false
    completed_scan_count_claimed: false

    evidence_states:
        CONFIRMED_REALITY
        ATTRIBUTED_CLAIM
        DEVELOPING_SIGNAL
        SHADOW_SENSOR
        UNKNOWN

    event_register:
        001_IRAN_DIRECT_COMBAT: 60_to_80_percent
        002_PAKISTAN_AFGHANISTAN: 45_to_65_percent
        003_ETHIOPIA_ERITREA: 20_to_35_percent
        004_INDIA_PAKISTAN: 20_to_35_percent
        005_CHINA_PHILIPPINES_LETHAL_EXCHANGE: 15_to_30_percent
        006_TAIWAN_ENFORCED_INTERDICTION: 10_to_20_percent
        007_TAIWAN_MAIN_ISLAND_INVASION: 5_to_10_percent
        008_RUSSIA_NATO_SUSTAINED_COMBAT: 5_to_15_percent

    resolution:
        apply_frozen_actor_geography_and_duration_rules
        exclude_events_before_baseline
        distinguish_event_date_from_report_date
        require_corroborated_evidence
        preserve_UNRESOLVED_when_evidence_is_insufficient
        do_not_score_nonoccurrence_before_window_end

    update:
        retain_original_forecast
        append_dated_evidence_and_reasons
        separate_fixed_horizon_from_new_rolling_year
        record_restraint_as_well_as_escalation
        retain_misses_and_wrong_mechanisms

    prohibited_shortcuts:
        repeated_headlines != independent_observations
        capability != chosen_operation
        exercise != enforced_blockade
        blockade != main_island_invasion
        collision != reciprocal_weapons_fire
        proxy_action != automatic_state_attribution
        ceasefire_announcement != durable_implementation
        price != physical_flow
        regional_probabilities != independent_world_war_formula

    purpose:
        improve_understanding
        preserve_civilian_choices
        observe_what_changes
        correct_the_record
}

8.1 How the source base should be read

Sources in this report perform different jobs. An official organisation is the primary source for its own statement or programme. A news agency can establish a reported incident while attributing competing accounts of responsibility. A research institution can provide a model, definition or earlier assessment without certifying this publication’s probability ranges. The hyperlinks identify those sources; they are not a claim that the organisations endorse the report.

The evidence base also contains different dates. September reporting informs the current regional baseline. The March intelligence assessment is explicitly retained as a dated assessment. The IPCC, ICRC and UNCTAD materials support general mechanisms and definitions. A historical mechanism reference must not be presented as a fresh measurement of the current conflict. A newer headline must not silently replace an older source’s observation window.

Where a cited account reports what an official, organisation or party said, that attribution remains part of the claim. An allegation does not become established merely because it appears in a reputable outlet. Where publications share a wire report or original statement, they belong to the same evidence family for that proposition. No source count is offered as a substitute for checking what each source actually supports.

8.2 Comparison with other conflict outlooks

The Council on Foreign Relations’ Conflicts to Watch in 2026 provides a useful external comparison. Its survey was conducted in November 2025, its contingencies concern calendar 2026, and its priority ranking includes potential harm to U.S. interests. Those choices differ from this report’s September-to-September horizon and civilian-consequence emphasis. Its Taiwan crisis and Russia–NATO clash categories must not be relabelled as this report’s invasion and sustained-combat questions.

Two forecasts can consequently have different numbers without being direct contradictions. One may ask about any clash; another about a campaign lasting a specified period. One may cover a full new year; another the remaining months of a fixed window. Before comparing accuracy, align the event, geography, time and resolution rules. Agreement on a general hotspot is not the same as agreement on a precisely defined outcome.

This report has not converted any outside institution’s risk label into its own numbers through an undisclosed formula. The eight ranges remain the publication’s initial judgments. External assessments are checks on reasoning and coverage, not borrowed authority for probabilities they did not issue. A future comparison should preserve this distinction even when another forecast appears to support the preferred conclusion.

8.3 Primary-source reading guide

Conflict classification: the ICRC’s international armed conflict explanation concerns legal classification, while UCDP’s definitions concern statistical recording. Neither uses this report’s editorial seven-day convention as the boundary for recognising all armed conflict. Consult the issuing institution’s terminology before comparing labels.

Humanitarian consequences: UNHCR’s 15 September Yemen release, WFP’s 17 September response update, and UNICEF’s 16 September account of children’s access to services describe different observations and programmes. Their populations and reporting periods should not be added together as though they were disjoint counts.

Implemented diplomacy: the PCA’s 16 September release records a meeting that occurred. It supports the maritime-process observation, not a claim that all Thailand–Cambodia disputes have been settled. The scope of the procedure is part of the evidence.

Singapore’s differentiated exposure: EMA’s energy-supply explanation, ISD’s September terrorism assessment, and MFA’s general travel guidance address different decisions. Energy exposure is not terrorism risk; terrorism risk is not a forecast of interstate invasion; general travel advice is not a destination-specific clearance for a particular journey.

Physical systems and civilian continuity: IPCC AR6 Working Group II, Chapter 16, the ICRC’s urban-services report, and UNCTAD’s 2025 maritime review provide mechanism-level context. They should be read with their dates and assumptions intact, rather than used as evidence that a particular future military action is inevitable.

8.4 The record that future editions must preserve

The permanent element of this edition is not a belief that the world must follow its scenarios. It is the dated record of what was judged, on what basis and over what window. A later edition may change every probability if the evidence warrants it. It must still allow readers to recover the original assessment and understand the revision.

Future updates should separate three possible failures. The source may have been inaccurate. The interpretation may have been unreasonable despite an accurate source. The forecast may have assigned a defensible probability to an outcome that nevertheless did not occur. These cases demand different lessons. A publication that treats all three as either total vindication or total failure will learn less than one that examines the actual error.

The report also retains a boundary around its tools. Public frameworks can organise questions about words, evidence, geography and institutions. They do not themselves establish facts about secret military intentions. AI-assisted synthesis can help compare material, but it is not an independent eyewitness or a substitute for source verification. No claim of exhaustive estate-wide research, one thousand completed checks, independent expert review or empirical forecast calibration is made for this edition.

To follow the wider work, return to the Purple Report and Latest News collection. For the underlying explanation, read How War Works with StrategizeOS, WarOS, Weather, Geography, and Environment. For the separate natural-hazard horizon, read the September 2026 disaster-forecast baseline. These are related reading routes, not duplicate versions of the same forecast.

Purple Report rule for this edition: define the event before predicting it, preserve the evidence before interpreting it, and keep the original forecast visible when reality requires a different answer.