Additional Mathematics Fatigue Curve | Why Accuracy Falls Late in an Examination

The Fatigue Curve Laboratory

The fatigue curve becomes trainable when late-paper deterioration is converted into controlled experiments. The laboratory does not attempt to manufacture exhaustion. Its purpose is to change one temporal or operational condition, observe which performance variable moves first and then test a repair.

A good fatigue experiment has four properties: the underlying mathematics is already substantially known; the early and late tasks are reasonably comparable; the load change is deliberate; and the review records behaviour before the final answer alone.

Lab 1: the relocation experiment

Select one structure that the student can execute reliably while fresh. Use a fresh parallel version early in one session and another parallel version late in a later session.

Record accuracy, first-move latency, execution time, working clarity and checking behaviour. If the late version is weaker across several trials, the temporal position is meaningful. If both versions are similar, the original late-paper error may have been caused by question difficulty, topic weakness or time debt rather than duration itself.

Lab 2: the time-debt neutralisation test

Many students appear fatigued only because they are behind schedule. To test this, create two comparable long sessions. In the first, allow an early question to consume extra time as it normally would. In the second, remove that early time sink by replacing it with an ordinary question or by imposing the trained stopping rule.

Now compare the final third. If late accuracy, checking or completion improves substantially when the time debt is removed, the original curve was partly a pacing curve.

This experiment prevents a student from trying to solve a timing problem only by increasing endurance.

Lab 3: the method-economy test

Take a question family where the student uses a long but valid route. Teach or consolidate a shorter reliable route. Then compare two extended sessions: one using the old route and one using the new route on fresh questions.

Do not judge only the individual question time. Observe whether later performance changes. A modest saving repeated across many questions can move the fatigue breakpoint much farther than a dramatic saving on one rare item.

Lab 4: the checking-load test

Tricia’s profile is useful here. Run one session with her ordinary checking habits. In another, preserve only high-yield checks: high-propagation setup, exact-value handoff, final constraint and surprising result.

Compare late accuracy and completion. If accuracy remains stable while more of the paper is finished, the earlier checking system was consuming fatigue reserve rather than protecting it.

The experiment should never encourage careless omission of checks required by the student’s actual assessment strategy. It is about removing low-yield repetition.

Lab 5: the governor-retention test

Kai Kai’s speed makes this test important. Identify three safeguards that protect high-severity errors—for example interval restoration, exactness preservation and a quick check before reusing a wide-node result.

Place those triggers early and late in extended practice. Measure whether the safeguards still activate without prompts.

If they vanish late, the problem is not that Kai Kai forgot the mathematics. The safeguards have not yet been compiled deeply enough to survive load.

Lab 6: the latency-retention test

Use short first-move drills at the beginning and near the end of the same extended session. Because the student does not need to complete full solutions, many decisions can be sampled quickly.

Track the median and the high-latency tail. A student may preserve ordinary decisions while only the most ambiguous questions become much slower late. That tail can create the time debt that later looks like general fatigue.

Lab 7: the propagation-retention test

Give comparable multi-stage questions early and late. Seed no errors. Simply record naturally occurring origins and how far they travel before detection.

If propagation distance increases late even when origin frequency stays similar, the containment system is the fatigue bottleneck.

The repair may be smaller than expected: preserve one late checkpoint at high-dependency states rather than adding more general accuracy drills.

Lab 8: the context-switch test

Some students tolerate duration when the work stays within one topic but deteriorate when the paper switches continuously. To test this, compare an extended topical session with an extended mixed session of similar overall mathematical difficulty.

If the mixed session produces earlier latency growth or more wrong-route starts, context switching is contributing to the fatigue curve.

Train the reset between questions: release the previous method, identify the new target and allow the new problem to earn its own representation.

Lab 9: the state-preservation test

Compare late questions requiring many linked states with late questions that are equally difficult but more self-contained. If the linked questions deteriorate more strongly, state management may be the weak curve.

Test whether clearer notation, boxed handoffs and visible constraints reduce the late penalty. If they do, the student needs better external state architecture rather than simply more stamina.

Lab 10: the recovery-after-disruption test

Insert one unusually demanding question before a set of otherwise ordinary late questions. The important measurement is what happens after the disruption.

Does the student return immediately to ordinary pace? Or do the next three questions show shorter working, more mistakes and higher latency?

A long recovery tail can make the fatigue curve appear worse than the underlying duration effect. Train the reset separately.

The fatigue signature atlas

Students often develop repeatable fatigue signatures. These signatures are more useful than the generic label “tired.”

The compression signature in detail

Late compression is tempting because it appears to save time. The student combines transformations, performs substitutions mentally and writes only final numerical states. Sometimes this is genuine expertise. Sometimes it is emergency behaviour.

Distinguish the two by error and recoverability. Expert compression preserves reliability. Emergency compression increases ambiguity and makes rollback harder.

Alicia should identify a minimum late-paper working standard: high-propagation equations remain explicit, sign-sensitive transitions get their own line when needed and exact handoffs stay visible. Everything else can remain compact.

The overchecking signature in detail

Tricia’s late response to uncertainty is often to spend more time proving that already-correct work is correct. The result is a paradox: she protects local accuracy and damages global completion.

Her fatigue strategy should define a verification ceiling. Once a low-risk answer has passed one suitable check, move on. Save deeper inspection for wide nodes or surprising results.

The exact checking strategy must match the actual paper, but the principle is stable: verification should not consume the marks it is trying to protect.

The underchecking signature in detail

Kai Kai sheds checks to preserve speed. This can work on low-risk routine questions and fail catastrophically at one high-dependency node.

His late strategy is trigger-based checking. If the result will be reused, if a candidate must be filtered, if a numerical output is surprising or if the model line controls the whole route, check. Otherwise continue.

Trigger-based checking survives fatigue better than a vague instruction to “be careful.”

The switching signature in detail

Under sustained load, students may lose tolerance for temporary ugliness. A valid route becomes uncomfortable, so they switch. The replacement route is not necessarily better, and the switch itself consumes time.

Late-paper method switches should require evidence. Algebra growing temporarily is not sufficient. A route becomes suspect when it no longer uses the controlling condition, creates unresolved variables without a path, contradicts known constraints or is clearly dominated by another representation.

This creates a fatigue-resistant persistence rule.

The exactness signature in detail

Exact expressions often feel heavier late because they demand symbolic attention. Students replace them with decimals to simplify the page.

But decimals can create a new chain of precision decisions. The better late policy is simple: if the value is still feeding later mathematics, preserve exactness or the required internal precision. If the quantity is genuinely final, output according to the board’s requirement.

The policy removes repeated late decisions.

The propagation signature in detail

Late-paper fatigue can reveal itself not through more origins but through worse detection. The student makes one error and carries it six lines instead of one.

Dependency-aware checkpoints should therefore be selected partly for temporal durability. A sophisticated check that disappears under fatigue is less valuable than a simple one that still runs automatically.

The throughput signature in detail

Some students become slower without noticing. They still produce good work, so the decline is hidden until the clock becomes critical.

Compare time per familiar question by paper region. If the same type of six-minute question becomes an eight-minute question late, inspect the extra two minutes. Are they rereading, checking, rewriting, deciding or calculating?

The intervention should remove the specific extra cost.

The fatigue budget metaphor

It can be useful to imagine that a student has a limited budget of high-quality deliberate control during a paper. This is only a metaphor; the mind does not contain a literal fixed tank that empties predictably. The metaphor is useful because it encourages allocation.

Do not spend high-control attention on every low-risk arithmetic step if the paper still contains difficult modelling and synthesis. Do not repeatedly inspect familiar answers because certainty feels good. Do not carry unresolved questions mentally into new ones.

Spend deliberate control where mathematical consequence is high.

Fatigue debt

Time debt has a close cousin: fatigue debt. When early work is unnecessarily expensive, the student arrives late with less capacity for the same remaining mathematics.

Fatigue debt can be created by:

• overlong methods;
• constant self-questioning;
• repeated checking;
• poor notation that forces reconstruction;
• method switching;
• doing mentally what the page could store;
• continuing dead routes too long.

Training should reduce avoidable debt before attempting to expand raw endurance.

Fatigue interest

Some fatigue debt compounds. A student becomes slightly slower, falls behind time, accelerates to catch up, makes an error, checks more and becomes even slower.

This feedback loop means a small early deterioration can produce a large late collapse.

Break the loop early. If latency begins to rise, rely on defaults. If checking expands, return to risk weighting. If one question becomes a time sink, use the stopping rule. Do not wait for the final score to reveal the loop.

The fatigue cascade

A common cascade is:

slower recognition → time debt → working compression → weaker verification → more propagation → confidence drop → even slower recognition.

Another is:

overchecking → throughput loss → late rush → exactness shortcuts → execution errors → global rechecking.

These cascades show why the first weak curve matters. Repairing the first transition can prevent the whole sequence.

The fatigue circuit breaker

A circuit breaker is a preplanned action used when an amber signal appears. It should be small enough not to interrupt the entire paper.

• latency rising → compress to target and controlling condition;
• time debt growing → use stopping rule on high-cost question;
• working becoming opaque → restore one-line-per-state at high-risk transitions;
• checking expanding → return to high-propagation checks only;
• method switching rising → default route unless evidence triggers alternative;
• one error discovered → localise descendants before rechecking globally.

The circuit breaker does not eliminate fatigue. It prevents fatigue from reorganising the whole examination strategy.

A paper can contain multiple fatigue breakpoints

The student may have one breakpoint for routine algebra, another for method selection and another for difficult synthesis. These do not need to occur at the same time.

For example, Tricia may preserve basic execution for the full paper but show checking inflation after forty minutes and first-move delay only after seventy.

A multidimensional profile is therefore more useful than one statement such as “my stamina lasts an hour.”

The fatigue matrix by topic and paper position

Build a simple matrix with topic families down one side and early, middle and late positions across the top. Mark where performance changes.

The question marks are deliberate. The matrix should store observations and hypotheses, not turn one paper into certainty.

Topic-specific fatigue can reveal carrier weaknesses

If only calculus deteriorates late, the student may not have a general stamina problem. The calculus carrier operations may simply be more expensive: algebra after differentiation, exact substitution, linked interpretation or long product/chain structures where relevant.

Reduce the cost of that topic family and retest. General endurance training is not always the most efficient answer.

Fatigue and representation switching

Late in a paper, students can become more reluctant to switch representation even when a graph, substitution or rearrangement would make the problem cheaper. They stay in the first form because reconstruction feels costly.

Train a small library of high-value representation switches until the trigger is familiar. The late student should not need to invent the switch from scratch.

Fatigue and constraint loss

Intervals, domains and contextual restrictions are often dropped late because they are not part of the main algebraic action. They sit at the edge of attention until the finish.

Move them into the state. Write the interval. Mark the domain. Carry the restriction forward. A visible constraint is less dependent on late memory.

Fatigue and question-contract loss

A student can perform the mathematics correctly and still answer the wrong object late because they no longer return to the wording.

Use a finish question: What did the problem actually ask me to return? A value, equation, coordinates, proof, sketch, interval or interpretation may require different final forms.

The finish question is cheap enough to remain part of the late-paper operating system.

Fatigue and route overfitting

When deliberate control becomes expensive, students may fall back on the most recently practised route. This can create method perseveration: using a familiar method simply because it is highly activated.

Mixed late-position drills should therefore test not only endurance but discrimination. The student must still let the new question choose the method.

Fatigue and route underfitting

The opposite problem occurs when the student becomes so cautious late that they refuse to use an efficient route unless it looks exactly like a textbook example. They revert to longer generic methods.

Strong trigger-based alternatives help here. If the mathematical condition is present, the shorter route is authorised even when the surface is unfamiliar.

Fatigue and error detectability

Late students may not make more errors; they may notice fewer. This distinction matters because detection can often be improved with independent plausibility signals even before raw execution accuracy changes.

• Does the sign fit the graph?
• Does the coordinate satisfy the original equation?
• Does the candidate satisfy the domain?
• Is the magnitude plausible?
• Does differentiating the antiderivative return the integrand?

The best late checks are short and structurally different from the original route.

Fatigue and confidence calibration

Some students become less confident late even when accuracy remains high. Others remain overconfident while safeguards disappear.

During practice, occasionally record confidence before marking. Compare confidence, correctness, time cost and checking behaviour by paper position.

If late low confidence repeatedly accompanies correct work, the student may be wasting time on reassurance checks. If late high confidence accompanies propagating errors, condition-triggered verification becomes more important.

Fatigue and the score floor

Late-paper robustness protects the score floor. Even when the final questions are difficult, routine and medium-difficulty marks should not disappear because one earlier event changed the entire operating state.

Protect the floor with low-cost defaults: clear handoffs, exactness policy, stopping rule, high-value checks and clean question resets.

Fatigue and the score ceiling

The ceiling also depends on endurance. Difficult synthesis questions often appear when substantial work has already been completed. A student can possess the conceptual capability and still fail to access it if decision and state-management costs have accumulated.

High-score training therefore needs late access to deep mathematics, not just fresh access.

The fatigue reliability ladder

1. Fresh correct: skill works at the beginning.
2. Mixed correct: skill survives topic switching.
3. Extended correct: skill survives longer continuous work.
4. Late correct: skill survives temporal relocation.
5. Timed late correct: skill survives realistic pace late.
6. Disrupted late correct: skill survives after a difficult preceding question.
7. Full-paper reliable: skill survives authentic interaction with the whole paper.

The ladder is not an official standard. It simply prevents fresh success from being mistaken for complete endurance.

Fatigue Curve Engineering: Measurement, Scheduling and Transfer

A useful fatigue model should do more than describe why late work feels harder. It should help decide what to measure, what to train, when to lengthen practice and when a once-fragile late-paper behaviour is reliable enough to move into maintenance.

The central discipline is comparison. Late performance means little without a suitable reference. A difficult final question may be difficult because of its mathematics, its position, the student’s accumulated time debt, or all three. Good fatigue engineering tries to separate those influences before prescribing a solution.

The fatigue fingerprint across several papers

One paper gives a snapshot. Several fresh papers can reveal a fingerprint: a repeatable pattern of what changes as the student moves through sustained work.

A fatigue fingerprint might look like this:

• first-move time is stable for most of the paper but rises sharply on mixed synthesis late;
• algebra accuracy remains high but exactness discipline weakens;
• checking becomes shorter after a major time sink;
• propagation distance increases in the final quarter;
• routine questions remain strong, while questions requiring representation switching deteriorate.

The fingerprint matters because it is more specific than “I get tired.” It tells the student which control functions need to survive later and which are already robust.

Do not average away the problem

An average score across the whole paper can hide temporal structure. A student may score ninety percent in the first half and sixty-five percent in the second, yet the overall percentage looks respectable. Another student may be equally accurate throughout but leave the final ten percent untouched.

For fatigue analysis, preserve position. Record where the change happened.

The purpose is not to generate complicated statistics. It is to avoid turning a time-dependent failure into an undifferentiated average.

Temporal bins

Instead of only early, middle and late, a tutor can divide a practice paper into smaller temporal bins—for example, roughly equal portions of the actual paper duration. The exact size depends on the assessment and should not be universalised.

For each bin, record only a few variables:

• questions attempted;
• first-move delays that were unusually long;
• avoidable execution errors;
• checking behaviour;
• method switches;
• unfinished accessible work;
• propagation events.

Then ask where the first sustained change appears. The bin is not a diagnosis by itself; it tells us where to investigate more closely.

Difficulty-adjusted comparison

A late section containing harder questions will naturally produce lower raw accuracy. Therefore, fatigue analysis should compare like with like wherever possible.

Useful comparison pairs include:

• routine algebra early versus routine algebra late;
• standard differentiation early versus standard differentiation late;
• similar candidate-filtering tasks at different positions;
• comparable multi-stage chains under fresh and delayed conditions;
• the same type of first-move classification early and late.

The closer the mathematical demand, the more interpretable the temporal difference becomes.

Position-swap testing

Position-swap testing is a stronger version of relocation. Across two practice sessions, place one structural family early in the first session and late in the second, while a second structural family does the reverse.

This helps reduce the risk that one family simply happened to be easier. If both families perform better when early and worse when late, the evidence for a temporal effect strengthens.

The design does not need to become a formal experiment. The underlying idea is counterbalancing: do not let topic and position always move together.

The protected late-paper zone

Not every operation needs identical protection late. Identify a small set of capabilities that must remain dependable because they carry many marks across topics.

A protected late-paper zone might include:

• sign-sensitive algebra;
• exact-value handoffs;
• substitution fidelity;
• interval and domain filtering;
• model validation before calculus;
• high-propagation derivative or parameter checks;
• clean reset between questions.

The exact set should be personal. Its purpose is to make late robustness concrete. If these carrier operations remain stable, a large part of the paper remains protected even when higher-level reasoning becomes slower.

Fatigue-sensitive operations

Some operations are more sensitive to sustained load for a particular student. The tutor should identify them empirically rather than assume the same list for everyone.

Common candidates include:

• negative signs across long symbolic chains;
• copying parameters between parts;
• deciding when to preserve exactness;
• restoring restrictions after transformations;
• calculator state and bracket entry;
• choosing among two plausible methods;
• interpreting a final result in context;
• maintaining notation across multiple pages of work.

A fatigue-sensitive operation deserves late-position practice even if its fresh accuracy is already excellent.

The minimum viable examination system

When sustained load rises, the student cannot carry an elaborate catalogue of reminders. The late-paper system needs a compact core that survives even when attention is expensive.

A minimum viable examination system might contain five rules:

1. Target first: know what the question wants before choosing operations.
2. Protect the handoff: when a value will travel, keep it clear and trustworthy.
3. Stop dead routes: do not allow one question to consume the whole paper.
4. Check wide nodes: spend verification where many later states depend on one result.
5. Return to the contract: before finishing, confirm the requested answer form and constraints.

The point is not that these five rules are universal. The point is that the student should possess a small durable core rather than fifty fragile instructions.

Alicia’s fatigue failure tree

Alicia’s late errors often look like accuracy problems, but her failure tree may start earlier:

unfamiliar question → longer recognition delay → time debt → faster execution → compressed working → sign-sensitive handoff unprotected → error propagates.

The first useful intervention is not “check more at the end.” It may be faster recognition or earlier stopping on high-latency questions.

Alicia’s fatigue programme should therefore measure upstream delay as carefully as downstream error.

Tricia’s fatigue failure tree

Tricia’s late failure may run differently:

slight uncertainty → additional checking → slower throughput → schedule pressure → more checking because time feels dangerous → final accessible question left untouched.

Her first intervention is checking economics. If one suitable check has already produced enough evidence, further repetition may not be worth the cost.

Kai Kai’s fatigue failure tree

Kai Kai can maintain speed while losing control:

sustained work → reduced willingness to inspect → first familiar cue chosen → route begins quickly → safeguard omitted → high-propagation state travels unchecked.

His repair is not global slowing. It is preserving a very small number of trigger-based governors at critical moments.

Time debt × fatigue interaction matrix

The matrix matters because late-paper collapse is often worst when both variables are high. The student is tired and behind. Improving only one variable can still create substantial benefit.

For some students, preventing large early time debt is the most efficient fatigue intervention available.

Late-paper pacing elasticity

Pacing elasticity describes how much the student can change speed late without losing mathematical control. This is a practical metaphor, not a formal psychometric measure.

A student with low late-paper pacing elasticity may remain accurate at one comfortable pace but deteriorate sharply if required to accelerate. A student with greater elasticity can recover modest lost time without crossing immediately into error-prone execution.

Train elasticity carefully. Use short late-position sets at slightly different paces and observe where accuracy, notation or checking changes. Do not train reckless acceleration.

Late-paper throughput economics

Near the end of a paper, every minute has opportunity cost. Spending a minute checking one low-risk answer means not spending that minute on an incomplete accessible question.

The late-paper question is therefore not “Can I make this answer even more certain?” but “Where will the next minute produce the highest expected mathematical value?”

Possible uses include:

• finishing an already-started accessible question;
• checking a high-propagation result;
• recovering a nearly solved deferred problem;
• restoring a missed final condition;
• attempting a fresh question with a clear first move.

The optimal choice depends on the actual assessment and student state. The principle is that late minutes should be allocated deliberately rather than emotionally.

Skipped accessible marks as a fatigue signal

If a student leaves difficult questions blank, that may simply reflect knowledge or difficulty. If they leave questions blank that they solve easily afterward, the paper may contain a fatigue, pacing or decision problem.

After marking, identify recoverable accessible marks: work the student can complete under fresh conditions but failed to reach or execute during the paper.

Then classify why:

• never reached because of earlier time debt;
• reached but first-move latency was too high;
• started but throughput collapsed;
• method was known but control was lost;
• answer was completed but final constraints were omitted.

This creates a more useful fatigue target than simply “finish faster.”

The return queue for deferred questions

When questions are deferred, the student may create a mental queue. Late in the paper, deciding which one to revisit can itself become expensive.

During practice, use a simple return rule based on state:

1. prefer questions with a visible route and meaningful partial state;
2. prefer high-value completions over completely opaque restarts;
3. preserve enough working that the route can be re-entered quickly;
4. do not return merely because the question feels emotionally unfinished.

The exact navigation strategy must match the real paper. The general goal is to lower late decision cost.

State preservation when leaving a question

A deferred question should be left in a recoverable state. If the student returns ten minutes later and must reconstruct everything, the earlier partial work has low value.

Before leaving, make clear:

• what has been established;
• what remains unknown;
• which route was being attempted;
• which intermediate value is trustworthy;
• where the next step should resume if it becomes visible.

This takes only a small amount of time if the working is already organised and can reduce restart cost substantially.

Late-paper checking triage

Checking should become more selective, not disappear. Rank checks by expected value per second.

The table is a prioritisation model, not a universal mark scheme. The actual assessment controls what work and evidence are required.

Late answer changes require evidence

Fatigued students sometimes distrust a correct answer because it looks unfamiliar, then replace it with a wrong one. Others refuse to change a wrong answer because correcting feels expensive.

Use an evidence rule. Change an answer when a contradiction or stronger independent check indicates that the current state is wrong. Do not change it merely because confidence has fallen with time.

Likewise, if evidence shows the state is wrong, do not preserve it merely because much work depends on it. Correct the origin and trace the descendants.

Calculator-state fatigue

Where calculators are permitted, sustained work can create device-state mistakes: an old stored value remains active, a bracket is missed, the wrong mode persists, or a previous expression is reused unintentionally.

The solution is not constant calculator checking. Protect important numerical handoffs with mathematical expectations. Before accepting a high-value result, ask whether sign, scale and range are plausible.

If the device is producing an answer that contradicts the mathematics, investigate before allowing the value to propagate.

Symbolic-state fatigue

Where non-calculator work or exact symbolic manipulation dominates, the late risk shifts toward sign, factor, fraction and exactness fidelity.

Protect symbolic state with clean line structure at critical transitions. Avoid unnecessary expansion when factored form preserves useful information. Delay approximation while exact structure remains active.

Symbolic economy is not merely elegance. It reduces the number of states that must be carried late.

Fatigue in proof and justification

Late in a paper, students can compress logical steps too aggressively. They know why a statement is true but fail to show the necessary link, or they reverse an implication without checking whether the converse holds.

For proof-like work, the minimum late standard is a visible dependency chain: what fact licenses the next step, and does the conclusion match the target?

Do not respond by writing essays. Preserve the critical logical bridge.

Fatigue in parameter problems

Parameters become dangerous late when the student manipulates symbols without keeping the controlling condition visible.

A simple late safeguard is to state the event the parameter must create before the algebra begins: repeated root, required intersection count, specified gradient, positivity or another syllabus-appropriate condition.

The condition acts as a compass when symbolic work becomes long.

Fatigue in inequalities

Inequalities often fail late at the transition from critical values to interval reasoning. The student solves the boundary equation correctly but treats the answer as isolated numbers instead of a region.

Preserve the answer object. Critical values are inputs; the interval set is the output. A sign chart or number-line representation can externalise the logic where appropriate to the course.

Fatigue in graph transformations

Graph problems can become more expensive late because the student must coordinate algebraic and visual representations. A familiar transformation rule may be recalled incorrectly when several directions or scales compete.

Use invariant checks: where should a known point move, what happens to intercepts, what feature must remain consistent? A quick test point can sometimes reveal a late representation error before it propagates.

Fatigue in multi-part dependencies

Multi-part questions become especially demanding late because the student must remember which earlier results are trustworthy and how they are intended to be reused.

Make the handoff explicit. Label the result from part (a). Preserve exact form if it will travel. If a later contradiction appears, return to the nearest dependency checkpoint rather than rebuilding everything.

The actual treatment of follow-through or method credit depends on the official marking scheme and should not be assumed globally.

The late-paper reliability shell

A reliability shell is the small set of habits wrapped around the student’s mathematics so that performance remains coherent when the internal state is less fresh.

• question reset;
• target and controlling condition;
• default method hierarchy;
• visible high-value handoffs;
• exactness policy;
• stopping rule;
• high-propagation verification;
• final answer-contract check.

The shell should be rehearsed until it is cheap. A safeguard that consumes too much deliberate attention can itself become part of the fatigue problem.

Practice scheduling: quality, mixed load and transfer

Not every study session should be a full-paper endurance test. A balanced programme can rotate among three session jobs.

The exact cadence should depend on the student’s stage, school workload and assessment date. The key is that long sessions test a system that shorter sessions have already improved.

Why maximum-duration practice every day is inefficient

If every session is a full endurance test, students can spend too little time repairing the actual mechanism. They repeatedly observe the same late collapse without changing it.

Long practice also has opportunity cost. Time spent reproducing an already-known fatigue pattern could have been used to stabilise the transition that causes it.

Use long sessions to test transfer. Use shorter targeted sessions to change the system.

The fatigue demotion test

A once-fragile operation should not remain a permanent training priority. When the student has repaired it, test whether it can be demoted from active repair to maintenance.

Demotion evidence might include:

• stable early and late performance across several fresh forms;
• the safeguard activates without prompting;
• late error frequency is low;
• propagation distance remains short;
• the operation survives after a difficult preceding question;
• the repair does not require excessive checking time.

Once demoted, probe it occasionally. If the old failure returns, move it back into active repair temporarily.

Maintenance probes

Maintenance should be small. A few fresh late-position questions can test whether the boundary still holds.

Good maintenance probes vary surface, position and topic context while preserving the mechanism being tested. The student should not know in advance that the old weakness is being inspected.

If the behaviour remains stable, keep the maintenance dose low and spend training time elsewhere.

The final-week fatigue strategy

Close to the examination, the objective is not to invent a new endurance system. It is to confirm that the existing one is stable.

Use fresh faithful practice to check:

• late first-move latency remains acceptable;
• routine carrier skills stay clean;
• high-propagation checkpoints still run;
• time debt is contained early;
• exactness and final constraints survive;
• deferred questions can be re-entered efficiently;
• one difficult event does not contaminate several later questions.

Final preparation should increasingly reinforce the reliable shell rather than add unfamiliar protocols. Broader revision and recovery planning belong to their own systems; this article’s job remains late-paper mathematical performance.

The full-paper fatigue reliability gate

A student is approaching robust late-paper readiness when several fresh, faithful practice papers show the following pattern:

• early-to-late accuracy change is within a personally acceptable range;
• late first-move latency does not explode on ordinary questions;
• throughput remains sufficient for the actual paper;
• high-value checking survives without excessive time cost;
• propagation distance remains contained;
• one difficult question does not create a long recovery tail;
• routine marks remain protected even on less friendly paper orders.

No single number defines readiness globally because assessments differ. The gate is a pattern of repeated functioning under the conditions the student will actually face.

Additional questions students ask about long mathematics examinations

How many full maths papers should I do?

There is no universal number. Use enough fresh full papers to test transfer and reveal recurring late patterns, but spend targeted sessions repairing those patterns. Repeating full papers without changing the mechanism can have diminishing value.

Why do I lose easy marks near the end?

Possible causes include time debt, compressed working, reduced checking, state loss or genuine duration effects. Relocate similar easy or medium operations earlier and later to identify what changes.

How do I stop rushing the final questions?

Prevent the rush upstream. Find the questions that create time debt, use stopping rules, reduce decision latency and make routine methods cheaper. Once late, prioritise accessible mathematical value rather than accelerating every operation indiscriminately.

How can I stay accurate in a long maths exam?

Build clean carrier skills, preserve a small set of high-value safeguards, practise known skills late, control time debt and use independent checks at high-propagation states. Accuracy late is a system outcome, not only a concentration instruction.

Should I slow down when I feel tired?

Not globally. Slow the fragile transition if evidence shows accuracy is at risk, while keeping routine safe work efficient. A blanket slowdown can damage completion. Use the first weak curve to decide what needs protection.

Why can I finish practice worksheets but not full papers?

Worksheets may reduce recognition switching, decision load, duration and paper-navigation demands. Build from topical capability to mixed work, then extended work and full-paper transfer instead of assuming worksheet speed automatically transfers.

The fatigue-curve operating loop

The complete operating loop is:

Baseline → Match → Relocate → Observe the first weak curve → Remove time debt → Reduce operating cost → Extend duration → Stress the late state → Retest in a full paper → Maintain.

Baseline the mathematics while fresh. Match comparable tasks so position can be interpreted fairly. Relocate them to expose temporal effects. Find the first variable that changes. Remove avoidable time debt. Make routine mathematics cheaper. Extend duration only after the mechanism is visible. Stress-test late performance, then return to authentic papers. Once stable, move the repair to maintenance.

The loop makes stamina measurable without pretending that every late error has the same cause.