SG EducationOS 12-Week Public Backtest Demonstration Template v0.1

(Intake → Plan → Runs → Outcomes → Calibration → Lessons)

Version: CivOS Unified Spec v1.x
Scope: Singapore; portable to any City/Nation instance

0) What this page is

A public, repeatable case template that proves the system works.

It must be possible for any parent/tutor/LLM to run the same structure.

1) Case Header (Copy-Paste)

			
[Backtest Case v0.1]
Case ID: SG-EDU-BT-YYYY-### 
Student Profile: (anonymous: Age/Level only)
Track: (Primary / Secondary / JC)
Primary Lane: (English / Math / Science / Humanities)
Exam Horizon: (PSLE / O / N / A / None)
TTC at Start: (1–2w / 4–6w / 8–12w / 12w+)
Start Date: (YYYY-MM-DD)
End Date: (YYYY-MM-DD)

		

2) Intake Snapshot (Week 0)

			
[SG-EDU Intake Snapshot]
Level:
School type:
Subject priority:
Current results:
Time available (hrs/week):
Tuition hours:
Parent support:
State Sensors:
Phase (P0/P1/P2/P3):
Backlog (Low/Med/High):
Timed stability (Stable/Shaky/Collapse):
Goal:
Target grade:
Career direction (optional):

		

3) Baseline Diagnostics (Week 0)

3.1 One-Line State Summary

Stage × Phase × TTC × Backlog × Timed Stability

3.2 Primary Failure Mode (pick ONE)

Choose from:

FM-R Retrieval weakness (knows it, can’t produce)
FM-C Concept gap (doesn’t understand)
FM-S Structure failure (writing / method)
FM-T Timed-load collapse (performance drops under time)
FM-X Execution failure (plan not followed)
FM-TR Transfer failure (can do drills, fails new questions)

3.3 Failure Mode Trace (Required)

			
Z0: (symptom)
→ Z1: (plan/ops issue)
→ Z2: (timed or transfer breakdown)
→ Z3: (lane instability outcome)
Repair:
(3 items only)

		

4) The 12-Week Plan (Locked Output)

4.1 This Week Plan (Week 1 Template)

(3–5 actions only, measurable)

			
Week 1 Actions:
A1:
A2:
A3:
A4 (optional):
A5 (optional):
Execution target: ≥80%

		

4.2 4-Week Targets (Milestones)

			
Week 4 target:
- Simulation pass rate:
- Error recurrence:
- Output quality marker:

4.3 12-Week Target (Outcome Definition)

			
Week 12 target:
- Grade/score band:
- Timed stability:
- Transfer integrity:

5) Simulation Ladder (ExamSim Block)

Select based on Timed Stability:

If Collapse:

Micro-set daily (10–15 min)
1.5× time (Week 1–2)
1.2× time (Week 3–4)
Full time (Week 5+)

If Shaky:

Mini timed set 2×/week
Full simulation 1×/week (Week 3+)

If Stable:

Full simulation 1×/week
Stretch set 1×/week

Pass threshold rule:

Week 2: ≥60%
Week 4: ≥70%
Week 8: ≥75%
Week 12: ≥80%
(Adjust via calibration)

6) Weekly Log Table (Weeks 1–12)

Fill this every week. This is the backtest dataset.

Week	Exec %	Sim Score %	Rewrite Count	Recurring Errors	Phase (P0–P3)	Notes

Minimal. Repeatable. Auditable.

7) HGW Forecast Panel (Week 0 → Week 12)

7.1 Scoring (0–5 each)

Hope

0–1 unclear target
2–3 partial target + weak alignment
4–5 clear target + aligned plan

Grind

based on Exec % and hrs/week
<50% ⇒ ≤2
70–85% ⇒ 3–4
> 90% ⇒ 5

Wisdom

mistake log + rewrite loop + reflection

7.2 Forecast Prints (Required)

			
Week 0 Forecast:
4w: H/G/W = _/_/_  (direction)
12w: H/G/W = _/_/_ (direction)
Risk signatures: (if any)

8) Calibration Updates (Week 4 and Week 8)

8.1 Calibration Trigger Rules

Overprediction: forecast ↑ but sim stagnates 2 weeks
Underprediction: sim ↑ and recurring errors ↓ and exec ≥85%
Drift spike: sim <60% twice or exec <50% twice

8.2 Calibration Output (Required)

			
Calibration Adjustment (Week 4):
Hope weight: +1/0/-1
Execution threshold: __% → __%
Simulation frequency: __ → __
Scope: Expand / Hold / Cut

		

Repeat at Week 8.

9) Final Outcome (Week 12)

9.1 Outcome Snapshot

			
Final Phase:
Final Timed Stability:
Final Simulation pass rate:
Final recurring error rate:
Final grade/score:

		

9.2 Did the system hit target?

✅ Hit / ⚠ Partial / ❌ Miss

9.3 Why (Causal, not emotional)

			
What caused success/failure:
- Factor 1:
- Factor 2:
- Factor 3:

10) Lessons → System Improvement (Promote to Canonical)

This is the point: each case improves the OS.

10.1 Promote a new rule if needed

			
New rule candidate:
Trigger:
Action:
Expected effect:

10.2 Update the Failure Atlas / Routing Playbook

Add a new collapse signature if discovered
Add a new repair loop if validated

11) “Publish Mode” Safety (Privacy + Integrity)

Remove identifying details
Show only level band + lane + outcomes
Keep logs numeric
Keep the structure consistent

This makes it scalable.

Start Here:

Start here if you want the full sequence:

Vocabulary OS Series Index:
https://edukatesg.com/vocabulary-os-series-index/

Fence English Learning System:

eduKateSG Learning Systems: