Rule Adherence Score — Grade Process, Not P&L

1 / Why this question showed up

2 / What “rule adherence” means here

Rule adherence is the share (or weighted degree) to which a trade follows predeclared entry, risk, management, exit, and state rules. The rules have to be observable: you — or another reviewer — should reach the same pass/fail from the same record.

“Take only good setups” cannot be scored. “Long only after the 15-minute close above the opening range, before 11:00, stop below the trigger low” can. If a score keeps causing debate, fix the written rule before chasing a higher percentage.

3 / Start with a minimum viable scorecard

The useful scorecard is short enough to finish on every trade. Begin with four to six rules that protect edge or account risk — not a duplicate of every sentence in the plan, and not cosmetic preferences scored as if they equal a hard risk limit.

1. Write one observable pass condition per critical rule.
2. State when evidence must be captured (before entry, during management, after exit).
3. Define whether a violation makes the whole trade off-plan.
4. Test the wording on a handful of historical examples for ambiguity.
5. Freeze the wording for a review cycle before comparing scores across weeks.

4 / Binary model, graded model, critical fails

Binary: pass or fail

Mark each applicable rule 1 if followed, 0 if violated. Non-applicable rules are N/A — not automatic passes.

Arithmetic illustration only: if five of six applicable rules pass, the score is 5 ÷ 6 × 100 ≈ 83.3%. That is definitional math, not a claim about any book.

Binary scoring is fast and repeatable for clear constraints (“risk ≤ 0.5%”, “no entry after 11:00”, “never widen the stop”). Its weakness is severity: missing a screenshot and doubling risk both score zero unless a critical-fail rule distinguishes them.

Critical fails without hiding the component score

Predeclare a small set of safety violations that make the trade off-plan regardless of the percentage — for example exceeding maximum risk, trading after a daily stop, removing the protective stop, or taking an unapproved setup. Keep the numerical score as detail; classify the trade off-plan. Do not invent a “critical” rule after a loss.

Graded / weighted (only where partial means something)

Some decisions are not naturally binary (documentation complete / partial / absent; exit discretion inside a defined band). A graded model can assign 0, 0.5, or 1 if each level has an observable definition.

• Prefer binary when rules are objective and speed matters.
• Use graded scoring only where partial compliance is well defined.
• Use weights sparingly and write down why one rule matters more.
• Keep critical risk limits as hard fails even in a graded model.
• Do not retune weights to make a disappointing month look disciplined.

5 / Trade score vs period score

A week’s adherence should aggregate applicable rule decisions — not merely average trade percentages when trades have different numbers of applicable rules.

Also report the share of trades that were fully on-plan. Component adherence and fully-on-plan share answer different questions: frequency of rule hits versus complete execution. One can be high while the other is mediocre if small violations are spread across many entries.

6 / Compare on-plan and off-plan groups — carefully

Once trades are classified, performance can be reviewed separately. Expectancy in R keeps different cash sizes comparable:

Compute that independently for on-plan and off-plan groups when you have enough trades to bother. Include costs and slippage in realized R. Report count beside expectancy so a three-trade subgroup is not treated as established.

Related reading on how people summarize results (without turning this memo into a metrics lecture): Performance metrics.

7 / Do not rewrite the strategy because off-plan trades lost

An off-plan loss is weak evidence about the planned strategy — the strategy was not executed. If a breakout rule requires a close above resistance but the trader anticipates the close and loses, changing the breakout stop or target from that loss uses contaminated evidence.

• On-plan trades test the strategy.
• Off-plan trades test the cost and triggers of deviation.

Fix execution from the second group; evaluate strategy rules primarily from the first. That does not mean ignoring the loss — record its R impact, review the trigger, install a process control. Strategy changes belong later, when on-plan evidence (and a stable rule version) actually supports them. Sample-size humility belongs here too; see the broader validation mindset in Validation gauntlet.

8 / Opportunity quality ≠ execution quality

Trade-only adherence can look perfect while valid setups are skipped. It can also punish restraint if every no-trade is scored as a miss. A fuller review starts from qualified opportunities: moments that met the plan whether or not a position was opened.

Define the scan universe and hours first, or the opportunity denominator becomes arbitrary. Keep setup quality and execution quality in separate fields.

9 / Compact mistake taxonomy (and careful “cost”)

The adherence score says whether a rule failed. A small mistake category says what kind of control should prevent recurrence. Prefer one primary category plus one specific behavior (“Risk → moved-stop”) over a pile of emotional labels.

• Selection — entered without required setup/context
• Timing — early, late, or outside approved session
• Risk — oversize, correlated exposure, missing stop, daily limit breach
• Management — unauthorized add, moved stop, unplanned partial
• Exit — fear exit, ignored invalidation, missed time rule
• Process state — revenge, FOMO, fatigue, checklist bypass
• Documentation — missing evidence that blocks reliable review

Do not encode the outcome (“bad loss”, “lucky win”) as a process category. On cost: realized R of mistaken trades is an impact total, not automatically the causal cost — an on-plan version of the same trade might also have lost. When a compliant alternative can be reconstructed without hindsight (e.g. stop should have closed at −1R but was widened to −2.2R), the excess is roughly the execution gap. When it cannot, report impact without claiming a counterfactual winner. Risk framing that sits next to this habit: Surviving risk.

10 / A practical review cadence

1. After each trade — score while memory is fresh; attach evidence; classify deviation. Do not redesign the strategy in the heat of the fill.
2. Daily — critical failures, daily loss status, unlogged opportunities, stop/go. Account protection, not strategy conclusions.
3. Weekly — component adherence, fully on-plan share, on-plan vs off-plan R (with counts), opportunity capture, highest-impact recurring mistake. Commit to one observable control for next week.
4. Monthly / per sample — rolling adherence by stable strategy version; audit score consistency; merge duplicate mistake tags; ask whether a repeatedly failed rule is unclear, impractical, or ignored. Strategy edits belong here only when on-plan evidence supports them.

A 100% adherence week can still lose money. A 60% adherence week can still make money. Process and outcome can diverge sharply in short windows; relate them only over a broad enough sample — and do not lower the pass threshold after a bad week, add rules solely because one trade lost, or remove a valid rule because breaking it produced a winner.

11 / What this page holds vs what it does not claim

12 / Continue on the site