Three checkpoints in this guide
Follow the full walkthrough in order, or jump directly to one of its main sections.
A losing trade does not prove that a good setup failed, and a winning trade does not prove that a bad decision was sound. The outcome is evidence about a distribution; it is not a complete diagnosis of the decision. A useful post-trade review asks what was knowable before entry, whether the documented rule was followed, whether execution matched the plan, and which explanation remains only a hypothesis.
This guide provides five audit buckets, an “unknown” escape valve, and a repeatable diagnosis record. It deliberately removes unsupported percentages for how often each failure occurs, fixed trade-count rules, and chart-reading shortcuts that assign causes after one loss.
Method checked September 9, 2026. Baron and Hershey's outcome-bias experiments and Fischhoff's hindsight study support separating decision quality from known outcomes. The CFA Institute's investment model validation guide supports out-of-sample and time-series validation for strategy-level changes. These are general decision and model-validation sources, not estimates of retail-trading failure-mode frequencies.
Quick answer: compare the trade with the rule and evidence that existed before the result. If the rule was followed and no process defect is documented, record a rule-compliant loss—not a guessed cause. Change a criterion only after a pre-written hypothesis is tested across all relevant trades and survives data that were not used to invent it.
Separate the Decision, Execution, and Outcome
A setup is a selection rule. Execution is what the order and trader actually did. Outcome is the market path after entry. Mixing these three makes every loss look like a setup defect and every win look like validation.
| Layer | Question | Best evidence |
|---|---|---|
| Decision | Was the trade eligible under the rule version active at entry? | Timestamped plan, checklist, setup tag, pre-entry note or image |
| Execution | Did the actual order match planned price, size, stop, target, and order type? | Orders, fills, modifications, fees, slippage, timestamps |
| Outcome | What path occurred, and was the result inside the strategy's observed range? | Market data, realized P&L in R, excursion and exit record |
Review the decision using only information available at the time. Then reveal the outcome. This simple order reduces the temptation to rewrite the setup around what the next candles happened to do. The journal-mistakes guide covers missing plans, inconsistent tags, and edited notes that make this audit impossible.
The Five Setup Failure Audit Buckets
These buckets organize evidence; they are not universal causes with known percentages. A trade may contain more than one issue. If the record cannot support a classification, use Unknown instead of forcing a story.
Failure Mode 1: Rule-Compliant Losing Outcome
The setup definition was met, the trade was authorized by the plan, execution stayed within tolerance, and no documented exclusion applied. The trade still lost. The defensible label is rule-compliant loss. Calling it “pure variance” would be stronger than the evidence: one outcome cannot identify the generating mechanism.
Evidence to Record
- Rule version and each eligibility field as it appeared before entry.
- Planned versus actual entry, stop, size, target, and risk.
- Known costs and whether the loss was within the plan's risk tolerance.
- Any information that arrived after entry, clearly separated from pre-entry facts.
Response: do not add a new criterion from this trade alone. Add the observation to the strategy distribution and continue the pre-committed review plan. Calculate the strategy-level expected value with the expectancy workflow, not with a verdict attached to one row.
Failure Mode 2: Eligibility or Criteria Deviation
The active rule would not have authorized the trade, or the reviewer cannot reproduce the eligibility decision. Examples include a required field left blank, a setup tag applied after the outcome, a minimum condition marked “close enough,” or a rule version that changed without a date.
Diagnostic Indicators
- A checklist field conflicts with the written plan.
- The trade lacks evidence for a required criterion.
- The rule used in review differs from the rule active at entry.
- Exceptions are remembered but were not logged before execution.
Response: fix compliance and evidence capture before changing the strategy. Tightening the rule does not solve a process that cannot prove which rule it followed.
Failure Mode 3: Execution or Market-Friction Deviation
The setup was eligible, but the filled trade differed materially from the plan. Possible evidence includes a late or partial fill, wrong quantity, wrong contract, missing stop, rejected modification, spread, slippage, commission, or an exit that did not follow the recorded rule.
What Not to Infer
A stop reached soon after entry does not by itself prove the entry was late. A hypothetical better price does not prove that price was obtainable. Diagnose execution from the order and fill trail, not from a cleaner point visible after the chart is complete.
Response: define an execution tolerance, identify the actual deviation, and choose a process fix—order template, instrument check, alert, or checklist. Keep the setup criteria unchanged unless a separate aggregate test supports a setup change.
Failure Mode 4: A Known Context Rule Was Missed
Context belongs in this bucket only when the condition and response were defined before the trade. For example, the plan may prohibit new entries around a specified event, require a particular session, or exclude a volatility state measured by a named rule.
Evidence to Require
- The context field has a reproducible definition and source.
- The data were available before entry.
- The active plan stated what to do when the condition was present.
- The review does not substitute a new chart narrative for the original rule.
Response: repair the context check or exception workflow. If the condition was not part of the plan, record it as a candidate hypothesis under the next bucket—not as a proven missed rule.
Failure Mode 5: Strategy or Regime Hypothesis
A cluster of rule-compliant losses may justify asking whether the strategy's opportunity set, market state, costs, or implementation has changed. It does not prove a “regime mismatch.” Regimes need operational definitions; otherwise the label merely renames a drawdown.
Build a Testable Hypothesis
- Name the condition and how it is measured without looking at future data.
- Explain why it could affect the setup's mechanism or execution.
- Specify the comparison, metrics, and decision before testing.
- Include all relevant observations, costs, and failed versions.
- Reserve later data or use an appropriate time-series validation design.
Response: continue, reduce, pause, or retire the strategy only under the risk and review rules you set in advance. The strategy-abandonment guide separates a tolerable drawdown from a breached premise and a broken process.
Use “Unknown” When the Evidence Is Missing
Unknown is not a failed review. It identifies the field or timestamp the next trade must preserve. Common reasons include missing pre-entry evidence, incomplete fill data, ambiguous timezone, unmatched account or instrument, an edited rule with no version history, and a manual import that omitted fees.
Do not distribute unknown trades among the five buckets to make the dashboard add up. Track the unknown share. If it is high, the priority is data quality—not strategy optimization.
Validate a New Criterion Without a Magic Trade Count
There is no universal “30 trades” threshold or win-rate gap that proves a condition predicts losses. Required evidence depends on outcome dispersion, dependence between trades, base rate, costs, search breadth, and the consequence of being wrong.
- Write the candidate rule. Define the characteristic, expected direction, metric, comparison, and proposed action.
- Audit the data. Confirm account, instrument, timezone, duplicates, fees, and rule version.
- Use the complete comparison set. Include trades with and without the characteristic and retain missing states.
- Measure more than win rate. Compare expectancy, payoff distribution, drawdown, outliers, cost, and opportunity frequency.
- Count every test. Multiple tried definitions make the best historical result less persuasive.
- Freeze and validate. Apply the unchanged rule to data that were not used to create it.
For systematic strategies, the backtest-versus-live guide covers leakage, implementation gaps, and why a historical fit is not an execution result.
The Failure Diagnosis Framework
| Step | Check | Output |
|---|---|---|
| 1. Integrity | Reconcile account, instrument, timestamps, quantities, fees, and duplicates | Usable record or Unknown |
| 2. Eligibility | Compare pre-entry evidence with the active rule version | Compliant or criteria deviation |
| 3. Execution | Compare plan, orders, fills, modifications, and exit | Within tolerance or execution deviation |
| 4. Context | Test only context exclusions documented before entry | No miss or known context-rule miss |
| 5. Outcome | Record R result, excursion, costs, and path without assigning causality | Rule-compliant loss record |
| 6. Hypothesis | If a pattern is suspected, pre-write an aggregate validation plan | Candidate strategy/regime test |
Assign a confidence level and cite the evidence. One trade can establish that a checklist was blank or an order size was wrong. It usually cannot establish that a new market condition predicts failure.
Aggregate the Audit Without Inventing a Normal Distribution
Review on a pre-committed horizon suited to your trade frequency. Report counts and rates for every bucket, including Unknown, but do not compare them with an invented “typical retail” distribution. The relevant baseline is the same strategy, rule version, account scope, and data process.
Investigate changes alongside exposure and opportunity. A higher number of execution deviations may reflect more trades, not a worse process. A cluster of rule-compliant losses may be compatible with the observed payoff distribution, or it may motivate a strategy hypothesis; the audit alone does not choose between them.
Trader's Second Brain is our product. It can preserve imported executions, setup and context tags, notes, screenshots, and review fields so the decision/execution/outcome split remains auditable. Its canonical source registry currently recognizes 330 structured broker, exchange, platform, and prop-export profiles. That is parser coverage, not a guarantee that every source contains every field.
Check your exact route in the live import directory, validate a representative file, and reconcile the imported record before diagnosing. Full Access includes review and rule-tracking workflows and offers a lifetime-access route; verify current terms rather than relying on a copied price.
Who Should Prioritize Failure Analysis
- Traders who change rules after individual losses: separate a process defect from an unfavorable result before adding complexity.
- Teams with inconsistent reviews: use the same fields, evidence hierarchy, and allowed labels.
- Traders who suspect execution leakage: compare plans with fills and costs rather than judging the completed chart.
- Strategies in a drawdown: distinguish data, compliance, execution, and hypothesis work before changing risk.
- Prop-program traders: keep program-rule breaches separate from whether the setup itself had positive expectancy.
Any response must stay inside the account's loss limits and the trader's risk capacity. The risk-management framework covers exposure, stop logic, drawdown capacity, and when investigation should not become another live experiment.
Methodology Note
The five buckets are an editorial audit taxonomy, not a validated universal distribution of trading losses. Outcome- and hindsight-bias research motivates the evidence order, but does not quantify trader behavior. CFA's model-validation guidance applies most directly to formal investment models; this guide borrows its core discipline of independent validation without claiming that discretionary trade review is equivalent to institutional model governance.
No fixed mode percentage, stop-timing signal, sample count, win-rate gap, regime duration, or review interval is presented as universal. Those claims require strategy-specific data and an analysis that accounts for dependence, costs, multiple testing, and changing conditions.
Final Verdict: Diagnose the Process Before You Rewrite the Setup
For one loss, prove what you can: data integrity, eligibility, execution, and known context compliance. Record the outcome without pretending its cause is observable from the chart alone. Use Unknown when the evidence is missing.
For a strategy change, raise the standard. State the hypothesis, use the complete comparison set, log every test, freeze the candidate rule, and validate it on later data. The best failure analysis does not explain every loss; it prevents an unsupported story from becoming a permanent trading rule.
