Three checkpoints in this guide
Follow the full walkthrough in order, or jump directly to one of its main sections.
The best way to review trades is to rebuild the decision before judging the outcome. Freeze the date range and plan version, reconcile executions with the broker or venue, preserve the evidence available at entry, mark each applicable rule as followed, violated, or unknown, compare like-for-like trades after costs, and finish with one dated action plus a later recheck. A review should be able to say “insufficient evidence”; otherwise it will manufacture lessons from incomplete data.
1. A Trade Review Starts With a Frozen Contract
A journal records events. A review connects those events to the plan and asks what the evidence supports. Before opening the equity curve or choosing a best and worst trade, write the scope:
- account, venue, strategy, and setup version;
- workspace timezone and session boundary;
- start and end timestamps;
- the plan and risk rules effective during that period;
- cost fields included and known data gaps;
- the question the review is allowed to answer.
Freezing scope prevents a familiar failure: changing the date range, setup labels, or exclusions until a clean story appears. If the plan changed during the period, split the review by version. Historical trades keep the authority that governed them; the current plan must not rewrite the past.
2. Separate Session Capture From Periodic Analysis
Capture and analysis serve different purposes. After the session, preserve facts that become harder to reconstruct: plan version, setup state, screenshots, order events, notes, exceptions, and the evidence visible at the decision. Do not force a strategic conclusion while the outcome is still emotionally or operationally fresh.
Run the deeper review when enough comparable observations exist and the evidence can still be reconstructed. Daily, weekly, and monthly can be useful labels, but none is a universal duration or schedule. An active intraday strategy may support frequent synthesis; a low-frequency strategy may need a longer window. The cadence follows evidence density and operational risk, not a generic timer.
The journal-building guide explains the minimum record that makes this later reconstruction possible.
3. Gather One Review Packet Before Interpreting It
A complete packet keeps conclusions traceable. Gather these layers from their authoritative sources:
Execution
Orders, fills, fees, positions, cancellations, corrections, and current account state from the broker or venue.
Plan
The dated strategy, setup, session, risk, stop, and exception rules that applied at the decision.
Decision evidence
What was known, missing, or contradictory at entry—before the result could change the story.
Review context
Trader-authored notes, screenshots, market labels, and explicit reflections, clearly separated from system facts.
If one layer is missing, say so. Notes can explain what the trader recorded, but they cannot replace fills. A broker statement can prove execution, but it may not prove setup permission. A calculated metric can summarize a cohort, but it cannot invent the decision-time reason.
4. Reconcile the Trade Record Before Calculating Performance
Reviewing an unreconciled log creates precise errors. Match imported or entered trades to the authoritative execution record. Resolve duplicate fills, partial exits, scale-ins, fees, currency conversion, reversals, overnight assignment, and corrections. Keep unresolved differences visible instead of silently dropping them.
- Confirm identity. Match account, instrument, order, execution, direction, quantity, and timestamp.
- Build the position. Define how fills become one trade or several decisions.
- Attach costs. Preserve commissions, venue fees, financing, spread or slippage evidence where available.
- Resolve boundaries. Apply the workspace timezone and declared session rule consistently.
- Publish coverage. Count matched, missing, ambiguous, and excluded records with reasons.
Only then calculate P&L, expectancy, drawdown, or setup results. The performance-analysis guide covers metric denominators and cost-complete comparisons in more depth.
5. Review Process Separately From Outcome
A winner can violate the plan. A loser can follow it. That does not make outcome irrelevant; it means process and outcome answer different questions. Score the process first, without looking for a grade that makes the P&L feel deserved.
For every rule that could have applied, use explicit states:
- Followed: the required action is supported by the record.
- Violated: the record supports a specific departure.
- Not applicable: the rule did not govern this decision.
- Unknown: the evidence cannot establish the state.
An A/B/C/D grade can be a convenient summary if its components are defined for one plan version, but it is not a universal standard. Publish the rule-level evidence, coverage, and unknown count beside any label. Do not turn trader-authored words such as “fear,” “revenge,” or “confidence” into facts unless the trader actually recorded them.
6. Compare Like With Like
Once the record is reconciled, group trades by a question that can change a decision: setup version, session, instrument, direction, holding logic, event condition, or plan adherence. Keep the comparison narrow enough that the groups mean the same thing.
Every comparison should show:
- trade and active-session counts;
- gross and net result with recorded cost coverage;
- outcome distribution rather than win rate alone;
- missing fields and exclusions;
- whether the slice was chosen before or after seeing results;
- a later, untouched window if the result will change the plan.
Do not cut the worst setups and increase exposure to the best ones from one retrospective ranking. A weak slice may reflect a regime, changed definition, incomplete sample, concentration, or chance. Treat the finding as a candidate until it survives a later check. The setup-confluence audit helps separate setup identity, filters, triggers, descriptive scores, and risk authority.
7. Review Winners, Losers, and Unknowns
Review all three outcome classes. Winners show whether the repeatable process was present or whether a deviation happened to pay. Losers show whether the plan was executed and whether the result sits inside the strategy's known distribution. Breakeven, cancelled, corrected, and unresolved trades can reveal execution or data problems that a winners-versus-losers filter misses.
Choose representative cases from a defined cohort, not only the most dramatic chart. For each case, reconstruct:
- the plan and setup version;
- the evidence known at the decision;
- the order and risk state;
- the action taken and any recorded exception;
- the resulting execution and costs;
- what remains unknown.
This makes a good loss useful without romanticizing it, and a bad-process win visible without pretending the outcome did not happen.
8. Promote a Pattern Only When the Evidence Earns It
A pattern needs a defined subject, comparison, minimum coverage, and limitation. “Afternoon trades are bad,” “Fridays kill performance,” or “I overtrade after losses” are hypotheses until the relevant trades are comparable and the alternative explanations have been checked.
Observation
A reproducible count, metric, or rule state in the frozen scope.
Comparison
A named baseline using the same strategy, account, and compatible market conditions.
Limitation
Missing data, sparse cohorts, changed definitions, selection, or confounding that narrows the claim.
Decision
Keep, investigate, or test one plan change—never automatic certainty.
The strongest review is often narrower than the first story. It may support “entries outside the active session were net negative in this cost-complete period” while refusing to claim why the trader took them or what every future session will do.
9. Fix the Review Failures That Create False Lessons
A review can be consistent and still be wrong. Audit the review process itself for these failure modes:
- Outcome-first grading. The result is known before the plan and decision evidence are reconstructed, so winners receive charitable explanations and losers receive harsher labels.
- Cherry-picked cases. Only the largest winner and loss are inspected, while ordinary trades and contradictory cases disappear.
- Label drift. A setup, session, or rule name changes meaning during the period, but all records are grouped as if the definition stayed fixed.
- Incomplete opportunity set. The review contains executed trades but not sessions with no trade, early stops, blocked setups, or apparently valid opportunities that were skipped.
- Cost leakage. Gross results are compared with net results, or missing fees are silently treated as zero.
- Multiple interventions. The trader changes the setup, session, risk, and execution rule together, making the next result impossible to attribute.
- Authority confusion. Journal labels are allowed to override live broker, venue, or account state.
Track these gaps as review debt. Each debt item should name the missing record or unresolved definition, the affected period, whether it blocks a conclusion, and the next recovery action. A review with visible debt can still support a narrow finding. A review that hides debt creates false confidence.
Do not punish the trader for unknown data. Separate “the rule was violated” from “the evidence does not show whether the rule was followed.” That distinction keeps the denominator honest and tells you whether the next improvement belongs in trading behavior, plan design, import coverage, or the review workflow itself.
10. Finish With One Action and a Recheck
A review without a decision becomes a report archive. A review with many simultaneous changes makes the next result uninterpretable. Choose one action linked to the clearest supported finding.
Write it as a measurement contract:
- Finding: the exact observation and supporting cohort;
- Action: the trader behavior to add, remove, or preserve;
- Condition: when the action applies;
- Authority: the plan, account, broker, or venue rule that controls it;
- Evidence: what must be recorded;
- Recheck: the later opportunities and cutoff that will evaluate it;
- Stop condition: the safety or data-failure state that ends the test.
Version the plan before the change takes effect. The discipline guide shows how to make the condition, action, authority, evidence, and exception executable during the session.
11. A Copyable Trade-Review Template
Use the same structure each time so reviews can be compared without forcing the same conclusion:
Scope and coverage
Account, strategy, plan version, period, timezone, matched trades, missing trades, excluded records, and cost coverage.
Deterministic results
Core metrics, distributions, drawdown, setup/session slices, and exact denominators.
Process evidence
Followed, violated, not applicable, and unknown rules with supporting trade references.
Decision record
One finding, alternatives checked, limitation, action, effective date, recheck, and stop condition.
Include at least one strong-process trade, one violated or ambiguous decision if present, and one result that challenges the leading story. That last case prevents the review from becoming a search for confirmation.
Before accepting the action, write the strongest alternative explanation in one sentence and identify the record that would distinguish it from the leading interpretation. If no available or future observation could change the conclusion, the review is not testing a finding; it is defending a belief. Store the rejected alternatives and their evidence beside the decision so a later reviewer can see why the plan changed.
12. Reviewing a Low-Frequency Strategy
Do not manufacture statistical confidence from a short calendar period. With few trades, emphasize reconstruction, data quality, setup permission, execution, and whether the planned opportunity set is complete. Keep results descriptive and roll comparable trades forward across a stable strategy version.
The review still has value: it can catch missing orders, unclear rules, inconsistent tags, unrecorded costs, and plan deviations. It simply may not support a performance ranking yet. State the evidence needed for the next decision rather than adopting a universal trade-count threshold.
13. Let Deterministic Analytics Hold the Numbers and Coach Explain Them
AI should not recompute P&L, expectancy, drawdown, setup grades, rule states, or the membership of a comparison from prose. Those belong to deterministic systems with visible scope and denominators. AI Coach is most powerful after that work is complete.
In Trader’s Second Brain, Trading Plan preserves dated rule authority; Journal holds imported or entered trade evidence; the analytics and Leak Map layers calculate observations; Daily Review closes the period; and Current Focus carries one accepted evidence-linked action into a later recheck.
Coach is the high-leverage decision layer. It connects the selected plan, authoritative metrics, supporting trades, coverage limits, and next action into an intelligible explanation. It can show why the evidence supports a conclusion, what remains unmeasured, and what deserves the next clean test. Its refusal to fabricate psychology, causality, missing facts, or recalculated metrics makes the guidance stronger: the reasoning stays anchored to the trader’s own evidence.
TSB has processed 600K+ imported trades across its import history, and its source registry recognizes 332 exact broker, exchange, platform, and prop-export profiles. Those values mean imported trades and recognized source routes—not users, guaranteed compatibility, or trades analyzed by Coach.
Open the review loop Reconcile the journal Ask Coach to connect the evidence
The Bottom Line
A winning trade review is not a fixed weekly timer or a motivational grade. It is a traceable chain from authoritative execution and dated plan rules to comparable metrics, rule evidence, one decision, and a later recheck. Review winners, losers, and unknowns; publish coverage; and let the process refuse conclusions the data cannot support.
Disclosure: Trader’s Second Brain is our product. Its Trading Plan fields, Journal evidence, deterministic analytics, Daily Review, Current Focus, Coach evidence boundary, and canonical public-truth values were checked against the local codebase on September 10, 2026. This guide is educational and does not provide investment advice or promise that a review cadence or plan change will improve returns. See our editorial methodology.
