Review habits: separate process, execution, and outcome

The short answer

Review in layers. First reconcile balances, fills, costs, and exposure. Then score the decision against its timestamped plan, score execution against the available route and order evidence, and record the outcome without letting it overwrite either score. Aggregate comparable records on a scheduled cadence before changing strategy rules.

The purpose is to produce testable action—not a story that makes the latest result feel inevitable. Profitable rule breaks remain rule breaks, and valid losing trades remain valid losses.

Reconcile facts before interpreting them

Join wallet balance deltas, signatures, partial fills, failed attempts, fees, token accounts, open orders, delegated approvals, transfers, and residual positions. Record the data source and retrieval time. Delayed indexers or an unexplained balance should remain unresolved rather than being forced into the preferred narrative.

Compare planned and actual entry, exit, size, route, minimum output, timing, and costs. This immediate reconciliation is operational work; it should not wait for a weekly review while capital or permissions may still be exposed.

Use separate review layers

• Decision quality: Did the information available at decision time satisfy the setup, risk, and security rules?
• Execution quality: Did route, size, order settings, monitoring, and fallback match the plan and available conditions?
• Rule adherence: Were entry, exit, session, and change-control boundaries followed?
• Data quality: Which inputs were verified, delayed, missing, stale, or learned only after the decision?
• Outcome: What net capital change, path, adverse excursion, and residual exposure actually occurred?

Keep the labels independent. The expected-value guide shows why one favorable or unfavorable branch cannot validate the decision model.

Give each cadence a different job

• After the trade: reconcile the economic position and flag any security, data, or execution issue requiring immediate action.
• After the session: classify trades and no-trades, rule breaks, interruptions, fatigue, urgency, costs, and session-limit events.
• Weekly or strategy review: group a defined cohort, examine distributions and repeated failure modes, and close prior action items.
• Scheduled change review: approve, reject, test, or roll back strategy changes outside live risk.

Use a cadence that matches the strategy's opportunity rate and risk. The important distinction is between urgent reconciliation and slower inference from a sample—not a universal calendar.

Review a defined denominator

Include every opportunity that met the setup rules, including skipped entries, failed transactions, partial fills, valid losses, profitable rule breaks, and positions with unresolved residuals. Reviewing only large wins, painful losses, or posted calls selects for emotional salience.

Group by strategy version, chain, venue, liquidity band, time window, and other conditions declared before analysis. Keep live, simulated, reconstructed, and backtested observations separate. A small or shifting sample supports a question, not a confident optimization.

Metrics should answer a review question

Useful measures can include rule-adherence rate, intended versus actual size, quote-to-fill difference, price impact, total costs, time in trade, maximum adverse and favorable excursion, planned versus actual loss, probability calibration, and results by setup version. Define every numerator and denominator.

A dashboard is not a verdict. Higher win rate can accompany worse payoff; lower slippage can reflect smaller positions or easier liquidity; and average return can hide one dominant outlier. Read distributions and underlying records before assigning cause.

Convert observations into controlled changes

Write the observed pattern, affected cohort, alternative explanations, proposed rule, expected effect, evidence threshold, test period, owner, review date, and rollback condition. Version the strategy so later trades can be attributed to the rules that actually governed them.

Do not rewrite an entry or exit rule while a losing position is asking for an exception. If an urgent security or operational issue requires action, protect capital first and review the strategy change separately after exposure is under control.

Design the review environment

Review outside the live terminal when possible. Remove open P&L, social feeds, notifications, and the ability to enter a trade while classifying the session. Start with the pre-trade plan, then reveal execution and outcome evidence in order.

A second reviewer can challenge classifications, but share only the minimum data required and protect wallet addresses, personal information, API credentials, and private notes. Disagreement should be saved as evidence, not averaged away.

A review workflow

1. Reconcile exposure. Confirm balances, permissions, orders, signatures, costs, and residual positions.
2. Freeze the standard. Retrieve the plan and strategy version saved before the outcome.
3. Score the layers. Classify decision, execution, adherence, data quality, behavior, and outcome independently.
4. Join the cohort. Add the complete record to a defined sample without excluding inconvenient outcomes.
5. Test the inference. Check denominator, uncertainty, alternate explanations, regime, and whether the pattern repeats.
6. Close the action. Assign an owner, test, review date, and rollback rule to any approved change.

What belongs in the journal

Preserve the original thesis and plan, strategy version, source timestamps, actual balances and fills, full costs, residual exposure, decision and execution classifications, rule adherence, missing data, behavioral state, outcome, comparable-cohort identifiers, metrics and definitions, reviewer disagreement, proposed change, evidence threshold, owner, due date, test result, and rollback.

Primary and official sources