Win-Stay, Lose-Shift Audit: Did One Outcome Rewrite Your Trading Rule?

One Outcome, One Apparent Rule Change

Did One Result Alter Your Next Trading Decision?

A result followed by a different decision establishes a sequence worth reviewing, not proof of causation. Keep three records separate: the rule documented before the result, the result itself and the first genuinely comparable later decision. A visible difference can establish deviation; its cause and usefulness each require further evidence.

Start with four distinct propositions: a rule existed, the later action differed, the outcome caused the difference and the change was useful. Each proposition needs its own support. A dated rule may prove the baseline, while a field comparison may prove deviation. Neither proves motive or future value.

Four-item event strip

Do not fill a missing field from hindsight. Once the event is separated from its explanation, you can define what this review can responsibly establish.

Keep This Review Educational, Not Clinical

What Can This Audit Responsibly Tell You?

This audit can show that a documented later decision differed from a documented baseline. It cannot diagnose stress, anxiety, compulsion or impaired control; prove an internal mechanism; validate a trading strategy; or predict performance. Where the records stop, the responsible conclusion is unknown rather than a confident psychological explanation.

A PubMed abstract reports a laboratory fMRI study involving 175 participants. It describes neural group differences under experimentally induced acute stress, while reporting no measured effect on behavioural decisions or self-control performance in that task. The study was not about trading, and it does not prove pressure is irrelevant elsewhere.

Can-and-cannot boundary

The boundary protects the reader from overclaiming. With it fixed, the response-pattern labels can be explained without turning them into instructions.

What Win-Stay And Lose-Shift Mean

How Can Feedback Shape Later Response Patterns?

Win-stay and lose-shift describe whether a response is repeated or changed after feedback within a defined task. Their inverse patterns—win-shift and lose-stay—also matter. These labels can help organise what happened next, but none explains an individual’s motive or tells a trader which action to take.

The Nature paper analysed 218 participants across three controlled Dice Dual experiments against computerised opponents. Its repeated fixed-feedback phases included 36 intermediate trials. Results varied by experimental condition, and the task’s useful response depended on the opponent pattern. These artificial games did not study trading or one isolated outcome.

Description-not-prescription map

Treat every label as a description to test, never as proof of discipline, flexibility or error. Because a label cannot supply a missing history, the next step is to retrieve the actual pre-outcome rule.

Retrieve The Original Rule First

What Was Your Rule Before The Outcome?

A before-and-after claim needs a dated pre-outcome baseline. Retrieve the rule wording, its source or version, the conditions in which it applied and any review trigger approved beforehand. If the evidence is missing, mark it UNKNOWN, NOT RECORDED or UNAVAILABLE instead of reconstructing a cleaner rule from memory.

Memory written after the result is weaker evidence than a contemporaneous journal, checklist or version history. Record both, but label their provenance differently. If no original wording survives, you may still improve future documentation; you cannot confidently claim that a later rule rewrote a baseline that cannot be demonstrated.

Baseline and provenance card

1. Copy the exact rule text and its record date.
2. Name the journal, checklist or version containing it.
3. Record the setup conditions and allowed discretion.
4. Record its invalidation or risk boundary.
5. Record any pre-approved review trigger.
6. Mark every gap UNKNOWN, NOT RECORDED or UNAVAILABLE.

Once the baseline is frozen, select a later decision by criteria declared in advance rather than by how neatly it supports your story.

Compare The First Later Decision

Where Did The Next Decision Depart From Baseline?

Choose the earliest later decision that meets your pre-declared comparison criteria, then inspect it field by field against the baseline. Mark each field same, different or unknown and retain its evidence location. A difference can establish a rule deviation, but it cannot establish that the preceding result caused the deviation.

Define “comparable” before inspecting the later outcome. Use relevant process fields such as setup family, decision stage and applicable rule version, while accepting that no two market moments are identical. Temporarily conceal the win or loss label: if your process judgement changes after revealing it, outcome knowledge may be shaping the explanation.

Before-and-after matrix

Reflection: Hide the result label, record your conclusion, then reveal it and note whether the explanation changed. A documented difference becomes useful only after its preferred cause survives a fair challenge.

Test, Then Retain, Repair Or Reject

Does A Counterexample Weaken Your Preferred Explanation?

Seek a comparable case that conflicts with your preferred explanation and record a plausible alternative cause. Weigh evidence for, evidence against and unknowns before choosing retain, repair or reject. Those labels describe a conditional documentation outcome, not a direction for a live position or a promise of improved performance.

Apply the same evidence standard after favourable and unfavourable results. The Nature study lacked a true continuation baseline and within-phase time-course analysis, while its 50% condition had a special meaning in a binary game. Those caveats reinforce caution; they do not establish a trading rule or individual cause.

Conditional disposition card

A counterexample can weaken a universal explanation without proving the opposite. “Deviation documented; cause unknown” may therefore be the most accurate conclusion. Close the review by making the uncertainty visible and setting the next evidence condition.

FAQs