Outcome bias is the tendency to evaluate a prior decision differently because its outcome is known, even when the decision-maker could not have known that outcome. The original outcome-bias experiments tested how outcome knowledge changed evaluations. In betting, review the pre-event process before revealing the result.
Decision quality and outcome are different records
| Record | Available before event? | Review question |
|---|---|---|
| Target and settlement rule | Yes | Was the event defined correctly? |
| Evidence and cutoff | Yes | Were sources relevant and time-safe? |
| Probability | Yes | Was it produced by the declared method? |
| Accepted price and exposure | At acceptance | Did the actual contract meet the rule? |
| Result and return | No | What happened and how was it settled? |
The outcome-bias research supports protecting decision evaluation from information that was unavailable at the time. It does not imply outcomes should be ignored forever. Outcomes are essential for later forecast evaluation; they should simply enter at the correct stage.
Four illustrative cases
| Process | Outcome | Correct interpretation |
|---|---|---|
| Rule followed, evidence complete | Win | One favourable outcome; process can be scored separately |
| Rule followed, evidence complete | Loss | One unfavourable outcome; not automatic proof of a bad process |
| Rule broken, evidence missing | Win | Favourable outcome does not repair the process failure |
| Rule broken, evidence missing | Loss | Both outcome and process are unfavourable, but causes still need evidence |
Expected value is an average over possible outcomes under estimated probabilities. A positive estimate does not promise a win, and a win does not prove that the estimate was positive. The expected-value definition makes that distinction explicit.
Blinded review workflow
- Hide the outcome, return, and post-event commentary.
- Check target, evidence cutoff, method version, and missing-data rule.
- Recalculate the forecast from the frozen record.
- Verify accepted price, maximum loss, and settlement terms.
- Score each process item using a scale written in advance.
- Reveal and reconcile the result.
- Evaluate forecasts across a consistent later sample.
For binary probabilities, the Brier score can aggregate forecast error. Use all eligible forecasts and a declared baseline; do not select only placed bets or dramatic results.
Language that preserves the distinction
Prefer "the decision met the recorded process and the selection lost" to "it was a good bet that got unlucky" unless luck and value have been defined and measured. Prefer "the selection won despite a missing evidence field" to "the instinct was right."
Next step
Use Betting Journal for the next part of this topic.
Continue learning
- Next guide: Sunk-Cost Fallacy in Betting
- Related guide: Tilt in Betting
Assumptions and limitations
The four cases are conceptual. A blinded checklist can reduce one source of hindsight but cannot remove selective records, invalid models, or other biases. Outcome data remain necessary for validation. Process adherence does not itself establish positive expected value or future success.
