Hindsight Bias in Journals
Hindsight bias is the tendency to believe, after the fact, that the outcome was predictable all along. In a trading journal it shows up as reviews that read 'obviously the level would hold' — statements nobody can verify because the outcome was already known when they were written.
The bias is corrosive for education specifically: it converts every review into a confirmation exercise. Entries stop producing new information because the reviewer already 'knows' the answer.
Time-Stamped Reasoning
The counterweight is reasoning written with timestamps before the result is visible. The pre-entry note, the mid-trade observation, and the post-session review should carry separate timestamps so each can be read as a product of its own moment.
When reviewing, read the timestamps in order and resist summarizing. The goal is to hear the earlier reasoning in its own voice, including its uncertainty — especially its uncertainty.
Separating Decision Quality From Outcome
Evaluate each entry twice, with two different questions. Question one: 'Given only what was knowable at decision time, was the reasoning sound?' Question two: 'What did the simulator display?' Keeping the questions in that order prevents the outcome from leaking into the quality assessment.
A decision can be sound and display a poor simulated result, or sloppy and display a good one. The journal's value is in separating those cells — a two-by-two table of quality versus outcome.
A Blind Review Exercise
A practical drill: cover the outcome column of the last ten entries and re-read only the pre-entry notes. Write, for each, whether you would take the same decision again. Then uncover the outcomes and note where your re-read judgment disagrees with your memory of the session.
The disagreements are the lesson. They show where outcome-knowledge was silently doing the evaluating — the exact mechanism hindsight bias uses.
Anti-Hindsight Techniques
| Technique | How It Works | What It Prevents |
|---|---|---|
| Timed entry review | Review written before replaying the outcome | Rewriting intent after seeing the result |
| Counterfactual log | Record what you'd have done at each decision point blind | 'I would have known' claims |
| Information blackout | Hide later-session data during review | Using future data to grade past calls |
| Two-reader pass | A peer reads only pre-written fields | Narrative smoothing by the author |
| Version diff | Compare plan vs. review text mechanically | Silent edits of the original intent |
Journal Checklist
- Was the review written before the outcome was replayed?
- Is the information available at each decision reconstructed, not assumed?
- Are 'I would have...' claims supported by pre-written fields?
- Is there any post-hoc editing of the original entry? (There must not be.)
- Does the review separate 'was knowable' from 'turned out'?
Review Note
Review decisions at the time they were made, not at the time the outcome became visible. The journal is a time machine only if its timestamps are honest.
Frequently Asked Questions
Only if you let it — procedural fixes exist: review before replay, blackout later data, diff versions. Hindsight is a procedure failure, not a human constant.
At each decision point, write what you'd do with only the information visible then — before unblinding. It turns 'I would have known' into a checkable claim.
Because the author can't see their own smoothing. A peer with only the pre-written fields can mark interpretation drift the author is blind to.
Continue Exploring
Costs and Assumptions in Simulation
Commissions, delays, and model limitations: how to document what the simulator charges, assumes, and quietly omits.
Separating Practice from Performance Claims
Why simulated results can never be presented as evidence of future returns — and what an educational journal may legitimately claim.
Paper Fills Are Not Live Fills
Why a simulated execution differs from a real market fill, and how to record the gap honestly in an educational trade journal.