Prediction-Market Trade Journaling: What to Record on Every Position
A trader who cannot reconstruct the reasoning for a past position cannot learn from it. The worst outcome on Polymarket is not a single loss; it is a series of losses that look identical in hindsight and were filed under different narratives at the time.
Why memory fails the trader
Kahneman's work on noise in judgment showed that the same person evaluating the same case on different days reaches materially different conclusions (Kahneman, Sibony, Sunstein, 2021, Noise, part III). Hindsight bias compounds the problem: once a contract resolves, the trader unconsciously rewrites the pre-trade reasoning to match the outcome. The position that lost is remembered as a stretch. The position that paid is remembered as obvious.
A journal pins the pre-trade state to a timestamp. When the contract resolves, the actual reasoning is on record, independent of the outcome. That is what makes calibration possible. Without it, a trader accumulates impressions, not data.
The six fields that matter
A Polymarket position journal needs six fields. Anything less is incomplete; anything more is journaling for its own sake.
Six fields. A disciplined trader fills them in under three minutes per position.
Common failure modes of trader journals
Skipping the probability estimate. The single most common mistake. A thesis without a probability is a narrative; it cannot be evaluated against what actually happened. This is also the field that makes Brier scoring (see the Brier post) possible after the fact.
Logging only the large positions. Small resolved positions are the best training data. A trader logging 100 sub-$100 positions per year learns calibration faster than one logging ten $5,000 positions. Volume and short feedback loops matter more than stake size for the journaling purpose specifically.
Never reviewing. The journal that is written and never read is a diary. A monthly review that scores each closed position on whether the trader was better or worse calibrated than the implied probability at entry is where the compounding begins.
Why spreadsheets cap out
A spreadsheet holds the six fields. What it does not do is compute the cohort-relative statistics that tell a trader whether they are actually any good. A single-trader spreadsheet cannot produce a posture percentile against 8,656 ranked wallets. It cannot compute the Hill tail estimator on the trader's own PnL. It cannot flag concentration percentiles that match the 10K-wallet cohort's top-quartile pattern.
The value of the journal is the inputs. The value of the analyzer is what the inputs become when benchmarked against a reference population. The Convexly wallet analyzer reconstructs the equivalent of the six-field journal directly from on-chain resolved positions, plus the cohort-relative ranking a spreadsheet cannot reach. For any trader on-chain, the journal mostly writes itself; the discipline that remains is the pre-trade probability estimate and the pre-mortem notes.
See how your wallet scores.
Paste any Polymarket address. Posture, conviction, and discipline percentiles against 8,656 benchmarked wallets. Free, no signup, 15 to 30 seconds.
Sources. Kahneman, Sibony, Sunstein (2021), Noise. Convexly (2026), Edge Score Methodology V1.