Blog
Operator notes for prediction-market traders. Calibration, concentration, and position sizing, anchored to the 8,656-wallet Polymarket cohort behind every Convexly finding. For the underlying research, see the research index.
Convexly Edge Score: not just another PnL dashboard
A plain-English note on what Convexly Edge Score measures, what it does not claim, and why method receipts and public failures matter.
Wallet Skill Brief #1: Most "whales" are not skilled
Across the Top-50 Polymarket cohort, the rank-correlation between Edge Score and realized PnL is Spearman ρ = −0.32 (p = 0.026, n = 50). One Polymarket wallet has $78K realized PnL and ranks 45/50 on Edge Score. The dollar leaderboard and the skill leaderboard disagree systematically.
What the discipline inversion means if you actually trade
Operator note on the updated Convexly V1-M finding: the recovered politics-excluded rerun reports -9.2pp lower concentration (95% CI [-12.8, -3.6], p = 0.0021). Causal language and the calibration secondary claim are superseded.
Stop Optimizing Calibration. Start Optimizing Conviction. What 10,000 Polymarket Wallets Tell You To Do.
Five operational rules for Polymarket traders, translated directly from the 8,656-wallet cohort. Calibration is table stakes. Concentration and survival math drive profit.
Unusual Flow Meets Unusual Skill. Ten Polymarket Wallets, Scored.
Ten wallets currently leading the Polymarket monthly-profit leaderboard, scored against Unusual Whales' Unusual Predictions criteria and Convexly's Edge Score. Four of ten match three or more criteria; the monthly rank-1 matches zero and made $7M anyway. Leaderboards select tail draws, not skill.
Base Rate Neglect on Prediction Markets
Traders price specific narratives and ignore historical frequencies. Incumbents, favorites, repeat performers. The base rate is the anchor. The wallet analyzer shows how often that anchor is ignored.
Overconfidence and Position Sizing on Polymarket
The 10K-wallet study shows calibration barely predicts profit, but overconfidence still kills. Under Hill α = 1.28, oversized bets blow up even when the probability looks right.
Kelly Criterion for Prediction Markets: Why Quarter-Kelly or Less
The Kelly derivation from Bell Labs to Thorp, and why Polymarket's fat-tailed returns demand quarter-Kelly sizing inside a Taleb barbell, not the half-Kelly defaults from finite-variance markets.
Prediction-Market Trade Journaling: What to Record on Every Position
A six-field template for every Polymarket position: thesis, base rate, probability estimate, stake, Kelly check, pre-mortem. The same fields the wallet analyzer captures automatically.
The Pre-Mortem: Running One on a Polymarket Thesis Before You Size Up
Gary Klein's pre-mortem, applied to a prediction-market position. Worked example on a $5K ETH contract. Why articulating failure modes typically cuts stake by a third.
The Brier Score: Necessary but Not Sufficient on Prediction Markets
The formula, the benchmarks, and the uncomfortable finding: across 8,656 Polymarket wallets, Brier explains only ~2% of profit-rank variance. Calibration tells you you won't blow up. It does not tell you you'll win.