Prediction-Market Thinking Errors
The most damaging prediction-market mistakes collapse probability, direction, price, realized PnL, and locked arbitrage into one misleading intuition.
Most beginner errors come from using one number to answer several different questions. Keep a small ledger of symbols: is estimated probability, is execution price, is settlement, is realized PnL, and is estimated gross EV.
Six recurring mistakes #
| Mistake | What it confuses | Correction |
|---|---|---|
| “The main thing is choosing the right side” | direction with value | Buy only when estimated payoff exceeds all-in price |
| “65% means this trade will win” | probability with certainty | It explicitly allows 35% losses |
| “Most trades won, so the strategy is profitable” | accuracy with money | Sum price-, size-, and cost-weighted PnL |
| “Price 0.70 means profit 0.30 with probability 70%” | payoff math with probability estimate | Winning PnL is 0.30; win probability requires separate |
| “Market price is true probability” | observable price with unknown reality | Treat price as an estimator with assumptions and frictions |
| “Positive spread guarantees profit” | estimated edge with locked payoff | Account for fill selection, costs, staleness, and unhedged resolution |
One diagnostic example #
Let and . YES is the likely side. If YES occurs, profit is +0.30; if NO, loss is -0.70. Expected value is
This one example disproves three intuitions: likely is not certain, likely is not necessarily cheap, and a 0.30 winning branch does not imply positive EV.
Life example #
A restaurant can be likely to be busy tonight yet be a poor investment at an absurd purchase price. Forecasting demand and valuing ownership are separate stages.
A five-question pre-trade check #
- What exact payoff and settlement text am I buying?
- What are , , costs, and the uncertainty around ?
- What are win PnL, loss PnL, net EV, and capital at risk?
- Is this position hedged, or only valued from a reference?
- Which other positions share its event, signal, and execution failure modes?
Application to no-hedge #
No-hedge concentrates all six mistakes if the reference is treated as certainty and the discount as locked profit. Preserve the complete chain: validated complement → fresh probabilistic estimate → executable maker price → net edge → unhedged binary exposure → clustered position limit → outcome-based calibration.
Self-check: q=0.40, p=0.20. Which side is more likely, and is buying YES attractive before costs?
NO is more likely, but YES has gross edge . Direction and value give different answers.