Kelly Criterion

To apply Kelly on a platform like Kalshi or Polymarket, first estimate the true probability p of the event independently of the market price. The market price p_m represents the Implied Probability. The Kelly fraction f* = (p minus p_m) / (1 minus p_m) tells you what percentage of your bankroll to commit. For example, if you estimate p = 0.70 and the market shows p_m = 0.55, then f* = (0.70 minus 0.55) / (1 minus 0.55) = 0.15 / 0.45 = 33%. You would stake 33% of your bankroll. After accounting for Trading Fees, the effective edge shrinks, lowering the recommended fraction. Most practitioners use half-Kelly (16.5% in this example) to cushion against estimation error.

The Kelly Criterion is theoretically optimal but requires perfect probability estimates and unlimited time to converge. In practice, prediction market traders face estimation error, correlated positions, and non-stationary market conditions. Overbetting above the full Kelly fraction increases variance faster than it increases expected return, creating a real risk of ruin even with positive Edge. This is why fractional Kelly strategies are standard in professional trading contexts. The criterion also presupposes that each trade is independent, which is violated when multiple open positions share the same underlying event or political theme. Combining Kelly sizing with holistic Bankroll Management rules, such as maximum per-trade exposure and drawdown limits, produces a more robust risk framework than Kelly alone.