Prediction Market Now • Live election odds • Crypto event markets • Sports prediction lines • Policy and macro event pricing • Real-time market sentiment •
← Back to blog

Blog Article

The Kelly Criterion for Prediction Markets: A Trader’s Math

Most traders on Polymarket and Kalshi blow through their bankrolls because they size positions with gut feel. The Kelly Criterion offers a mathematical answer to the oldest question in trading: how much should you risk? In prediction markets, where every contract is a probability bet, Kelly sizing separates long-term winners from quick burnouts. This guide breaks down the math, shows you how to adapt it for binary contracts, and explains why the pros rarely use full Kelly.

Why Kelly matters more in binary markets

Prediction markets trade binary contracts that settle at $1.00 or $0.00. A contract priced at 62 cents implies a 62% probability. If your model says the true chance is 70%, you have an edge. Kelly tells you exactly how much of your bankroll to risk on that edge. Traditional sports betting uses Kelly too, but prediction market mechanics make it simpler: no juice, no spreads, just pure probability versus price.

The wisdom of crowds usually prices contracts efficiently, but information asymmetries and news shocks create windows. When you spot mispricing, Kelly keeps you in the game long enough to capture your edge repeatedly. Overbetting kills you on the inevitable losing streaks. Underbetting leaves profit on the table. Kelly finds the sweet spot.

The Kelly formula, adapted

The classic Kelly formula is f = (bp – q) / b, where f is the fraction of your bankroll to bet, b is the odds received, p is your win probability, and q is 1 minus p. For binary markets, the formula simplifies. If you buy a Yes contract at 0.62 and believe the true probability is 0.70, your edge is 0.08. Kelly says bet (0.70 × 1.00 – 0.30 × 0.62) / (1.00 – 0.62) = 0.158, or about 16% of your bankroll.

The math assumes you can repeat the bet infinitely and that your probability estimate is accurate. In practice, neither holds perfectly, which is why traders modify the approach.

Calculating edge from your model

Your edge is the difference between your estimated probability and the market price. If Kalshi prices an event at 45 cents and your model says 55%, your edge is 10 percentage points. Plug that into Kelly along with the contract payout structure. Most platforms pay $1.00 on a win, so the math stays clean. The hard part is building a model that beats the collective intelligence of the crowd consistently.

Fractional Kelly and why pros use it

Full Kelly maximizes long-term growth rate but produces wild swings. A string of bad luck can cut your bankroll in half even when you have genuine edge. Most professional traders use half Kelly or quarter Kelly. This sacrifices some growth for dramatically lower volatility. Half Kelly means you bet 8% instead of 16% in the earlier example. Your bankroll grows slower, but you sleep better and survive the inevitable model errors.