John L. Kelly Jr. published his criterion in 1956 while working at Bell Labs, and it remains the most mathematically rigorous answer to the question: "How much should I bet when I have edge?" The formula is deceptively simple. The implications are profound, and the failure to follow them is one of the most common causes of bankroll destruction among otherwise intelligent bettors.
The Kelly Formula
Kelly stake (as fraction of bankroll) = (bp − q) / b, where b is the net odds received (profit per unit staked), p is your estimated probability of winning, and q is the probability of losing (1 − p). On a prediction market where you buy YES at 55¢ and believe the true probability is 65%: b = (1−0.55)/0.55 = 0.818, p = 0.65, q = 0.35. Kelly = (0.818 × 0.65 − 0.35) / 0.818 = 0.143. Bet 14.3% of bankroll.
Full Kelly maximises long-run growth but creates enormous short-term volatility. Most professional bettors use half-Kelly (stake 50% of the Kelly amount) as a practical compromise. The maths of over-staking are merciless — exceeding full Kelly reduces long-run growth compared to the optimal.
Practical Kelly on Prediction Markets
- →You must be honest about your edge — overconfidence destroys Kelly's benefits
- →Uncorrelated positions allow Kelly to be applied across multiple simultaneous markets
- →Fractional Kelly (1/4 to 1/2) is preferred for most real-world applications
- →Kelly assumes exact probability knowledge — treat your estimates as ranges, not points
The Kelly Criterion applied to Boromarket positions gives you a mathematically defensible staking plan. It doesn't guarantee winning — it guarantees that if you have genuine edge, you're extracting it at the fastest sustainable rate.