LineWhat Kelly actually solves
Every staking question reduces to a tension: bet too little and your edge compounds slowly; bet too much and one bad stretch destroys the capital your edge needed to work on. The Kelly criterion, published by John Kelly at Bell Labs in 1956, is the mathematical answer to that tension — the stake size that maximizes the long-run growth rate of a bankroll, given two inputs: your win probability and the odds you're being paid.
The formula for a simple win/lose bet: f = (bp − q) / b, where b is the decimal profit per dollar staked (at +120, b = 1.2), p is your win probability, and q is 1 − p. The output f is the fraction of bankroll to stake. If you think a +120 dog wins 48% of the time: f = (1.2 × 0.48 − 0.52) / 1.2 = 4.7% of bankroll. If f comes out negative, Kelly is telling you the bet has no edge — the correct stake is zero.
Why full Kelly is a trap in sports betting
Kelly is mathematically optimal if the inputs are true. That's the catch: in blackjack, the win probability is known; in sports betting, your "55%" is an estimate, and bettors systematically overestimate their edge. Overbetting Kelly is asymmetric — staking double the true Kelly fraction produces zero long-term growth with enormous swings, while staking half of it keeps 75% of the growth with a fraction of the volatility. Since your estimate error only ever hurts you in one direction, the professional standard is half-Kelly or quarter-Kelly: run the formula, then divide by 2 or 4. The calculator above shows both.
Even fractional Kelly produces stakes that feel aggressive. A 55% edge at -110 gives full Kelly of 5.5% — half-Kelly is still 2.75% per play, more than double the flat 1% unit sizing most disciplined bettors use. That's not a contradiction; it's a confession. Flat 1% units are what quarter-Kelly looks like for someone honest about not knowing their true win rate yet. Kelly graduates from theory to practice only after a few hundred graded plays tell you what your percentage actually is — which is one more argument for keeping a real record.
Where Kelly earns its keep anyway
Even if you never stake exact Kelly fractions, the formula is worth internalizing for what it forces you to do: put a number on your edge. The moment you have to write down "I think this wins 56%," you can be wrong in a checkable way — and comparing your written probabilities against results over time is the fastest self-scouting tool in betting. It also gives you a principled reason to pass: any bet where your honest probability produces a negative f isn't a small bet, it's no bet. Traders will recognize this as position sizing with a different accent — same discipline, same failure modes, same cure.