Kelly Criterion: The Math Behind Optimal Bet Sizing
Flat betting is leaving money on the table. Here's the formula that tells you exactly how much to risk on every play.
Most bettors do one of two things: they either bet the same flat amount on every play regardless of edge, or they follow their gut and size up when they "feel good." Both approaches are wrong, and the math can prove it.
The Kelly Criterion is a formula developed by mathematician John Kelly in 1956. It tells you exactly what percentage of your bankroll to wager on any given bet to maximize long-run growth while avoiding blowing up your account. It's used by professional gamblers, hedge funds, and anyone who takes the math seriously.
The Formula
f* = (bp - q) / b
Where: f* is the fraction of your bankroll to bet. b is your net odds (what you win per dollar risked). p is your estimated true probability of winning. q is 1 - p (probability of losing).
For prediction markets trading in cents, it simplifies to this: **f* = (true probability - implied probability) / (1 - implied probability)**
That's it. Your edge divided by what you stand to lose if wrong.
A Real Example
Say a Kalshi market is pricing YES at 42¢ — the market implies a 42% probability. You've done your research and think the true probability is closer to 58%. That's a 16-point edge.
Kelly says: f* = (0.58 - 0.42) / (1 - 0.42) = 0.16 / 0.58 = 27.6% of your bankroll
That's a massive bet. Which is why most pros use a fraction of Kelly.
Full, Half, and Quarter Kelly
Full Kelly maximizes long-run bankroll growth in theory. In practice, it produces gut-wrenching variance. You'll be right on the math and still watch your bankroll cut in half before it recovers.
Half Kelly is the professional standard. Same expected long-run growth rate, roughly half the variance. If you're unsure, default here.
Quarter Kelly is the conservative play — useful when your edge estimate is shaky or the market is less liquid. It sacrifices some growth for significantly lower drawdowns.
FadeMe's picks page shows Kelly sizing at all three fractions automatically once you enter your bankroll.
When to Ignore Kelly
Kelly assumes your probability estimate is accurate. The bigger your uncertainty, the more you should shade down. If you're guessing at the true probability rather than deriving it from data, use Quarter Kelly at most.
Also: Kelly says nothing about correlation. If five of your picks all resolve on the same news event, don't bet full Kelly on all five simultaneously.
The Hard Part Isn't the Math
The formula is simple. A calculator can do it in two seconds. The hard part is actually following it when your gut says to size up on a "sure thing" or size way down on a pick that mathematically deserves a real bet.
Discipline is the edge. The math just tells you where it goes.
Want to put this analysis into practice?
Analyze a Bet Now