Kelly Criterion Explained: Optimal Bet-Sizing Formula
Expert Analysis

Kelly Criterion Explained: Optimal Bet-Sizing Formula

The Board·Jul 10, 2026· 7 min read· 1,716 words

The Kelly Criterion is a formula that tells you how much of your bankroll to wager on a favorable bet. It answers the question every gambler, trader, and prediction-market participant eventually runs into: I have an edge — so how much do I actually bet? Bet too little and you leave growth on the table. Bet too much and you go broke over time, even when every single bet is in your favor.

Developed by physicist John L. Kelly Jr. at Bell Labs in 1956, the criterion has since been adopted by legendary investors — Ed Thorp used it to beat blackjack and then run a hedge fund that compounded for decades; Warren Buffett and Charlie Munger have described their concentrated bet-sizing in Kelly-like terms. This guide walks through what it is, the formula, a worked example, fractional Kelly, and how to apply it in prediction markets, sports betting, and trading — plus the mistakes that quietly destroy bankrolls.

What the Kelly Criterion actually tells you

The Kelly Criterion finds the bet size that maximizes the expected logarithm of your wealth — which, over many repeated bets, maximizes your long-run compound growth rate. That last part is the key. Kelly does not maximize your expected profit on the next bet (that would tell you to bet everything). It maximizes how fast your bankroll grows when you make the same kind of bet again and again.

The intuition: growth is multiplicative, not additive. Halve your bankroll and then add back half of what remains, and you are not back to even — you end up down a quarter. Because losses compound against you, protecting the downside matters as much as capturing the upside. Kelly is the mathematically optimal balance between the two.

The Kelly formula

For a simple bet with two outcomes, the Kelly fraction is:

f* = (bp − q) / b

Where:

  • f* = the fraction of your bankroll to bet
  • b = the net odds received (bet $1 to win $b; 2-to-1 odds means b = 2)
  • p = your probability of winning
  • q = your probability of losing = 1 − p

The even-money shortcut

For an even-money bet (b = 1), the formula collapses to something you can do in your head:

f* = p − q = 2p − 1

That is simply your edge. A win rate of 0.55 on an even-money bet gives 2(0.55) − 1 = 0.10, so Kelly says stake one-tenth of your bankroll.

A worked example

Suppose you are offered 2-to-1 odds (b = 2) on a coin you believe lands heads half the time (p = 0.5, q = 0.5).

f* = (2 × 0.5 − 0.5) / 2 = (1 − 0.5) / 2 = 0.5 / 2 = 0.25

Kelly says bet a quarter of your bankroll. On a $10,000 bankroll, that is a $2,500 wager. Notice the formula rewards both a bigger edge and better odds: improve either your win probability or the payout, and Kelly tells you to bet more.

Now flip it. If those same 2-to-1 odds applied to a coin that wins only three times in ten (p = 0.3):

f* = (2 × 0.3 − 0.7) / 2 = (0.6 − 0.7) / 2 = −0.05

A negative Kelly fraction means the bet is unprofitable — do not take it (or bet the other side if you can). This is the criterion's built-in discipline: it refuses to size a bet with no edge.

Why betting more than Kelly is a trap

The single most important lesson in Kelly is what happens when you overbet. Plot long-run growth rate against bet size and you get a curve shaped like a hill. Growth rises as you bet more — up to the Kelly fraction, the peak — and then it falls. Bet past Kelly and your growth rate declines. Bet at roughly twice the Kelly fraction and your expected growth rate drops to zero. Bet more than that and, despite having a genuine edge, you are mathematically expected to go broke.

This is why "bet big when you're confident" is such dangerous advice without a framework. Confidence is not the same as edge, and even a real edge has an optimal size beyond which more is strictly worse.

Fractional Kelly: why professionals bet half

In practice, almost no serious bettor bets full Kelly. They bet fractional Kelly — most commonly half-Kelly. Here is why that is the smart move, not a timid one:

  • Your probabilities are estimates, not facts. The formula assumes you know your true win probability. You don't. Overstate your edge and full Kelly overbets — and the growth curve punishes overbetting far more harshly than underbetting.
  • Half-Kelly keeps most of the long-run growth while cutting the volatility roughly in half. Because the growth curve is flat near its peak, halving your bet sacrifices only a small slice of return while dramatically smoothing the ride and shrinking drawdowns.
  • Drawdowns are survivable. Full Kelly can produce gut-wrenching drawdowns — half your bankroll or more — that end careers before the math has time to work. Fractional Kelly keeps you in the game.

The rule of thumb among experienced practitioners: never bet more than full Kelly, and usually bet half.

Kelly in prediction markets

Prediction markets are a natural home for Kelly because prices are probabilities. A contract trading at $0.60 implies a 60-in-100 market-assessed chance of the event. You only have an advantage when your own probability estimate differs from the market's price. As of 2026, the rapid growth of prediction markets has made Kelly-style sizing newly relevant to everyday traders, not just professionals.

The workflow:

  1. Estimate your own probability p for the outcome.
  2. Read the market price, which sets your odds. Buying "Yes" at $0.60 to settle at $1.00 gives net odds b = (1 − 0.60) / 0.60 ≈ 0.67.
  3. Only bet when p is meaningfully higher than the implied price. Plug p and b into the Kelly formula.
  4. Apply a fractional multiplier — your probability estimate in a messy real-world market is far less reliable than in a coin flip, so half-Kelly or less is prudent.

The hard part is not the formula; it is honestly estimating p. Overconfidence in your own forecast is the fastest way to overbet. For a deeper treatment of finding an advantage in prediction markets — arbitrage, calibration, and exploiting behavioral bias — see our guide to prediction market trading strategies.

Kelly for sports betting

Sports bettors use Kelly to convert an edge over the sportsbook's line into a stake. The odds b come straight from the payout; the probability p comes from your model. Because sportsbook lines are sharp and public models are noisy, disciplined bettors almost always use quarter- or half-Kelly, and they account for the vig (the book's built-in margin) when computing whether an edge exists at all. A "value bet" that ignores the vig usually isn't one.

Kelly for trading and investing

In continuous markets, Kelly generalizes: the optimal fraction is approximately expected excess return divided by variance. Ed Thorp applied exactly this logic to move from card counting to managing money. The same cautions apply, only stronger — returns are non-stationary, correlations shift, and estimation error is large — so investors typically run well below full Kelly. Position sizing at "half-Kelly on a conservative return estimate" is a common professional posture.

Common mistakes that blow up bankrolls

  • Overstating your edge. Garbage-in probabilities produce oversized bets. When unsure, shade p toward the market.
  • Using full Kelly with uncertain probabilities. Full Kelly is optimal only if your inputs are exact. They never are. Bet fractional.
  • Ignoring correlation. Kelly assumes one bet at a time. Placing several correlated bets simultaneously is really one big bet — size the portfolio, not each leg.
  • Treating Kelly as a profit guarantee. Kelly maximizes long-run growth given an edge. With no edge, no bet size saves you; with an edge, short-run losing streaks are still guaranteed.

Kelly Criterion calculator

To compute your own Kelly fraction, you need three inputs: your win probability (p), the net odds (b), and your fractional multiplier (start at one-half). Then:

  1. Compute full Kelly: f* = (b·p − (1−p)) / b
  2. If f* is zero or below, don't bet.
  3. Multiply by your fraction: stake = bankroll × f* × multiplier

Example: a $5,000 bankroll, p = 0.58, even-money odds (b = 1), half-Kelly. Full Kelly = 2(0.58) − 1 = 0.16. Half-Kelly stake = $5,000 × 0.16 × 0.5 = $400.


Frequently asked questions

What is the Kelly Criterion in simple terms? It is a formula for how much to bet when you have an edge, chosen to grow your money as fast as possible over the long run without risking ruin.

Is the Kelly Criterion better than betting a fixed amount? Over many bets with a real edge, Kelly grows a bankroll faster than any fixed-fraction strategy that bets more, and more safely than any that bets more aggressively. It is provably growth-optimal.

What is fractional Kelly? Betting a set fraction (often one-half) of the amount full Kelly recommends. It gives up a little long-run growth for a large reduction in volatility and drawdown, and it protects you from overbetting when your probabilities are wrong.

Can the Kelly Criterion lose money? Yes. Kelly optimizes growth given an edge; it does not eliminate losing streaks, and if your edge is overstated (or negative), Kelly can lose money like any strategy. It only shines when your probabilities are well-calibrated.

Who invented the Kelly Criterion? John L. Kelly Jr., a researcher at Bell Labs, in a 1956 paper. It was later popularized in gambling and investing by mathematician Ed Thorp.

The bottom line

The Kelly Criterion turns a vague instinct — "bet more when the edge is bigger" — into a precise, growth-optimal number. The formula is simple; the discipline is not. The edge you plug in is only an estimate, overbetting is punished far more than underbetting, and correlation quietly turns several small bets into one large one. Bet fractional, respect the downside, and let compounding do the work.