Kelly Criterion Gambling: Optimal Bet Sizing Explained

Kelly criterion gambling is the mathematically optimal strategy for sizing bets to maximize the long-term growth rate of a bankroll. John Kelly published it at Bell Labs in 1956, and it has been the quiet backbone of every serious advantage player and professional sports bettor since. The catch is that for most casino gambling, Kelly tells you something you probably do not want to hear.

What the Kelly Criterion Actually Says

The formula is deceptively small: f* = (bp - q) / b. Here, f* is the fraction of your bankroll to wager, b is the net odds received (decimal odds minus one), p is your probability of winning, and q is the probability of losing (1 - p). The output is a percentage of bankroll, not a dollar amount, which is why Kelly forces you to resize every bet as your balance moves.

A worked coin flip: Imagine a coin weighted so heads comes up 60% of the time, and you get even money on heads. Then b = 1, p = 0.60, q = 0.40, and Kelly = (1 × 0.60 − 0.40) / 1 = 0.20. The optimal bet is 20% of bankroll on each flip — a shockingly aggressive number, which is exactly the point. Kelly is not polite.

What Kelly optimizes: It maximizes the expected geometric growth rate of wealth over infinite repeated bets. In plain English, it compounds your money as fast as mathematically possible without ever sending the bankroll to zero. What it does not do is eliminate short-term volatility, guarantee any individual win, or produce a positive bet on a negative-expectation game.

Kelly Criterion Gambling: Why Kelly Hates Casino Games

Every casino game with a house edge has negative expected value for the player, so Kelly returns a negative number. A negative Kelly fraction is not "bet a little less" — it is the formula telling you to be the house, not the player. Since you cannot bet negative money, the correct interpretation is simply: do not bet.

Roulette on a single-zero wheel: Red pays even money, so b = 1. Your win probability is 18/37 = 0.486 and q = 19/37 = 0.514. Kelly = (1 × 0.486 − 0.514) / 1 = −0.028. The optimal bet is −2.8% of bankroll, which does not exist, so Kelly's answer is zero.

Blackjack on basic strategy: Against a ~0.5% house edge, p ≈ 0.495 and q ≈ 0.505. Kelly = (1 × 0.495 − 0.505) / 1 = −0.01. Still negative, still zero. Blackjack only flips positive if you are counting cards and the true count gives you a measurable edge — the narrow window where Kelly starts to matter again.

Where Kelly genuinely works: Card counting at a favourable count, sports betting with a verified information edge, poker (skill creates long-run edge against opponents), and structural bonus situations with positive EV. Outside of these, Kelly's answer is a clean, unambiguous $0. For more on when you actually have an edge, see our advantage play guide.

Kelly Numbers for Common Gambling Scenarios

Here is what Kelly actually says for the bets most gamblers consider. This is the table that separates people who use Kelly from people who understand it.

Game/Bet	Win Probability	Odds (b)	Kelly Fraction	Optimal Bet (of $1,000 bankroll)	Verdict
Roulette: Red/Black (single zero)	48.6%	1.0	-2.7%	$0	Don't bet
Roulette: Red/Black (double zero)	47.4%	1.0	-5.3%	$0	Don't bet
Blackjack: Basic strategy	49.5%	1.0	-1.0%	$0	Don't bet
Blackjack: Card counting (true count +2)	51.0%	1.0	+2.0%	$20	Bet — small
Blackjack: Card counting (true count +5)	52.5%	1.0	+5.0%	$50	Bet — moderate
Sports bet: 55% edge at -110	55.0%	0.91	+5.5%	$55	Bet — if edge is real
Sports bet: 52% edge at -110	52.0%	0.91	-0.7%	$0	Edge too small at these odds
Dice (1% house edge): Over/under	49.5%	1.0	-1.0%	$0	Don't bet
Coin flip: Fair coin, 2:1 payout	50.0%	2.0	+25.0%	$250	Bet aggressively (if real)
Poker tournament: Skilled player	~55-60%	Varies	Positive	2-5% of bankroll	Bet — bankroll management critical

Notice how narrow the positive-EV column is. Outside card counting, a verified sports model, or skilled poker, everything else is $0. That is not pessimism, that is just the arithmetic doing its job.

Fractional Kelly: Growth Versus Survival

If you do have an edge, the next question is how aggressively to press it. Full Kelly maximizes the theoretical growth rate, but short-term volatility is brutal — 50%+ drawdowns are not rare, they are expected. Most serious practitioners back off to fractional Kelly because they distrust their own edge estimates and their own nerves.

Why shrink the bet: Your edge estimate is imprecise, variance is vicious at full Kelly, and humans under drawdown pressure deviate from their strategy — and deviating is strictly worse than running a smaller, consistent fraction. Half-Kelly has become the professional default because it preserves most of the growth while massively reducing the ulcer rate.

Example with a sports bet: You believe Team A wins 55% of the time and the book offers +110 (b = 1.10, p = 0.55, q = 0.45). Kelly = (1.10 × 0.55 − 0.45) / 1.10 = 0.14, or 14% of bankroll. But if you are slightly wrong — 52% gives Kelly = 6.5%, and 50% collapses it to 0% — using half-Kelly (7%) absorbs most of that estimation error before it wrecks you.

Kelly Fraction	Bet Size (5% edge)	Expected Annual Growth	Probability of 50% Drawdown	Emotional Reality
Full Kelly (100%)	$50 per $1,000	Maximum (theoretical)	~50% at some point	Terrifying swings
3/4 Kelly (75%)	$37.50	~94% of max growth	~25%	Still rough
Half Kelly (50%)	$25	~75% of max growth	~6%	Manageable for most
Quarter Kelly (25%)	$12.50	~44% of max growth	~0.4%	Smooth ride

The professional consensus: half-Kelly is the standard for serious advantage players, and quarter-Kelly is where people go when the edge is soft or the bankroll cannot tolerate a bad month. Full Kelly is a whiteboard ideal, not a live strategy.

The Five Ways People Break Kelly

Using Kelly without an edge. The most common misuse is applying the formula to roulette, slots, or any flat negative-EV game and being surprised when it returns a negative number. Kelly is a sizing rule, not an alchemy spell — it cannot manufacture an edge that isn't there.

Overestimating your edge. Sports bettors convince themselves they are 55% when they are really 51%, and poker players feel invincible after a heater. Kelly is unforgiving here: the penalty for overbetting grows faster than the reward, which is exactly why fractional Kelly exists to protect you from yourself.

Ignoring variance. Even a correctly-sized full-Kelly bet on a real edge produces 50%+ drawdowns as a routine matter. If that ends your willingness to follow the system, a smaller fraction will leave you richer long-term simply because you will actually stick with it.

Applying Kelly to correlated bets. The formula assumes independence. Betting three markets on the same football match means those bets move together and the effective Kelly fraction is smaller than the formula suggests. Treating them as independent leads to stealth overbetting.

Not recalculating as the bankroll moves. Kelly is a fraction, not a fixed dollar amount. After a win your bets should grow; after a loss they should shrink. People who calculate once and then flat-bet yesterday's number are running a slowly broken facsimile of Kelly, not Kelly itself.

What Recreational Players Should Do Instead

Since Kelly returns $0 for almost every casino game, recreational gamblers need a different framework. You are not optimizing wealth, you are buying entertainment, and the right question is how to get the most entertainment per dollar without damaging yourself.

Entertainment budget: Decide in advance what you can afford to lose in a given month — say, $200 — and treat it as the price of a night out, not an investment. Divide it into sessions ($50 × 4 trips, for example), and when a session's envelope is empty, you are done. This is not Kelly optimization, it is harm reduction.

Session length and bet sizing: Smaller bets mean longer sessions, and for entertainment gambling, time-at-the-table is usually the thing you actually want. A $1 slot spin or a $5 blackjack hand stretches $50 into real playtime; a $25 minimum evaporates it in minutes. Pick the stakes that maximize enjoyable session length, not action.

Hard loss limits: Set maximum losses per session and per month in advance, and treat them as non-negotiable. A pre-committed limit is made by a calm version of you and enforced on a tilted version of you — rewrite it mid-session and it was never a limit at all.

Kelly for Advantage Players

For the narrow group with a genuine edge, Kelly becomes a working tool rather than a thought experiment. Card counters use it implicitly when they spread bets: minimum units at neutral counts, scaling up as the true count rises and the edge with it. Sports bettors with a verified model do the same thing in a different venue — model true probabilities, compare to market odds, calculate Kelly, then deliberately scale down to half- or quarter-Kelly to absorb estimation error.

Poker sits in a similar neighbourhood. Pure Kelly does not map cleanly to tournaments because buy-ins are discrete and variance is extreme, but the spirit of the rule — never risk more of your bankroll than your edge justifies — produces the standard guidance of 20-50 buy-ins for your chosen stakes.

The Bottom Line

Kelly is mathematically optimal for positive-expectation bets, genuinely useful for advantage players, and ruthlessly honest about negative-EV casino games: the optimal bet is zero. If you are playing for entertainment, use an entertainment budget; if you have a real edge, use fractional Kelly and respect the variance.

The Kelly Criterion's greatest insight is also its bluntest: the math says stop. If you continue anyway, at least be honest about why. For more on the underlying numbers, see our guides to provably fair gambling, RTP in casino games, and why Martingale fails.

Frequently Asked Questions

What is the Kelly Criterion and how does it work?

The Kelly Criterion is a mathematical formula developed by John L. Kelly Jr. at Bell Labs in 1956 for determining the optimal bet size to maximize long-term bankroll growth. The formula is: f* = (bp - q) / b, where f* is the fraction of bankroll to wager, b is the odds received (decimal odds minus 1), p is the probability of winning, and q is the probability of losing (1 - p). For example, if you're offered 2:1 odds on a coin flip that lands heads 60% of the time, Kelly recommends betting 40% of your bankroll: (2 × 0.6 - 0.4) / 2 = 0.40. The formula works by balancing growth rate against risk of ruin — betting more than Kelly increases volatility without improving long-term returns, while betting less than Kelly is mathematically safer but grows your bankroll slower. The critical requirement: you must have a positive expected value (edge) for Kelly to recommend any bet at all.

How do you calculate the Kelly Criterion for betting?

To calculate Kelly, you need two inputs: your true probability of winning and the odds being offered. Step 1: Determine your edge — if you believe a team has a 55% chance of winning and the sportsbook offers even money (2.00 decimal), your edge is positive. Step 2: Apply the formula: f* = (bp - q) / b. With decimal odds of 2.00, b = 1. So f* = (1 × 0.55 - 0.45) / 1 = 0.10 or 10% of your bankroll. Step 3: Adjust for reality — most professionals use "fractional Kelly" (typically quarter-Kelly or half-Kelly) because the formula assumes you know the true probability exactly, which you never do. Overestimating your edge by even a few percentage points leads to dramatically oversized bets. For casino games with a known house edge, Kelly always returns a negative number, meaning the mathematically optimal bet is zero. Kelly only produces a positive recommendation when you have a genuine, quantifiable advantage.

Should you use full Kelly or fractional Kelly?

Almost everyone should use fractional Kelly — typically quarter-Kelly (25% of the calculated amount) or half-Kelly (50%). Full Kelly maximizes theoretical long-term growth rate but comes with extreme volatility: the probability of losing 50% of your bankroll at some point approaches certainty even with a genuine edge. Half-Kelly reduces the growth rate by only 25% but cuts the volatility roughly in half, making it far more psychologically sustainable. Quarter-Kelly reduces growth rate more but provides even smoother results. Professional sports bettors and poker players almost universally use fractional Kelly because: you never know your true edge exactly (any overestimate leads to ruinous overbetting), drawdowns are emotionally devastating and lead to poor decisions, and the practical difference in long-term wealth between half-Kelly and full Kelly is less important than actually sticking with the system through inevitable losing streaks.

Does the Kelly Criterion work for casino games?

No — and this is the most important thing to understand about Kelly. The Kelly Criterion requires a positive expected value (edge over the house) to produce a positive bet recommendation. Standard casino games — slots, roulette, baccarat, craps, blackjack without card counting — all have a negative expected value for the player. When you input a negative edge into the Kelly formula, the result is negative, meaning the mathematically optimal bet is zero. Kelly doesn't "fail" for casino games; it correctly identifies that no bet size can make a losing proposition profitable long-term. The only casino-adjacent applications where Kelly works: card counting in blackjack (where specific counts create a temporary positive edge), poker (where skill creates long-run edge against other players), and sports betting (where superior information can identify mispriced lines). For recreational gambling at negative-expectation games, a fixed entertainment budget approach is more practical than Kelly.

What is the biggest risk of using the Kelly Criterion?

The biggest risk is overestimating your edge, which causes catastrophic overbetting. The Kelly formula assumes you know the exact probability of winning — in practice, you never do. If you believe your win probability is 55% but it's actually 50% (no edge), Kelly calculates a 10% bet on what is actually a coin flip, guaranteeing long-term ruin through systematic overbetting. This "garbage in, garbage out" problem is why fractional Kelly exists and why professional bettors are conservative with their edge estimates. The second major risk is psychological: even with a genuine edge and correct Kelly sizing, you will experience significant drawdowns — losing 30-40% of your bankroll is mathematically expected, and most people abandon their strategy during these periods. Kelly also assumes you can make unlimited sequential bets with consistent odds, which doesn't match reality. Used correctly with verified edges and fractional sizing, Kelly is the gold standard for bankroll management. Used carelessly with assumed edges, it accelerates losses faster than flat betting.