There is a number that represents the mathematically optimal fraction of your bankroll to wager on any given bet. Most players have never calculated it. A smaller group has heard of it but dismissed it as theory. Almost nobody at a casino table or sportsbook is actually using it—and that gap between what people bet and what they should bet explains a lot of consistent, preventable losses.
The concept is the Kelly Criterion, developed by physicist John Kelly in 1956 and adopted widely in professional gambling and investing. The core formula is simple: divide your edge by the odds you're receiving. If you have a 2% edge on a bet paying even money, the Kelly formula tells you to bet 2% of your current bankroll. Not 5%, not 10%, not whatever you brought to feel comfortable—2%.
Why Most Players Overbet
The intuitive bet size and the correct bet size almost never match. Players anchor to round numbers, to what they've always wagered, or to what feels proportionate to the excitement of the moment. None of those anchors have anything to do with edge or bankroll survival.
When you overbeta—say, wagering 10% of your bankroll on a bet where Kelly says 2%—you're not just risking more money in an obvious sense. You're altering the trajectory of your results in a way that makes ruin vastly more likely even when you're winning. This is counterintuitive enough that it's worth stating precisely: overbetting can turn a positive-edge situation into a losing one over time, purely through the mechanics of compounding losses at the wrong scale.
This happens because of the asymmetry between gains and losses. Lose 20% of your bankroll, and you need to win 25% back just to break even. Lose 50%, and you need 100% just to recover. The larger each individual bet is relative to your bankroll, the more any losing streak compounds into a hole that your winning rate cannot mathematically dig you out of fast enough.
Fractional Kelly and the Real World
Full Kelly is theoretically optimal but practically brutal. It maximizes the long-run growth rate of your bankroll, but the variance it generates along the way is severe enough that most people abandon the approach—or their bankroll—before the math has time to work. The swings are real and they're large.
Professional gamblers and sharp sports bettors almost universally use a fraction of full Kelly—commonly half Kelly or quarter Kelly. Betting half the Kelly amount roughly halves the variance while reducing long-term growth by only about 25%. That tradeoff is worth it for most people because it produces a ride you can actually stay on.
Half Kelly on a 2% edge at even money means betting 1% of your bankroll. On a $2,000 session bankroll, that's $20 per bet. Not $100. Not $200. Twenty dollars. For most recreational players, that number feels uncomfortably small. That discomfort is the problem.
When You Don't Know Your Edge
In casino games where you're playing against a house edge, Kelly still applies—but the fraction goes negative, which tells you something important: you should not be betting at all if you're trying to grow a bankroll. You're managing a controlled drawdown, not optimizing growth.
In that context, Kelly's value isn't about sizing up—it's about sizing down. The math gives you a ceiling on how much of your bankroll to expose per decision. Exceeding that ceiling doesn't make sessions more exciting in any durable sense. It accelerates the rate at which the house edge converts your bankroll into their revenue.
For players in games with genuine skill edges—poker, sports betting, advantage play situations—the edge is real but often uncertain. You might estimate you have a 3% edge on a particular sports bet, but that estimate carries its own error range. Betting full Kelly on an uncertain edge is doubly dangerous because you're calibrating to a number you don't actually know with precision. This is another argument for fractional Kelly: it provides a buffer against your own miscalculation of edge, which is almost always optimistic.
Bet Sizing Across Sessions
Kelly is technically a dynamic formula. As your bankroll grows, your bet size grows. As it shrinks, your bet size shrinks. This recalibration is what produces the compounding growth properties the math promises. Players who treat it as a fixed-dollar amount instead of a fixed percentage are using it incorrectly and will see worse results.
In practice, recalibrating every single bet is impractical at a table. The reasonable approximation is to recalibrate at the start of each session based on your current total bankroll—not your session buy-in. What you brought to the table today is not the right reference point. Your total gambling bankroll is.
This distinction matters more than it might appear. A player who loses three sessions in a row and keeps buying in for the same amount is effectively increasing their Kelly fraction relative to a shrinking bankroll. They're getting more aggressive precisely when the math says to get more conservative.
The Practical Takeaway
You don't need to be a mathematician to apply this. The steps are concrete:
First, establish a total gambling bankroll—a real number that represents the money you've allocated to this activity across all sessions, not just tonight.
Second, estimate your edge honestly. In negative-edge games, accept that the math recommends the smallest bets compatible with the table minimum and the entertainment value you're buying. In skill games, use conservative edge estimates.
Third, calculate half Kelly as a percentage of your total bankroll. That's your bet size ceiling for the session.
Fourth, do not adjust upward mid-session because you're running hot. The Kelly formula doesn't change because you've won a few hands. Your bankroll has changed slightly; recalculate if you must, but don't use a winning streak as a reason to abandon the framework.
The reason this works when intuitive sizing doesn't is that Kelly is derived from the actual math of compounding outcomes over time—not from any theory about what feels right or what should happen in the short run. It's one of the few places in gambling where the math is both accessible and directly actionable. Most players leave it on the table.