Most bettors obsess over picking winners and barely think about how much to stake on each one. That's backwards. How you size your bets often decides whether you grow a bankroll or blow it long before your picks ever get the chance to prove themselves.
The Kelly Criterion is a staking formula built to answer exactly that question: given an edge, how much should you bet to grow your bankroll fastest without going broke? It's mathematically elegant and genuinely powerful – and also one of the easiest tools to misuse in a way that wrecks you faster than no system at all. Let's break down what it is, how it works with a real example, and whether you should actually use it.
The Kelly Criterion is a formula that tells you what percentage of your bankroll to stake on a bet, based on two things: how big your edge is, and the odds you're getting. The core idea is that your bet size should scale with your advantage – bet more when your edge is large, less when it's small, and nothing at all when you have no edge.
The standard formula for betting looks like this:
f = (bp – q) / b
Where:
f is the fraction of your bankroll to bet
b is the decimal odds minus 1 (your profit per unit staked if you win)
p is your estimated probability of winning
q is the probability of losing (which is just 1 – p)
Don't let the letters intimidate you – it's simpler in practice than it looks, and most people use a calculator for it anyway. The key thing to understand is what the formula is doing: it weighs your edge against the payout and spits out a stake size designed to maximize long-term growth.
The most important word in that last sentence is edge. Kelly only works if you genuinely have one – meaning your estimated probability of winning is more accurate than the odds the sportsbook is offering. If you don't have an edge, Kelly tells you to bet zero, and no amount of math creates an advantage that isn't there.
Say you're looking at a moneyline bet priced at decimal odds of 2.50 (that's +150 in American odds). The sportsbook's odds imply about a 40% chance of winning. But based on your own analysis, you believe the true probability is 50%. That gap is your edge.
Plug it in:
b = 2.50 – 1 = 1.5
p = 0.50
q = 0.50
f = (1.5 × 0.50 – 0.50) / 1.5 = (0.75 – 0.50) / 1.5 = 0.25 / 1.5 ≈ 0.167
Kelly says to stake about 16.7% of your bankroll on this bet. If your bankroll is $1,000, that's a $167 bet.
Now notice something important: 16.7% of your bankroll on a single bet is enormous by any sane bankroll-management standard. That number alone should tell you why full Kelly is rarely used as-is in real betting – more on that below. The formula is correct; the problem is that it assumes your 50% estimate is exactly right, and in sports betting, it almost never is.
Kelly's appeal is that it solves a real problem that flat-staking and gut-feeling both get wrong. Bet too small relative to your edge, and you leave growth on the table. Bet too big, and a normal losing streak can wipe you out before your edge plays out. Kelly is the mathematically optimal balance point between those two failures – it maximizes the long-term growth rate of your bankroll while, in theory, never risking total ruin.
This is why Kelly has a serious following beyond betting. It originated in information theory in the 1950s (developed by John Kelly at Bell Labs) and was later adopted by professional investors and gamblers, including some legendary names in both fields. The math is sound, and over a long enough horizon with an accurate edge, no other staking strategy grows a bankroll faster.
The catch – and it's a big one – is buried in that phrase "with an accurate edge." Kelly's output is only as good as your probability estimate going in. And that's exactly where sports bettors get into trouble.
In a casino game like blackjack, a card counter can calculate their edge precisely, because the math of the deck is fixed and knowable. Sports betting is nothing like that. Your "50% true probability" is an estimate, a guess informed by analysis but still a guess. Nobody knows the real probability of a sporting event.
This matters enormously because Kelly is extremely sensitive to overestimated edges. If you think your edge is bigger than it really is – which bettors do constantly, thanks to overconfidence – Kelly tells you to bet too much, and overbetting is precisely how bankrolls get destroyed. A small error in your probability estimate translates into a large error in your stake size, always in the dangerous direction.
Full Kelly also produces wild swings. Even when your edge is real, staking the full Kelly amount leads to gut-wrenching volatility – it's common to see large drawdowns of your bankroll along the way to long-term growth. Most people cannot stomach watching 40% or 50% of their bankroll evaporate during a downswing, even if the math says it'll recover. Emotional bettors abandon the system mid-drawdown, which is the worst possible time, locking in losses.
Because of these problems, almost nobody who uses Kelly seriously uses full Kelly. Instead, they use fractional Kelly – betting a fraction of what the formula recommends, most commonly half (Half Kelly) or a quarter (Quarter Kelly).
The reasoning is smart. Fractional Kelly dramatically reduces volatility and the risk of ruin while keeping most of the growth benefit. Research on Kelly shows that betting half of the recommended amount keeps roughly three-quarters of the growth rate while cutting the bankroll swings substantially. It also builds in a safety margin for the inevitable overestimation of your edge – if you're betting half Kelly, you can be meaningfully wrong about your edge and still not be overbetting catastrophically.
In our earlier example, full Kelly said 16.7% of bankroll. Half Kelly cuts that to about 8.3%, and Quarter Kelly to about 4.2%. Even Quarter Kelly's 4.2% is on the aggressive side for a single sports bet, which tells you most disciplined bettors effectively run a very conservative fraction of Kelly, if they use it at all.
Here's the honest, practical take. Kelly is a useful concept for serious, disciplined bettors who genuinely have a measurable edge and can estimate their probabilities with some rigor. For that group, fractional Kelly is a sensible framework for sizing bets in proportion to confidence. The principle alone – bet more when your edge is bigger, less when it's smaller, nothing when you have none – is worth absorbing even if you never touch the formula.
For most casual and recreational bettors, full Kelly is the wrong tool, and even fractional Kelly can be risky, for one simple reason: it requires you to accurately know your edge, and most bettors don't have a reliable edge in the first place. If you can't honestly estimate your probability better than the sportsbook can, Kelly has no edge to size and will likely just lead you to overbet on bad estimates.
A reasonable middle path for most people: use simple, flat bankroll management instead – staking a fixed small percentage (commonly 1–3%) of your bankroll per bet. It's not mathematically optimal, but it's far more forgiving of the estimation errors that real bettors make, and it keeps you in the game. If you later develop a proven, tracked edge over a large sample of bets, then a conservative fractional Kelly approach becomes worth considering.
Kelly is a bankroll-growth tool, not a winning system – it tells you how much to bet, never what to bet, and it cannot turn losing picks into profit. Whatever staking method you choose, a few things stay true. Only ever bet money you can afford to lose, set firm limits on your time and spending before you start, and treat betting as entertainment with a cost rather than a way to make money. Chasing losses by increasing stakes – the opposite of what any sound system tells you to do – is one of the clearest warning signs of a problem.
If betting is causing financial stress or stops feeling like a choice, support is available. In the US, the National Problem Gambling Helpline is reachable at 1-800-522-4700, free and confidential, 24/7.
Is the Kelly Criterion guaranteed to make me money? No. Kelly only optimizes how you size bets when you already have a genuine edge – it creates nothing on its own. If your picks aren't actually beating the odds, Kelly can't fix that, and using it on overestimated edges can lose money faster than flat staking.
What's the difference between full Kelly and fractional Kelly? Full Kelly stakes the exact percentage the formula recommends, which produces maximum growth but extreme volatility and high risk if your edge estimate is off. Fractional Kelly (like Half or Quarter Kelly) stakes a portion of that amount, sharply reducing the swings and the risk of ruin while keeping most of the growth. Serious users almost always use a fraction.
Why does Kelly sometimes recommend huge bets? Because the formula assumes your probability estimate is exactly correct. When you input a large edge, it returns a large stake. In real sports betting your edge is uncertain and usually smaller than you think, which is exactly why full Kelly's recommendations are often dangerously large and why fractional Kelly exists.
Is flat-percentage staking better than Kelly for beginners? For most beginners, yes. Staking a fixed small percentage (around 1–3%) per bet is simpler, far more forgiving of the estimation errors beginners inevitably make, and keeps your bankroll alive long enough to learn. Kelly becomes worth considering only once you have a proven, tracked edge.
Can I use the Kelly Criterion for parlays or accumulators? It's much harder and riskier. Parlays involve combined probabilities and larger effective edges-and-errors, so estimation mistakes compound. Most experienced bettors who use Kelly apply it to single bets where the probability estimate is cleaner, and treat parlays as low-stake entertainment rather than Kelly-sized positions.
The Kelly Criterion is a genuinely smart staking formula that sizes your bets according to your edge, maximizing long-term bankroll growth in theory. In practice, its biggest strength is also its biggest weakness: it depends entirely on knowing your true edge, which sports bettors almost never do precisely. Full Kelly is too aggressive for nearly everyone, fractional Kelly is the sensible real-world version, and flat-percentage staking is the safer choice for most recreational bettors. Understand the principle – bet in proportion to your edge, and bet nothing when you don't have one – and you'll be ahead of most bettors regardless of which formula you use. Whatever you choose, bet responsibly and only with money you can afford to lose.
Kelly, J.L., Bell System Technical Journal – A New Interpretation of Information Rate (original Kelly paper): https://ieeexplore.ieee.org/document/6771227
Thorp, E.O. – The Kelly Criterion in blackjack, sports betting, and the stock market: https://www.eecs.harvard.edu/cs286r/courses/fall12/papers/Thorpe_KellyCriterion2007.pdf
Investopedia – Kelly Criterion definition and formula explained: https://www.investopedia.com/articles/trading/04/091504.asp
National Council on Problem Gambling – Help and resources: https://www.ncpgambling.org/help-treatment/
Pinnacle – How to use the Kelly Criterion for betting (educational): https://www.pinnacle.com/en/betting-articles/Betting-Strategy/the-real-kelly-criterion/D5XAQGFFNZGB48VL