Kelly Criterion for Basketball Betting: Optimal Stake Sizing Mathematics
Available in US
Early in my betting career, I discovered a spread I was certain offered value. My model suggested 60% win probability at the available odds – a substantial edge. So I bet heavily. Very heavily. The bet lost, and that single loss wiped out weeks of careful, profitable betting. The edge was real, but my stake sizing was catastrophically wrong. That’s when I started studying Kelly.
The Kelly Criterion is a mathematical formula for optimal bet sizing that maximises long-term bankroll growth. Developed by John Kelly at Bell Labs in the 1950s, it tells you exactly what percentage of your bankroll to stake given your edge and the odds offered. In theory, Kelly staking produces faster wealth accumulation than any other approach while theoretically avoiding bankruptcy. In practice, it’s more complicated.
Basketball bettors face particular challenges applying Kelly. We estimate edges rather than knowing them precisely. Our win probability assessments contain error. Spreads involve push possibilities that the basic formula doesn’t elegantly handle. And full Kelly staking produces volatility that most bettors – including professionals – find psychologically unbearable.
This guide covers the mathematics of Kelly betting, why most practitioners use fractional approaches, and how to adapt these concepts to NBA and basketball betting specifically.
The Kelly Formula Explained
The basic Kelly formula is: f = (bp – q) / b, where f is the fraction of bankroll to bet, b is the decimal odds minus 1 (your net profit per unit if you win), p is your probability of winning, and q is your probability of losing (1 – p). For a bet at decimal odds of 2.00 where you estimate 55% win probability, Kelly suggests: (1 * 0.55 – 0.45) / 1 = 0.10, or 10% of your bankroll.
The formula optimises for logarithmic utility of wealth, which means it maximises the expected growth rate of your bankroll over time. It’s mathematically optimal in the sense that no other fixed-fraction staking strategy produces faster long-term growth given accurate edge estimates. This optimality has made Kelly influential among serious bettors and investors.
Kelly naturally scales stakes to edge size. Large edges warrant large bets; small edges warrant small bets; zero or negative edges produce zero or negative stake recommendations (meaning don’t bet). This proportionality makes intuitive sense and prevents the common mistake of betting the same amount regardless of edge quality.
The formula also has a critical property: it theoretically prevents bankruptcy. Because you’re always betting a fraction of your current bankroll, you can never lose everything on a single bet. A losing streak reduces your stakes proportionally, preserving capital for recovery. This contrasts with fixed-stake approaches where a bad run can eliminate your entire bankroll.
The catch is that Kelly assumes you know your true edge. In sports betting, you don’t – you estimate it. If your 55% estimate is actually 52%, Kelly tells you to overbet massively. If your estimate is actually 58%, Kelly tells you to underbet. The formula’s optimality depends on inputs that you can never know with certainty.
Fractional Kelly: Reducing Volatility
Full Kelly staking produces terrifying volatility. Even with accurate edge estimates, drawdowns of 50% or more occur regularly. Sequences of losses that feel impossible happen frequently enough over long betting careers to guarantee you’ll experience them. Most bettors – including professionals who understand the math – can’t tolerate this variance emotionally.
Fractional Kelly addresses this by betting a fixed fraction of the Kelly-recommended stake. Half-Kelly (50% of the recommended bet) cuts volatility substantially while sacrificing relatively little expected growth. Quarter-Kelly provides even smoother equity curves. The sacrifice in growth rate is sublinear – half-Kelly gives you more than half the growth rate of full Kelly because reduced volatility prevents the deep drawdowns that slow compounding.
The choice of Kelly fraction involves a trade-off between growth and comfort. Aggressive bettors might use 50-75% Kelly; conservative bettors might use 25% or less. The right fraction depends on your risk tolerance, your confidence in your edge estimates, and how much variance you can endure without abandoning your strategy during inevitable losing stretches.
I use quarter-Kelly for most of my basketball betting. This feels excessively conservative to aggressive bettors, but it lets me sleep during bad runs and maintain discipline when variance works against me. The theoretical growth sacrifice is modest; the psychological benefits of reduced volatility are substantial. My longest losing streaks have never threatened my ability to continue betting at meaningful stakes.
Another approach adjusts Kelly fraction based on confidence. High-confidence edge estimates might warrant half-Kelly; uncertain situations might warrant eighth-Kelly or less. This adaptive approach acknowledges that not all edge estimates are equally reliable, sizing bets proportionally to conviction as well as estimated edge size.
Applying Kelly to NBA Betting Decisions
Basketball betting presents specific challenges for Kelly application. Spread bets can push, creating three possible outcomes rather than two. The basic Kelly formula assumes binary win/lose outcomes; pushes complicate the math. Various modifications exist to handle pushes, but they all involve either approximation or more complex calculations than the simple formula.
Edge estimation is the critical input, and it’s inherently uncertain. Your model might suggest 54% win probability on a spread bet, but that estimate has a confidence interval. Maybe you’re 90% confident the true probability is between 51% and 57%. Full Kelly on a 54% estimate produces dramatically different stakes than Kelly on a 51% estimate. Conservative bettors should use the lower bound of their confidence interval rather than the point estimate.
Correlated bets require adjustment. If you’re betting three NBA games on the same night, and all three involve betting against publicly overvalued teams, your bets are partially correlated – they might all lose together if your public overvaluation thesis is wrong. Kelly assumes independent bets; correlated exposure should be treated as partially overlapping, requiring stake reductions.
Practical implementation involves calculating Kelly for each bet, summing total recommended exposure, and potentially scaling down if total exposure exceeds comfortable levels. If Kelly suggests 5% on three different games, the combined 15% exposure might feel excessive. Many bettors cap total daily or weekly exposure regardless of what Kelly recommends for individual bets.
Tracking results against Kelly recommendations helps calibrate your approach. If you’re consistently overbetting relative to actual edges (meaning your estimated edges exceed your actual win rates), you need either better edge estimation or smaller Kelly fractions. Building this feedback loop improves both your staking and your edge estimation over time.
When Kelly Gets It Wrong
Kelly’s greatest vulnerability is edge estimation error. The formula takes your probability estimate as truth and calculates accordingly. If you believe you have a 57% edge but actually have a 50% edge, Kelly tells you to bet heavily on a coin flip. This isn’t Kelly’s fault – it’s optimally processing the wrong information – but the outcome for your bankroll is disastrous.
Overconfidence in edge estimates is the most common betting mistake, and Kelly amplifies this error. Most bettors overestimate their edge at least some of the time. When you apply Kelly to inflated edge estimates, you bet too much on positions that don’t warrant the exposure. The formula can’t protect you from yourself.
Model uncertainty compounds estimation error. Your model might be systematically biased in ways you don’t recognise. If it consistently overvalues certain teams or underweights certain factors, every bet benefits from that bias, and Kelly tells you to overbet systematically. By the time you recognise the model flaw, significant bankroll damage may have occurred.
Liquidity constraints rarely bind in recreational betting but matter for larger operations. Kelly might recommend a stake larger than markets can absorb at the available price. Moving substantial volume shifts lines against you, meaning the odds change as you bet. True edge decreases as you bet more, making full Kelly recommendations invalid at scale.
The psychological reality of drawdowns exceeds what mathematical descriptions convey. Knowing that 50% drawdowns are expected under full Kelly is different from experiencing one. Most bettors abandon their strategies during severe drawdowns, locking in losses at the worst possible time. Kelly’s theoretical optimality assumes you’ll maintain discipline through any variance – an assumption that fails for most humans.
Finding Your Optimal Staking Approach
Kelly provides a mathematical framework rather than rigid rules. Understanding why Kelly works – scaling stakes to edge, preventing bankruptcy, maximising growth – matters more than mechanically applying the formula. You can capture Kelly’s benefits through approaches inspired by its logic without strict adherence to its output.
Start conservative. New bettors don’t know their actual edge, and even experienced bettors overestimate. Quarter-Kelly or less provides a safety margin that protects against estimation error while still scaling stakes appropriately to perceived edge quality. You can always become more aggressive as your track record provides evidence that your edge estimates are calibrated.
Combine Kelly thinking with flat-bet simplicity. Instead of calculating precise Kelly fractions, categorise bets by confidence level: low confidence gets 0.5 units, medium confidence gets 1 unit, high confidence gets 2 units. This captures Kelly’s proportionality in a simpler framework that’s easier to implement consistently.
The comprehensive bankroll management guide covers how Kelly concepts integrate with broader financial management for basketball betting. Stake sizing is one component of a sustainable betting practice that also includes budgeting, record keeping, and psychological discipline.
What is the Kelly criterion formula?
The basic Kelly formula is f = (bp – q) / b, where f is the recommended fraction of bankroll to stake, b is the decimal odds minus 1, p is your estimated win probability, and q is your loss probability. For a bet at 2.00 odds with 55% estimated win probability, Kelly recommends staking 10% of your bankroll.
Should I use full or fractional Kelly?
Most serious bettors use fractional Kelly, typically 25-50% of the full Kelly recommendation. Full Kelly produces optimal long-term growth mathematically but creates extreme volatility that most bettors cannot tolerate emotionally. Fractional Kelly sacrifices modest growth for dramatically reduced variance and drawdowns that won’t derail your strategy.
Is Kelly too risky for most bettors?
Full Kelly is too risky for most bettors because it assumes perfect edge estimation, which is impossible in sports betting. When you overestimate your edge – which happens frequently – Kelly tells you to overbet, potentially causing serious bankroll damage. Conservative Kelly fractions like quarter-Kelly provide Kelly’s benefits while building in protection against estimation error.
Prepared by the Best Basketball Betting Strategy editorial staff.
