Kelly Criterion for First Basket Staking: A Fractional Approach for UK Bettors
Available in USThe Formula That Tells You How Much, Not Whether
The first time I encountered the Kelly criterion I made the mistake of trying to apply it at full strength to a first basket bet, and I lost roughly a fifth of my month’s bankroll on a single Tuesday-night fixture that had nothing structurally wrong with my edge calculation. The math was right. The application was reckless. Kelly is a formula for optimal bet sizing when you know your edge with confidence, and “with confidence” is not a phrase that applies to high-variance markets like NBA first basket.
The criterion itself is straightforward. It tells you the percentage of your bankroll to stake on a bet given the price you are getting and the true probability you have estimated. Apply it correctly and over the long run you grow your bankroll faster than any other staking system. Apply it incorrectly – particularly by trusting your edge estimate too much, which is the most common error – and you will be ruined faster than any other staking system. For first basket markets specifically, the working version of Kelly that survives contact with reality is the fractional version, typically Kelly divided by four, and even that requires bankroll discipline most UK bettors do not have at the start.
This piece walks through the formula, why full Kelly is dangerous on first basket markets, why fractional Kelly at one-quarter strength is the working default, a worked example with realistic numbers, and the comparison with flat staking that most casual UK bettors use. The aim is to give you a staking framework that survives a normal first basket loss streak without breaking your bankroll, while still giving you the compounding advantages of edge-aware sizing.
The Kelly Formula Explained
The Kelly criterion in its standard form is: stake percentage equals (bp – q) divided by b, where b is the decimal odds minus 1 (the net profit per unit staked), p is the probability of winning, and q is the probability of losing (which equals 1 minus p). The output is the percentage of your current bankroll to stake on this bet for optimal long-run growth.
A worked numerical example: you estimate a player has a 22% probability of scoring the first basket. The book is offering decimal odds of 6.0, which means b = 5 (6.0 minus 1). Plug in: (5 × 0.22) – 0.78 = 1.10 – 0.78 = 0.32. Then divide by b: 0.32 / 5 = 0.064. Full Kelly tells you to stake 6.4% of your bankroll on this bet. On a £1000 bankroll, that is a £64 stake.
The formula has elegance because it self-regulates. If your estimated probability equals the implied probability from the price (no edge), the formula returns zero – Kelly tells you not to bet. If you have a small edge, Kelly tells you to bet a small percentage. If you have a large edge, Kelly tells you to bet a larger percentage. The output scales with both your edge and your confidence in the price.
The mathematical proof behind Kelly was developed in the 1950s in a context of information theory and gambling, and it is the only staking formula with a rigorous proof of long-run optimality given perfect knowledge of probabilities. The phrase “perfect knowledge” is doing a lot of work in that sentence, and it is what breaks the formula in real-world application to high-variance markets.
Why Full Kelly Is Dangerous Here
Full Kelly is mathematically optimal only when your probability estimate is exactly correct. In any market where your probability estimate has uncertainty around it, full Kelly oversizes your bets relative to the actual edge. The oversize is amplified by variance – markets with high variance have more painful drawdowns even when the edge is real, and full Kelly does not account for the bettor’s psychological tolerance for those drawdowns.
First basket markets are about as high-variance as NBA prop markets get. Even the most likely first basket scorer in any given game converts at less than 20%, while a player priced as 15% likely to score first will miss roughly 85% of attempts. Those numbers are sobering when you start applying them to staking. If your bet has a true 22% probability of winning, you will lose 78% of similar bets. A staking system that requires you to size confidently against a 78% loss rate had better be calibrated for the loss streaks that come with that probability profile.
The math of loss streaks at this probability is uncompromising. At 22% true probability of winning, the chance of losing eight bets in a row is 0.78 to the eighth power, or about 14%. The chance of losing ten in a row is around 8.5%. Across a season of 200 first basket bets, you will hit at least one ten-loss streak more often than not. Full Kelly applied to these conditions produces drawdowns that, even with positive expected value, take the bettor out of the game emotionally and sometimes financially before the long-run advantage materialises.
The other danger is estimation error. Full Kelly assumes your probability estimate is correct. In reality, your estimate has an error band – perhaps you think a player is 22% likely but the true probability is 19%. Full Kelly applied to a 22% estimate is the right size if 22% is correct, but is the wrong size – too large – if true probability is 19%. The systematic effect of estimation error in a high-variance market is that full Kelly chronically overstakes, and the overstaking compounds across many bets.
Fractional Kelly at One-Quarter as Default
The working solution for high-variance first basket markets is fractional Kelly, typically at one-quarter strength. Calculate the full Kelly stake and bet 25% of it. So in the worked example above, full Kelly said 6.4% of bankroll; quarter Kelly says 1.6% of bankroll. On a £1000 bankroll that drops the £64 stake to £16. Less aggressive on every bet, but radically more survivable across loss streaks.
The mathematical property of quarter Kelly is that you give up a meaningful share of the long-run growth rate that full Kelly delivers, but you also reduce variance dramatically. The trade-off is roughly: full Kelly maximises growth but carries severe drawdown risk; half Kelly retains most of the growth at much lower drawdown; quarter Kelly retains roughly two-thirds of the growth at drawdowns most bettors can actually tolerate. For markets where probability estimation is uncertain – which is every first basket market – quarter Kelly is the regime that survives.
The other property of fractional Kelly is psychological. Bettors who are constantly hitting Kelly-sized losing streaks will at some point break discipline – either by abandoning the system, by reducing stakes below quarter Kelly informally, or by chasing losses with stakes above the system. Quarter Kelly produces a stake size that most UK bettors can stomach for the duration of a real loss streak without breaking, and that survivability is the actual edge of the system. The most theoretically optimal staking plan you abandon halfway through is worse than a less optimal plan you stick with.
A Worked Example with Real Numbers
Let me run a realistic UK first basket scenario through the framework. Jalen Brunson posted a 21.2% first-basket rate across 80 starts and took 23.8% of his team’s opening shots, with the New York Knicks winning the opening tip at 53.4% and converting that to a 61.4% team first-basket rate. Suppose tonight’s matchup gives him a typical opening alignment, the Knicks are facing an opponent of average tipoff strength, and your model produces a 22% probability for Brunson to score first.
The book is offering Brunson at 4/1, which is decimal 5.0, implied probability 20%. Your estimated 22% versus the implied 20% is a 2-point edge. Plug into Kelly: b = 4, p = 0.22, q = 0.78. Calculation: (4 × 0.22) – 0.78 = 0.88 – 0.78 = 0.10. Divided by b: 0.10 / 4 = 0.025. Full Kelly stake is 2.5% of bankroll. Quarter Kelly stake is 0.625% of bankroll.
On a £1000 bankroll, full Kelly says stake £25. Quarter Kelly says stake £6.25. The £6.25 figure looks small relative to the £25 figure, but it is the size that survives the realistic loss profile of first basket betting at a 22% true probability rate. Across a 200-bet season at this stake size, a typical realised result might be 45 wins and 155 losses, generating roughly 45 × £25 wins minus £6.25 across all 200 bets – the math works out to a meaningful positive return on quarter Kelly applied consistently, even though the win rate looks bleak in isolation.
The same calculation across a portfolio of bets per night produces the staking discipline that distinguishes survivable bankrolls from unsurvivable ones. If you are placing four first basket bets on a given night with edges of similar magnitude, you do not stake 0.625% on each – you stake the calculated quarter Kelly per bet but cap the total nightly exposure at, typically, 5%-7% of bankroll. Even quarter Kelly without portfolio caps can produce nights where you have committed too much bankroll to a single slate’s correlated outcomes.
Kelly vs Flat Staking
Most casual UK bettors stake flat – the same amount on every bet regardless of edge or odds. Flat staking has the virtues of simplicity and discipline; it cannot oversize because it never adjusts. The disadvantage is that flat staking ignores the information about edge and price that should drive bet sizing. A 2-point edge at 5/1 and a 5-point edge at 5/1 are not the same bet, and flat staking sizes them identically.
The comparison over a long sample favours quarter Kelly meaningfully. On a portfolio of first basket bets with realistic distribution of edges, quarter Kelly delivers 30%-40% better long-run return than flat staking at equivalent average stake size, primarily because it concentrates capital in the highest-edge opportunities and avoids overcommitting to thin-edge bets. The trade-off is more stake-size variability and slightly higher peak-to-trough drawdown than flat staking, although the drawdown gap is much narrower than for full Kelly.
For UK first basket bettors deciding between flat and Kelly, the practical recommendation depends on calibration confidence. If your probability estimates are reliable across a large sample, quarter Kelly is the better system. If your probability estimates have not been validated against historical results, flat staking is the safer default – flat staking does not punish you for miscalibrated edge estimates, while Kelly does.
The drawdown profile of even a well-calibrated first basket Kelly approach can still take you through stretches that look like the system is broken when it is not, and understanding the realistic loss-streak distributions is essential for staying disciplined through them – I have laid out the math of expected drawdowns and how to size bankrolls to survive them in drawdown and loss streaks in first basket betting.
Kelly as a Tool, Not a Religion
Kelly is a powerful staking framework when applied with the discipline of fractional sizing and the humility to recognise that probability estimates are uncertain. Used well, it grows bankrolls faster than flat staking with manageable drawdowns. Used badly – at full strength, with overconfident estimates, in markets too high-variance for the assumption set – it destroys bankrolls. The difference between the two outcomes is application discipline, not formula sophistication.
For UK first basket bettors, the working recommendation is quarter Kelly with portfolio caps, validated probability estimates, and a willingness to drop to flat staking when calibration confidence is low. The math will compound in your favour over a real season’s volume, but only if you respect the variance and stay sized for the loss streaks that come with the territory.
Should I use full Kelly if my edge estimates are very accurate?
Even with accurate edge estimates, full Kelly produces drawdowns most bettors cannot stomach across realistic first basket variance. Quarter Kelly retains most of the long-run growth advantage with much lower psychological and bankroll risk. Half Kelly is occasionally appropriate for very calibrated bettors with strong drawdown tolerance, but full Kelly is rarely the right answer for high-variance NBA prop markets.
How do I calculate Kelly stakes for multiple first basket bets in the same game?
Calculate the Kelly stake for each bet independently, then apply a portfolio cap on total game exposure – typically 3% to 5% of bankroll across all bets in a single game. The cap prevents overexposure when multiple positively-priced selections produce correlated outcomes you have not fully modelled.
Prepared by the nba First Basket Bets editorial staff.
