Bankroll Management for Sports Betting: Flat Staking vs Kelly vs Proportional

Three staking systems dominate serious sports betting. Flat staking bets a fixed dollar amount every game. Proportional staking bets a fixed percentage of your current bankroll. Kelly staking calculates a variable percentage based on your estimated edge and the odds. Each one makes a different assumption about what you know and what you are willing to risk. Getting the assumption wrong costs you real money.

This article shows the math for all three, compares them over a 500-bet sequence you can verify yourself, and identifies the one condition that makes full Kelly dangerous even when you have a real edge.

Three Systems, Three Assumptions

The first thing to understand is what each system assumes about the bettor.

Flat staking assumes you do not know your edge well enough to size bets dynamically. Every bet gets the same dollar amount. The bet size does not change as your bankroll grows or shrinks. This is the right starting point for bettors who track fewer than a few hundred bets and have no stable edge estimate.

Proportional staking assumes you have a consistent edge but no precise measurement of it. You bet a fixed percentage of your current bankroll, not a fixed dollar amount. The stake compresses automatically during a losing run and expands during a winning run. No edge estimate is needed.

Kelly staking assumes you know your edge with reasonable accuracy. The formula tells you exactly what fraction of your bankroll to bet to maximize long-run growth. The larger the edge, the larger the stake. But the system is only as good as the edge estimate feeding it.

Flat Staking

Flat staking is simple and transparent. Pick a unit size at the start of a betting season, for example 2% of your initial bankroll, and stick to it regardless of whether you are up or down.

On a $1,000 bankroll, flat staking at 2% means every bet is $20. A win at -110 American odds nets $18.18. A loss costs $20.

The math over 500 bets at a 55% win rate:

275 wins × $18.18 = $4,999.50 gained 225 losses × $20.00 = $4,500.00 lost Net profit: +$499.50 Final bankroll: $1,499.50

Two things are worth noting. First, the math is straightforward and predictable. Second, your nominal "2% stake" becomes a larger share of your bankroll after a drawdown. If your bankroll drops to $500, your $20 flat bet is 4% of remaining capital. Flat staking does not protect you from stake inflation during bad runs.

Proportional Staking

Proportional staking fixes this problem. You always bet the same percentage of your current bankroll, so stakes shrink after losses and grow after wins.

The expected growth per bet under proportional staking uses the log-growth formula. For a 2% proportional stake at -110 odds with a 55% win rate:

G = p × ln(1 + f × b) + q × ln(1 - f) G = 0.55 × ln(1.01818) + 0.45 × ln(0.98) G = 0.009909 - 0.009091 G = 0.000818 per bet p = win probability, f = stake fraction, b = net odds (10/11 at -110), q = loss probability After 500 bets: e^(500 × 0.000818) = e^0.409 ≈ 1.505 Final bankroll ≈ $1,505

At 2%, proportional staking produces almost the same dollar result as flat staking over this sequence. The difference shows up in bad variance scenarios. A 20-bet losing run on flat staking costs exactly $400 regardless of timing. The same losing run on proportional staking costs less as stakes compress with each loss, protecting a larger share of capital.

Kelly Staking

Kelly staking was derived by J.L. Kelly Jr. in his 1956 paper "A New Interpretation of Information Rate," published in the Bell System Technical Journal. The formula maximizes the long-run exponential growth rate of a bankroll.

f* = (b × p - q) / b f* = optimal fraction of bankroll to bet b = net odds (profit per $1 staked) p = probability of winning q = probability of losing (1 - p) At -110 American odds: b = 10/11 ≈ 0.9091 With 55% win rate: p = 0.55, q = 0.45 f* = (0.9091 × 0.55 - 0.45) / 0.9091 f* = (0.5000 - 0.45) / 0.9091 f* ≈ 5.5% of bankroll

For a bettor at -110 with a 55% win rate, full Kelly prescribes roughly 5.5% of bankroll per bet. The expected growth:

G = 0.55 × ln(1.05) + 0.45 × ln(0.945) G = 0.026835 - 0.025456 G = 0.001379 per bet After 500 bets: e^(500 × 0.001379) = e^0.6895 ≈ 1.993 Final bankroll ≈ $1,993

Full Kelly more than doubles the expected bankroll over 500 bets where flat staking returns 50%. That is the appeal. But the expected value comparison misses a critical problem addressed in the next section.

Edward Thorp's 2006 book chapter "The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market" demonstrated the asymmetry of the Kelly growth curve: overbetting is worse than underbetting. Betting more than the Kelly fraction reduces the expected growth rate, and doing so by enough turns the bankroll toward zero even with a positive edge. Betting less than Kelly reduces growth, but preserves capital. The function is not symmetric around the optimum.

The 500-Bet Comparison

Here is the full comparison across all five common approaches, all using the same setup: 55% win rate at -110 American odds, starting bankroll of $1,000, over 500 bets.

Strategy Stake per Bet Expected Final Bankroll Growth
Flat (2% of starting bankroll) $20.00 fixed $1,500 +50%
Proportional (2% of current bankroll) ~$20 → adjusts $1,505 +51%
Quarter Kelly (1.4% of bankroll) ~$14 → adjusts $1,352 +35%
Half Kelly (2.75% of bankroll) ~$28 → adjusts $1,677 +68%
Full Kelly (5.5% of bankroll) ~$55 → adjusts $1,993 +99%

On paper, full Kelly wins easily. In practice, it is the system most likely to wipe you out.

A 2023 simulation study by Beggy, Kim, Mucaj, and Nordell published through the Wharton Sports Analytics Initiative found that full Kelly led to bankruptcy in 100% of simulated realistic betting scenarios. The same study found that partial Kelly with a coefficient of 0.50 and a conservative selection threshold produced roughly 80% annualized returns over an 11-year simulation period.

An empirical study published by Springer in their ISACE 2026 proceedings examined 17,403 football matches from the top five European leagues between 2015 and 2025. Full Kelly produced losses exceeding 80% of starting bankroll. The optimal fractional Kelly coefficient in that dataset was 15% of Kelly, which generated +10.5% ROI with drawdowns below 52%.

The gap between expected-value calculations and real-world results has one cause: edge estimation error.

Edge Estimation Error: Where Kelly Breaks Down

The Kelly formula gives you the correct fraction when you know your edge correctly. When your edge estimate is wrong, Kelly amplifies the error.

Here is a concrete example. You track 300 bets, see a 55% win rate, and calculate your Kelly fraction as 5.5%. But your sample is too small and your real long-run win rate is 52%. What happens over the next 500 bets?

First, check whether 52% is even a positive edge at -110:

Break-even win rate at -110: 52.38% Your actual win rate: 52% True Kelly fraction: (0.9091 × 0.52 - 0.48) / 0.9091 = -0.008 → bet zero A 52% win rate at -110 is a small negative edge. True Kelly says do not bet.

But you believe you have a 55% edge, so you bet 5.5% of bankroll per game. The log-growth function tells you exactly what happens:

G = 0.52 × ln(1.05) + 0.48 × ln(0.945) G = 0.025371 - 0.027154 G = -0.001783 per bet (negative) After 500 bets: e^(-0.8915) ≈ 0.41 Final bankroll: $410 Starting from $1,000, you lose 59% of your bankroll.

Now compare what flat staking does with the same incorrect 52% win rate:

260 wins × $18.18 = $4,727.27 gained 240 losses × $20.00 = $4,800.00 lost Net: -$72.73 Final bankroll: $927.27 A 7% loss. The error is contained.

A three-point error in your win rate estimate, a difference between thinking you are at 55% when you are at 52%, costs flat staking 7% over 500 bets. The same error costs full Kelly 59%.

This is not a flaw in the Kelly formula itself. The formula is correct given correct inputs. The flaw is the assumption that you know your edge precisely, which almost no bettor does from a sample of 300 to 500 bets.

A 300-bet sample at 55% win rate has a wide confidence interval. The 95% confidence interval around a 55% win rate over 300 bets runs roughly from 49.3% to 60.7%. That range includes multiple scenarios where full Kelly bets destroy your bankroll.

The Interactive Staking Calculator

Staking Comparison Calculator

What Sharp Bettors Use and Why

The research converges on one answer: fractional Kelly, calibrated to account for edge uncertainty.

Uhrín, Šourek, Hubáček, and Železný at Czech Technical University reviewed optimal betting strategies across horse racing, basketball, and soccer in their 2021 paper published at arxiv.org. They found that adaptive fractional Kelly outperformed both flat staking and full Kelly across all three sports. The fraction varied by setting, but reducing the theoretical optimal fraction was necessary in every case.

Thorp's analysis shows why. The Kelly growth curve is steeper on the "under-bet" side than on the "over-bet" side when you move away from the optimum. Betting half Kelly costs you 25% of your maximum growth rate. Betting double Kelly costs you your entire bankroll given enough time, even with a real positive edge.

The practical question is not "Kelly or not Kelly" but "how much of Kelly." Four factors narrow the answer:

Sample size of your edge estimate. Fewer than 500 bets means wide uncertainty. At 500 bets, even a real 55% win rate has a standard error of about 2.2 percentage points. Use a lower fraction until you have 1,000 or more bets with a stable record.

Correlation between bets. Kelly assumes independent bets. Same-game parlays, multi-game parlays, and bets within the same game or night are correlated. Correlated bets require a lower effective Kelly fraction or separate tracking as a single event.

Your actual vs estimated win rate. The edge estimation error scenario above shows what a three-point overestimation does. If you are uncertain whether your 55% sample will hold, cut the Kelly fraction in half or more.

Account longevity goals. Full Kelly produces the fastest expected growth but also the largest swings. A drawdown of 50% or more is not unusual at full Kelly even with a genuine edge. A bettor who closes accounts during drawdowns, or who loses discipline during bad variance, will perform better at a lower fraction.

Quarter Kelly (25% of the calculated optimal fraction) is the starting point most practitioners recommend for recreational bettors with fewer than 1,000 tracked bets. It gives you about 35% of full Kelly's expected growth while absorbing edge estimation errors that would bankrupt a full Kelly bettor.

Which System to Choose

Three different bettors warrant three different answers.

You are new to tracking bets or have fewer than 500 in your record. Use flat staking at 1% to 2% of your initial bankroll. Track every bet. Build the sample. Do not use Kelly until you have enough data to know whether your edge estimate is real.

You have a consistent edge but do not trust your win rate estimate enough to size bets dynamically. Use proportional staking at 1% to 2% of your current bankroll. The auto-compression during losing runs provides downside protection. The compounding during winning runs improves on flat staking. No edge estimate required.

You have tracked 1,000 or more bets with a documented, stable positive return. Start at quarter Kelly and move toward half Kelly as your sample grows and your edge estimate stabilizes. Never use full Kelly unless you have a very large sample, an edge you can measure precisely, and the psychological tolerance for 50%+ drawdowns.

The math of Kelly is correct. The assumption of precise edge knowledge is almost never correct. That gap between the theory and the input quality is where bankrolls go to zero.

Key Numbers to Carry

At 55% win rate, -110 American odds, $1,000 starting bankroll, 500 bets:

  • Flat staking at 2%: $1,500 expected final bankroll
  • Proportional staking at 2%: $1,505 expected final bankroll
  • Quarter Kelly: $1,352 expected final bankroll
  • Half Kelly: $1,677 expected final bankroll
  • Full Kelly: $1,993 expected final bankroll

If your actual win rate is 52%, not 55%, full Kelly produces a final bankroll of $410 over those same 500 bets. Flat staking produces $927. The edge estimation error costs full Kelly 59% of the bankroll and costs flat staking 7%.

The Kelly formula maximizes expected growth given accurate inputs. In the real world, bettors rarely have accurate inputs. Fractional Kelly is not a compromise. It is the correct answer to a problem with uncertain inputs.

The breakeven win rate at -110 is 52.38%. If your actual win rate is below that, no staking system generates profit. Focus first on finding a genuine edge. Staking systems are a multiplier on edge, not a substitute for it.

For the full derivation of the Kelly formula and a deeper look at why quarter-Kelly became the standard among professional bettors, see the Kelly criterion guide in The Kennel. For context on why your edge estimate depends on closing line value, see the CLV guide.

Sources

  1. Kelly, J. L. (1956). A New Interpretation of Information Rate. Bell System Technical Journal, 35(4), 917-926. Open-access mirror: archive.org/details/bstj35-4-917.
  2. Thorp, E. O. (2006). The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market. In Handbook of Asset and Liability Management, Vol. 1, pp. 385-428. Elsevier Science. Semantic Scholar record.
  3. Beggy, J., Kim, D., Mucaj, K., Nordell, J. (2023). An Investigation of Sports Betting Selection and Sizing. Wharton Sports Analytics Initiative. wsb.wharton.upenn.edu.
  4. Risk Parity vs. Kelly Criterion: An Empirical Evaluation of Bankroll Allocation Strategies in Football Value Betting. (2026). Springer ISACE Lecture Notes in Computer Science. link.springer.com. 17,403 matches, 2015-2025.
  5. Uhrín, M., Šourek, G., Hubáček, O., Železný, F. (2021). Optimal sports betting strategies in practice: an experimental review. arXiv, 2107.08827. arxiv.org/abs/2107.08827.
  6. Kim, S.-K. (2024). Kelly Criterion Extension: Advanced Gambling Strategy. MDPI Mathematics, 12(11), 1725. mdpi.com.