Flat Staking vs Kelly Criterion: A Practical Comparison for Serious Bettors
- Kelly Criterion maximizes long-run bankroll growth but requires accurate edge estimation — miscalculate your edge and Kelly becomes destructive
- Flat staking (fixed unit size) is simpler, more robust to edge estimation errors, and psychologically easier to sustain
- Fractional Kelly (25–33% of full Kelly) is the practical middle ground most professionals use — Kelly growth with reduced variance
- For bettors who can accurately estimate their edge: fractional Kelly. For bettors uncertain about their edge: flat staking until you have sufficient data
- The difference in long-run performance between flat and fractional Kelly is real but smaller than many expect at typical edge levels
What Each System Does
Flat Staking
Flat staking means betting a fixed amount (or fixed percentage of bankroll) on every bet, regardless of the perceived edge. If you bet 2% of bankroll per bet, every bet is 2% — a heavily favoured price gets the same size as a slight edge on an uncertain price.
The key property of flat staking: it separates the staking decision from the edge estimation problem. You need to know whether you have edge to bet at all — but you don't need to quantify the exact size of that edge to size the bet.
Kelly Criterion
The Kelly Criterion calculates the mathematically optimal fraction of bankroll to stake on each bet, given your estimated edge. The formula:
f = (b × p − q) / b
Where: f = fraction of bankroll to stake | b = decimal odds − 1 (net profit per unit) | p = estimated probability of winning | q = 1 − p
Kelly tells you the exact stake that maximises long-run bankroll growth. But it requires knowing p — your true probability estimate — with reasonable accuracy. If p is wrong, Kelly is wrong, and the consequences can be severe.
The Core Trade-off
| Factor | Flat Staking | Full Kelly | Fractional Kelly (33%) |
|---|---|---|---|
| Long-run growth (if edge estimate is correct) | Sub-optimal | Maximum | Strong (75–90% of Kelly max) |
| Drawdown during variance | Predictable, controlled | High — can reach 50%+ during bad runs | Moderate |
| Sensitivity to edge estimate errors | None | High — overestimating edge is ruinous | Low-moderate |
| Stake consistency | Consistent (fixed %) | Varies per bet significantly | Varies moderately |
| Implementation complexity | Simple | Complex | Moderate |
The Edge Estimation Problem
Kelly's theoretical superiority depends entirely on accurate edge estimation. In practice, estimating your true probability advantage on a given bet is extremely difficult. The number you use as "p" in the Kelly formula is based on your model or judgment — and models have errors, especially on specific events.
Bet: Manchester City -0.5 AH at odds 1.90
True probability: 57% (genuine +EV bet)
Kelly stake: (0.90 × 0.57 − 0.43) / 0.90 = (0.513 − 0.43) / 0.90 = 9.2% of bankroll
Overestimated probability: 65% (overconfident model)
Kelly stake: (0.90 × 0.65 − 0.35) / 0.90 = (0.585 − 0.35) / 0.90 = 26.1% of bankroll
If the true edge is 7% but you're staking 26% of bankroll per bet, you're massively over-staked. Three consecutive losses = loss of 61% of bankroll, even though each bet was individually +EV.
This is why full Kelly is rarely used by professionals in practice. The asymmetry of outcomes (bankroll goes to zero faster than it compounds) combined with edge estimation uncertainty makes full Kelly too aggressive for most real-world betting environments.
Access Asian Bookmakers Through a Single Account
AsianConnect gives you access to PS3838, SBOBET, ISN, MaxBet and more from one wallet — the widest Asian book coverage of any broker. Competitive commission from 0.5%.
Open AsianConnect AccountFractional Kelly: The Professional Standard
Most professionals who use Kelly use a fraction of it — typically 25–33% of the full Kelly stake. The rationale:
- If your edge estimate is correct: fractional Kelly gives 75–90% of maximum growth rate with substantially lower variance
- If your edge estimate is 2× what it actually is (common with overconfident models): fractional Kelly still produces positive expected growth, while full Kelly would be over-bet
- Psychologically: smaller drawdowns are easier to sustain without abandoning the system during a losing run
Full Kelly: 9.2% of bankroll
33% Fractional Kelly: 3.1% of bankroll
Flat stake equivalent (2% rule): 2.0%
At this stake level, the difference between flat and fractional Kelly is 1.1 percentage points per bet — meaningful over 1,000 bets but not dramatic in any single session.
When Flat Staking Makes More Sense
Flat staking is the better choice when:
- You have fewer than 1,000 historical bets with your current methodology — insufficient data to estimate edge precisely
- You're testing a new approach or market — edge estimation is speculative until proven in practice
- Your bet selection is qualitative rather than model-driven — estimating exact probabilities requires more precision than qualitative judgment provides
- You want to simplify operations — running multiple strategies simultaneously is easier to manage with flat stakes
A 2% flat stake per bet is the most common starting point for serious bettors. It survives 50 consecutive losses (theoretically — the probability approaches zero as losses accumulate), allows for 50 active positions simultaneously if needed, and removes any over-staking error from the equation.
Practical Comparison Over 500 Bets
Starting bankroll: €10,000 | Average odds: 1.90 | True win rate: 53%
| Staking Method | Ending Bankroll (Median) | Max Drawdown (Typical) |
|---|---|---|
| Flat 2% | ~€16,500 | ~18% |
| 33% Kelly | ~€19,200 | ~22% |
| Full Kelly | ~€22,000 (but high variance) | ~45%+ |
Figures are illustrative medians from simulation — individual outcomes vary significantly. Key takeaway: 33% Kelly outperforms flat by ~16% over 500 bets, while full Kelly offers only ~15% more than fractional but with 2× the typical drawdown.
The Decision Framework
Use the following framework to choose your staking approach:
- Do you have a quantitative edge estimate backed by 500+ historical bets? → Use fractional Kelly (25–33%)
- Are you testing a new approach with limited data? → Use flat staking (1–2%) until you have edge evidence
- Is your edge estimate highly uncertain? → Use flat staking or low fractional Kelly (10–15%)
- Are you running multiple concurrent strategies? → Flat staking simplifies operations; fractional Kelly requires per-strategy edge estimates
FAQ — Flat Staking vs Kelly
Does Kelly Criterion guarantee higher profits than flat staking?
In theory, yes — Kelly maximizes long-run growth for a bettor with a known, constant edge. In practice, no — because edge estimates are imprecise, and overestimating your edge while applying full Kelly causes severe over-staking. Fractional Kelly typically outperforms flat staking over large samples, but the advantage requires reasonably accurate edge estimates to materialise.
What percentage of bankroll is a safe flat stake?
1–3% per bet is the standard range. 2% is the most common professional starting point. Above 3%, variance risk starts to impact long-run outcomes materially. At 5% flat stakes, a 20-bet losing run (statistically certain to occur over enough bets) reduces bankroll by 64%, making recovery difficult without additional capital.
Should I adjust flat stakes when my bankroll grows?
Yes. Flat staking as a percentage of bankroll (not a fixed amount) automatically adjusts as your bankroll grows or shrinks. Betting €200 when your bankroll is €10,000 (2%) and continuing to bet €200 when your bankroll has grown to €20,000 (1%) under-stakes your edge. Recalibrate your unit size regularly — monthly or after significant bankroll movements.
Do professional bettors actually use Kelly?
Some do, almost always in fractional form. Many professionals use Kelly as a framework to think about relative stake sizing (higher edge → larger stake) but apply it qualitatively rather than mathematically. Pure flat staking is also widespread among professionals who value simplicity and robustness over theoretical optimality.