Most bettors obsess over finding winners. They chase tipsters with the highest advertised win rate and ignore the single most important factor in long-term profitability: how much you bet. The Kelly Criterion, developed by John Kelly at Bell Labs in 1956, provides a mathematical answer to this question, and it changes everything about how you should think about betting.
The Problem with Flat Stakes
Flat-unit betting means wagering the same amount on every bet regardless of your edge. If you bet 1 unit on a coin flip at even money, you expect to break even. If the coin is biased to 55% heads, flat betting still works, but it is far from optimal.
The core issue is that flat stakes treat a bet where you have a 2% edge the same as a bet where you have a 15% edge. You are under-betting your strongest plays and over-betting your weakest ones. Over hundreds of bets, this costs you significant compound returns.
The Kelly Formula
The Kelly Criterion tells you to bet a fraction of your bankroll equal to your edge divided by the odds:
Where f* is the fraction of bankroll to wager, b is the decimal odds minus 1, p is your estimated probability of winning, and q is the probability of losing (1 - p).
For example, if you estimate a team has a 60% chance of winning and the bookmaker offers 2.00 (even money), the Kelly stake is: (1 x 0.60 - 0.40) / 1 = 20% of bankroll.
Why Full Kelly Is Too Aggressive
Full Kelly maximizes the long-term growth rate of your bankroll, but it comes with brutal variance. A few bad beats can cut your bankroll in half. In practice, most professional bettors and funds use fractional Kelly, typically betting between 25% and 50% of the full Kelly amount. This sacrifices a small amount of theoretical growth in exchange for much smoother equity curves.
Our 5-Tier System
At CalibrSports, we use a 5-tier fractional Kelly system that maps model confidence to Kelly fractions:
- Probability 90%+: Kelly fraction = 0.50 (half Kelly)
- Probability 80-89%: Kelly fraction = 0.30
- Probability 70-79%: Kelly fraction = 0.18
- Probability 60-69%: Kelly fraction = 0.12
- Probability 50-59%: Kelly fraction = 0.08
This tiered approach ensures that our highest-conviction bets, where the model sees a large edge, receive meaningfully larger stakes than marginal plays. A bet with 92% estimated probability might receive a 4-5% bankroll allocation, while a 52% edge bet might receive 0.5%. The daily cap is 50% of bankroll exposure, and no single bet can exceed 10%.
Why 55% WR + Kelly Beats 60% WR + Flat
Consider two bettors over 500 bets at average odds of 2.00:
- Bettor A: 60% win rate, flat 2% stakes. Expected profit: 500 x 0.02 x (0.60 x 2.00 - 1) = +2.0 units.
- Bettor B: 55% win rate, Kelly-sized stakes averaging 3.5% on strong bets and 0.8% on weak ones. The compound growth from larger bets on higher-edge plays generates +3.8 units despite the lower hit rate.
The math is clear: optimal sizing on smaller edges beats flat sizing on larger edges. This is counterintuitive, which is why most recreational bettors never adopt it.
Risk Management
Kelly sizing also provides a natural risk management framework. When edges are thin, stakes are small. When the model is uncertain, exposure is minimal. You never blow up because the system self-regulates. Compare this to a flat-stake bettor who risks the same amount on a shaky 51% play as on a rock-solid 85% play.
If you want to see Kelly sizing in action across real bets, visit our performance page where every bet, stake, and result is published transparently.