What Is the Kelly Criterion?
The Kelly Criterion is a mathematical formula that determines the optimal percentage of your bankroll to stake on a bet, based on the perceived edge you have over the bookmaker.
Developed by John L. Kelly Jr. at Bell Labs in 1956, it was originally designed for information theory but quickly found applications in gambling and investment.
When applied correctly, Kelly maximizes the long-term growth rate of your bankroll while minimizing the risk of ruin. It's the mathematically proven optimal staking strategy.
The Kelly Formula
The basic Kelly Criterion formula is:
f = (bp - q) / b*
Where:
- f* = fraction of bankroll to stake
- b = decimal odds minus 1 (i.e., the net odds)
- p = your estimated probability of winning
- q = probability of losing (1 - p)
Worked Example
You find a bet at decimal odds of 2.50 and estimate the true probability of winning at 45%.
- b = 2.50 - 1 = 1.50
- p = 0.45
- q = 0.55
f = (1.50 × 0.45 - 0.55) / 1.50* f = (0.675 - 0.55) / 1.50* f = 0.125 / 1.50* f = 0.0833 or 8.33% of your bankroll*
This means you should stake 8.33% of your current bankroll on this bet for optimal long-term growth.
Why Kelly Works
Maximizes Long-Term Growth
Kelly is proven to produce the fastest bankroll growth over time. No other staking strategy grows your bankroll faster while maintaining mathematical rigor.
Automatically Adjusts for Edge Size
The bigger your edge, the bigger the Kelly stake. If you have a small edge (say 2%), Kelly recommends a small stake. If you have a huge edge (say 20%), Kelly recommends a larger stake. This natural scaling is powerful.
Prevents Overbetting
Kelly will never recommend staking more than your edge warrants. If you have no edge (the bet is fair value or negative EV), the formula returns zero or negative — telling you not to bet.
Adapts to Bankroll Size
Because Kelly stakes a percentage of your current bankroll, your absolute stake naturally:
- Increases when you're winning (compounding gains)
- Decreases when you're losing (protecting your bankroll)
The Problem with Full Kelly
While mathematically optimal, full Kelly is too aggressive for most bettors:
- Variance is enormous — Full Kelly produces wild bankroll swings. A 50% drawdown is common even with a genuine edge.
- Probability estimates are imperfect — Kelly assumes you know the true probability. In reality, your estimates have uncertainty.
- Psychological toll — Watching your bankroll drop 40-50% (even temporarily) is mentally devastating.
Fractional Kelly — The Practical Solution
Most professional bettors use fractional Kelly, typically staking 25-50% of the full Kelly recommendation.
Quarter Kelly (25%)
Using our earlier example (full Kelly = 8.33%):
- Quarter Kelly stake = 8.33% × 0.25 = 2.08%
Half Kelly (50%)
- Half Kelly stake = 8.33% × 0.50 = 4.17%
Why Fractional Kelly Is Better in Practice
| Metric | Full Kelly | Half Kelly | Quarter Kelly |
|---|---|---|---|
| Growth rate | Maximum | ~75% of max | ~56% of max |
| Max drawdown | ~50-60% | ~25-35% | ~15-20% |
| Risk of ruin | Low but possible | Very low | Near zero |
| Emotional stress | Very high | Moderate | Low |
The trade-off: You sacrifice some growth rate for dramatically smoother bankroll progression. For most people, this is a worthwhile trade.
Step-by-Step Kelly Implementation
Step 1: Estimate the True Probability
This is the hardest part. Sources for probability estimation:
- Your own analysis — xG data, form, team news
- ScoreSage AI predictions — our AI model provides probability estimates for each match
- Odds comparison — averaging odds across 10+ bookmakers gives a reasonable market probability
- Historical data — base rates for similar situations
Step 2: Calculate the Edge
Edge = (True probability × Decimal odds) - 1
If Edge > 0, you have a value bet. If Edge ≤ 0, don't bet.
Step 3: Apply the Kelly Formula
Use the formula above with your estimated probability and the available odds.
Step 4: Apply Your Kelly Fraction
Multiply the full Kelly result by your chosen fraction (0.25 to 0.50 recommended).
Step 5: Place the Bet
Stake the calculated percentage of your current bankroll (not your starting bankroll — Kelly adjusts dynamically).
Kelly Criterion Limitations
1. Garbage In, Garbage Out
Kelly is only as good as your probability estimates. If you consistently overestimate your edge, Kelly will lead you to stake too much and lose money faster.
2. Simultaneous Bets
The basic Kelly formula assumes sequential bets. When placing multiple bets simultaneously (e.g., a Saturday afternoon with 10 matches), you need to adjust. A simple rule: divide your Kelly stake by the number of simultaneous bets.
3. It Doesn't Work for Accumulators
Kelly is designed for single bets. For accumulators, the maths becomes extremely complex. Stick to flat staking or unit-based systems for accas.
Kelly vs Other Staking Methods
| Method | Pros | Cons |
|---|---|---|
| Kelly | Mathematically optimal; scales with edge | Requires probability estimates; volatile |
| Flat staking | Simple; consistent | Doesn't account for edge size |
| Percentage staking | Adjusts to bankroll size | Doesn't account for edge |
| Fibonacci | Structured recovery system | Can lead to huge stakes after losses |
The Kelly Criterion is the most powerful staking tool available, but it requires honest probability assessment. Use ScoreSage AI's confidence ratings alongside your own analysis to estimate probabilities, then let Kelly tell you how much to stake.