The Kelly criterion for football bettors — explained without the math PhD

Most bettors stake the same amount every time. The Kelly criterion stakes by edge — and is the formula behind every serious bankroll-growth strategy.

The Kelly criterion for football bettors — explained without the math PhD

Most bettors stake the same amount on every bet. Some scale by "feel." Neither maximises growth. The Kelly criterion does — and it works on exactly two numbers you should already have: your edge and the odds.

BANKROLL GROWTH · 500 SIMULATED BETS · 5% EDGE 3x 2x 1x Flat 2u stake Half-Kelly (recommended) Full Kelly (volatile)

The formula

Kelly tells you what fraction of your bankroll to stake on a given bet:

f* = (bp − q) / b

Where:

  • f* — fraction of bankroll to stake
  • b — odds − 1 (the net payout per 1u staked, e.g. 1.10 for odds 2.10)
  • p — your model probability of the bet winning
  • q — 1 − p (probability of losing)

Why half-Kelly

Full Kelly maximises long-run growth — but the swings are brutal. A 5% edge bet at full Kelly stakes ~10% of your bankroll on one fixture; losing three in a row hurts. Half-Kelly captures ~75% of the growth with a fraction of the variance, which is why every serious practitioner runs at 0.25–0.5× the formula output.

What this means in practice

Suppose you have a €1,000 bankroll and you spot a bet at 2.30 that your model thinks has a 50% probability of winning.

  • b = 2.30 − 1 = 1.30
  • p = 0.50
  • q = 0.50
  • f* = (1.30 × 0.50 − 0.50) / 1.30 = 0.115 (i.e. 11.5%)
  • Half-Kelly → stake €57.50 (5.75% of bankroll)

Without an edge, Kelly returns zero or negative — which is its great virtue: it forbids betting on coin-flips, no matter how exciting they look.

The pitfalls

Kelly is unforgiving if your edge estimate is wrong. Overestimate p by even a few percent and you'll over-stake. This is why GSS uses the model probability from the same Bayesian framework that backtested the verified system — not gut feel.