The Kelly Criterion tells you how much of a bankroll to stake when you believe you have an edge. It is powerful, but full Kelly assumes your probability estimate is accurate. Football rarely gives you that luxury.
What Kelly is trying to do
Kelly aims to maximise long-term bankroll growth by staking more when the edge is larger and less when the edge is smaller. The danger is obvious: if your edge is overstated, Kelly also overstates the stake.
For decimal odds, the usual version is: `stake % = ((odds - 1) x probability - (1 - probability)) / (odds - 1)`. Half-Kelly simply takes the result and halves it.
Why half-Kelly is often better in practice
Half-Kelly accepts that your edge estimate might be noisy. A player may be rotated, a tactical setup may change, or the market may have information you have not priced in. Cutting the stake gives up some theoretical growth in exchange for smoother survival.
The Ressq view
Ressq separates selection from staking. RR can highlight a value gap, but the stake should reflect bankroll size, confidence in the estimate and personal risk tolerance. A smaller stake on a good edge is usually better than an oversized stake that changes your behaviour after one loss.
Kelly is not a responsible gambling tool on its own. Set hard limits, avoid borrowing to bet, and never increase staking because a formula says the next edge looks bigger.