Bankroll management: the part nobody looks at
Two people with the same model and different stakes do not get the same result. Often, the gap comes from there and not from the model.
By The PRIZIFOOT team
Bet selection decides the expectation. Stake size decides the probability of lasting the distance. These are two separate problems, and the second is regularly ignored.
The flat stake
Staking a constant percentage of the starting bankroll (1 to 2% is a common order of magnitude) is the simplest method and the hardest to apply badly. It mechanically bounds the damage of a bad run.
The Kelly criterion, and its compulsory caution
Kelly suggests staking a fraction f = (p × c − 1) ÷ (c − 1) of the bankroll, where p is the estimated probability and c the price. That formula maximises long-run growth, but only if p is right.
And p is an estimate. Overstating its accuracy leads Kelly to recommend enormous stakes, and the formula meant to grow the bankroll destroys it. That is why common practice is “half-Kelly” or less: you accept slower growth in exchange for far less sensitivity to your own errors.
The useful reminder
A long losing run is not proof that a method is bad, and a winning run is not proof of the opposite. On short samples, both happen to everyone. The only judge is the number of bets, and it is larger than one hopes.
Move from theory to the match
Pick two teams and see what the model makes of it, without creating an account.
Analyse a match