Expected value answers one question: if you could place this exact bet a million times, what would the average outcome be? Multiply the probability of winning by the amount you win, subtract the probability of losing times your stake, and you have the EV.
A concrete example: a bookmaker offers 2.20 on an outcome whose true probability is 50%. Half the time you win 1.20 units, half the time you lose 1 unit. EV = 0.5 × 1.20 − 0.5 × 1.00 = +0.10 — ten cents of theoretical profit per unit staked. The same bet at 1.90 has an EV of −0.05: a guaranteed long-run loss dressed up as a coin flip.
Notice what EV does not care about: whether this particular bet wins tonight. A positive-EV bet that loses was still right; a negative-EV bet that wins was still wrong. Results arrive with brutal variance — EV is the signal underneath.
The hard part is the probability. Bookmakers publish the payout, but nobody hands you the truth about how often the outcome hits. Every serious approach to betting is really a fight over estimating that number — which is why we anchor ours to the devigged sharp market, the most accurate public estimate that exists, and flag value only when a soft book strays above it.
See how we apply this on every match in our methodology. Methodology →
EV = (fair probability × decimal odds) − 1. If the fair probability is 40% and the price is 2.70, EV = 0.40 × 2.70 − 1 = +8%. The hard part is the fair probability, which is why we derive it from the devigged sharp market rather than from a model.
Constantly. EV describes the average over many repetitions; any single bet is decided by the result. A +5% EV bet at odds of 3.00 still loses two times in three. Judge the process by EV and closing line value, never by the outcome of one bet or one week.
See these ideas at work in our public track record — every call logged before kick-off, losses included.