EV is the single number that tells you whether a bet is mathematically sound — independent of whether it actually wins.
Where P(Win) is your estimated true probability of the outcome occurring (not the bookmaker's implied probability).
Decimal odds of 3.00 (implied 33.3%) on a $50 stake. You believe the true probability is 40%.
Positive EV means this bet, repeated many times under identical conditions, profits on average — even though it will still lose 60% of the time.
Same odds (3.00), but your true probability estimate is only 28% — below the bookmaker's 33.3% implied probability.
This is a losing proposition in the long run, even if it happens to win this time.
EV calculations are only as good as your probability estimate. If your estimate of 40% in the first example was actually wrong — and the true probability was really 30% — you'd have a negative EV bet that merely looked profitable on paper. This is why accurate probability modelling (not the EV formula itself) is the hard part of profitable betting.
A single +EV bet can still lose. A single −EV bet can still win. EV only describes the average outcome across a large number of repetitions. This is precisely why disciplined bankroll management matters — surviving the variance long enough for a real edge to play out requires careful stake sizing, not just identifying +EV opportunities.
Strip out the bookmaker's vig first to find the true fair price — your baseline for any EV calculation.