A value bet exists when a bookmaker's odds imply a lower probability than the outcome's real chance of happening. If a team's true chance of winning is 50% but you can back it at odds of 2.20, the price is 'wrong' in your favour: over many repeats of that situation you would win more than the odds assume, so the bet has positive expected value.
Value has nothing to do with how likely the bet is to win. Backing a 5.00 outsider can be a value bet while backing a 1.20 favourite can be terrible value — what matters is the gap between price and probability, not the probability itself. That is why disciplined bettors hunt prices, not winners. Finding value systematically requires an independent estimate of the true probability, which is exactly what odds models and screeners attempt to produce.
Suppose the fair chance of a home win is 40% (fair odds 2.50), but one bookmaker offers 2.75. Expected value per £1 staked = (0.40 × 2.75) − 1 = +0.10, i.e. a 10% edge. Individually the bet loses 60% of the time — but at scale, prices like this are where profit comes from.