Betting Education

Expected value (EV) in football betting explained

Expected value is the average profit or loss a bet would produce if the same situation repeated indefinitely. It is calculated from your probability estimate and the price, and it is the number that decides whether a bet is worth placing at all.

Probability curve rising across a dark chart, illustrating betting odds and implied probability

By BetBuddy Editorial Team · Editorial & research · Published · Last updated · 3 min read

The EV formula

Expected value weights the winning outcome by its probability and subtracts the losing outcome weighted by its probability. Everything hinges on the probability estimate; the price is simply given to you.

EV = (p × (odds − 1) × stake) − ((1 − p) × stake)

A worked €10 example

Take a €10 bet at 2.60 on an outcome you estimate at 42%. The winning branch pays €16.00 profit with probability 0.42; the losing branch costs €10.00 with probability 0.58.

EV = (0.42 × 16.00) − (0.58 × 10.00) = 6.72 − 5.80 = €0.92 per bet, or 9.2% of stake. That is the long-run average, not the outcome of any single match — this bet still loses 58% of the time.

EV per €10 staked at 2.60, by probability estimate
Your probabilityImplied by priceEV per €10
36%38.5%−€0.64
38.5%38.5%€0.00
42%38.5%+€0.92
46%38.5%+€1.96

EV, edge and yield are different numbers

Edge is EV expressed as a fraction of stake. Yield is realised profit divided by turnover once bets have settled. Return on investment, as BetBuddy reports it, is profit divided by the starting bankroll. Confusing them makes results look better or worse than they are.

  • EV: expected profit per bet, before anything happens.
  • Edge: EV ÷ stake, a percentage of what you risked.
  • Yield: realised profit ÷ total turnover, after settlement.
  • ROI: realised profit ÷ starting bankroll — the figure used across BetBuddy statistics.

Why positive EV is not a promise

EV assumes your probability estimate is correct. If the estimate is biased by two percentage points, a modest edge disappears entirely. Model error, stale prices and thin markets are the usual reasons a positive-EV bet on paper is a negative-EV bet in reality.

Betting carries financial risk and no staking plan removes it. Historical performance does not guarantee future results, and BetBuddy runs in paper mode: stakes are simulated and no bookmaker account is connected.

A worked expected-value calculation

Suppose the model estimates a 40% chance for a selection priced at 2.80. The fair price implied by 40% is 1 / 0.40 = 2.50, so the market is offering more than the model thinks the outcome is worth. Expected value per €1 staked is (0.40 × 1.80) − 0.60 = €0.12, an edge of 12%.

Now hold the model estimate constant and let the price drift to 2.40. Expected value becomes (0.40 × 1.40) − 0.60 = −€0.04. The same opinion about the same match has flipped from a strong bet to a losing one on a price move of four ticks. Edge is a property of the pairing of estimate and price, never of the selection alone.

This sensitivity is why BetBuddy re-prices every candidate close to kick-off and drops any bet whose edge has evaporated, rather than honouring a decision made hours earlier.

EV per €1 = (p × (odds − 1)) − (1 − p)

Why positive expected value still loses for long stretches

A 12% edge on a 40% shot is a distribution, not a promise. Six losses in a row from that selection profile happen roughly once in every twenty-two sequences of six bets — often enough that any real bettor meets it. Variance at these odds is large relative to the edge, which is why a sample of twenty settled bets says almost nothing about whether a method works.

The practical consequences are unglamorous: keep stakes small relative to the bankroll, judge the process rather than the last result, and treat closing-line value as the faster feedback signal while the profit-and-loss sample is still thin.

Related articles

About the author

BetBuddy's editorial team writes the education library and reviews every article against the production system it describes. Formulas are taken from the code that runs the staking engine, and any performance figure quoted comes from the tracked results ledger rather than from an example.

Editorial policy · Methodology

Back to all articles