Why expected goals is not a betting system
xG tells you how good a team's chances were. It does not tell you what the next match will do, and it says nothing about the price a bookmaker is offering. Here is what the number can and cannot carry.
Expected goals has gone from the back room of analytics departments to the graphics on a Saturday night highlights show. Every shot gets a number between 0 and 1, every match gets a total, and a team that "won the xG" is said to have deserved more than the scoreline gave them. Used well, it is one of the most useful tools in football analysis. Used as a forecast for the next ninety minutes, or as a reason to back a team at a price, it is much weaker than it looks.
What xG actually measures
An expected goals model looks at a shot and asks a narrow question: of all the similar shots in its training data, how many were scored? Distance and angle to goal do most of the work. Most models then add the body part, whether the chance came from a cross, a through ball or a set piece, whether it was a counter-attack, and in the better models where the defenders and the goalkeeper were standing.
The result is a probability for the chance, not for the finish. A tap-in from four yards might be worth 0.8, a shot from thirty yards 0.03. Add up a team's shots and you get the number of goals an average finisher would have been expected to score from them.
That makes xG a description of process. Over a run of matches it is a better guide to how well a team creates and concedes chances than goals alone, because goals are rare and noisy. A side scoring well above its xG for half a season is usually finishing at a rate it will not sustain; a side well below it is often better than its position.
Why it struggles with a single match
The same noise that xG smooths out over a season is exactly what decides a single game. A team can produce 2.1 expected goals and score none; another can score twice from 0.6. Neither result is a mistake by the model. A probability of 0.3 on a chance means it will be missed seven times in ten.
Three further limits matter for anyone reading a match total as a prediction:
- Game state changes behaviour. A team leading 2-0 sits deeper and concedes more low-quality shots; the trailing side's xG climbs without its chances of winning rising much.
- Small samples mislead. Ten matches of xG can still be distorted by one penalty, one red card or one chaotic second half.
- Models differ. Providers weight the same shot differently, so "the xG" for a match can vary noticeably depending on whose model produced it.
A season of xG says something real about a team. One match of xG mostly describes that match.
Why it is not an edge on the odds
The step from "this team's xG is better than its results" to "this team is good value" sounds small. It is not. Bookmakers and the betting markets behind them use expected goals, and considerably richer data, to set their prices. By the time a team's underlying numbers are visible on a public graphic, they are already in the odds.
Beating a price needs information or judgement the market does not already have, and a widely published metric is the opposite of that. Add the bookmaker's margin, which is built into every price, and a reader who bets on public xG alone should expect to lose over time, not win.
If you do bet on football, the practical questions are less glamorous than any model: who you are betting with, what their terms say, and how much you can afford to lose. For readers who use bookmakers licensed outside Great Britain, Daily All In, a sister title, keeps a shortlist of operators it has checked. Those operators sit outside the British system, including the GamStop self-exclusion scheme, so check their licence and withdrawal terms yourself before depositing.
How to read xG well
Treat it as a lens on how a team plays, not a verdict on what it deserves or a tip for what comes next. Look at it across ten matches or more, alongside who the opponents were and how the games unfolded. Use the gap between goals and xG as a question to investigate, not an answer. And remember that the best teams in the world regularly lose matches they "won" on expected goals; that is not a flaw in the number, it is football.
18+ only. Gambling carries a real risk of losing money. If it stops being fun, free and confidential help is available from BeGambleAware, GamCare or the National Gambling Helpline on 0808 8020 133.