Expected goals (xG) explained, and how to actually use it

Expected goals assigns every shot a probability of being scored, based on the characteristics of the chance: distance, angle, body part, the type of pass that created it, and the defensive pressure at the moment of the shot. A tap-in from six yards might carry an xG of 0.7 — seven times out of ten it goes in. A speculative effort from thirty yards might carry 0.03. Adding a team's shots gives their xG for a match. A side that produced 2.4 xG created chances that would, on average, produce roughly two and a half goals. If they scored one, they underperformed on the day. What this is really measuring: the quality and volume of chances a team creates and concedes. Over a single match that tells you little, because finishing is genuinely variable. Over fifteen or twenty matches it becomes one of the most useful public numbers in football, because chance creation is far more repeatable than finishing.

This is the core of the method. Compare a team's actual goals against their xG over a meaningful sample. A team scoring well above its xG has been finishing unusually well. That is sometimes skill — some strikers genuinely outperform models — but far more often it regresses. The league table flatters them, and the market prices the table. A team scoring well below its xG has been creating chances and missing them. Unless something structural is wrong, that tends to correct, and the market is often still pricing the poor results. The same applies defensively. A side conceding far fewer goals than their xG against suggests an inspired goalkeeper or a run of luck, neither of which is a plan. How to act on it: • Look for teams whose league position and underlying numbers disagree sharply • Bet the correction before the results correct, not after — once the table catches up, so has the price • Require a real sample. Ten matches is a starting point; three is noise

Goals markets are where xG applies most directly, because xG is literally a model of chance creation. The match winner market involves more that xG does not see. A practical filter: before backing a favourite, check whether their record is built on chance quality or on finishing. A team winning narrowly every week on 0.9 xG is a team about to lose.

The four things it misses: 1. Game state. A team leading 3-0 stops attacking. Their xG falls, and it means nothing about their quality 2. Individual finishing quality. Models treat shots as average. Elite finishers really do outperform, consistently 3. Goalkeeping. Standard xG says nothing about the keeper, and post-shot models that do are less widely available 4. Context the data cannot see — injuries mid-match, red cards, a manager change, a squad rotated for a cup tie Use xG as one input, not as an oracle. The bettors who lose money with it are the ones who treat it as a prediction rather than as a description of what has happened. A weekly routine: 1. Pull the xG table for your league and compare it with the actual table 2. List the three biggest overperformers and the three biggest underperformers 3. Check whether anything structural explains the gap — a new signing, an injury, a tactical change 4. Look at the coming fixtures for those six teams, and compare the price with what the underlying numbers suggest 5. Bet only where the gap is large. Small discrepancies are noise, and the margin eats them