Goals are the first number you see and the least reliable. They do not say how long the player was on the pitch, or how good the chances were. Here are three things to check first.
Per 90 minutes and sample size
Two players do not play the same minutes. So we divide by minutes and multiply by 90. That lets you compare per match, not per spell on the bench. Sample size still matters: 15 matches’ worth of minutes is small and the number moves a lot.
xG and xA
xGi (expected goals) gives each shot a probability of scoring, based on position and type. The sum shows how good the chances were. xA does the same for passes that led to shots. Chances are steadier than goals.
| Measure | Striker A | Striker B |
|---|---|---|
| Goals | 12 | 12 |
| Minutes | 2700 | 1350 |
| 90-minute matches | 30 | 15 |
| xG | 8.4 | 10.5 |
| Goals per 90 | 0.40 | 0.80 |
| xG per 90 | 0.28 | 0.70 |
| Goals minus xG | +3.6 | +1.5 |
Overperformance and regression to the mean
Someone who scores far above xG usually does not keep it up. Part of it was luck. Future goals tend to move toward xG. It is not a rule for every player: some do finish better than average. You can only tell with many seasons.
How to do it in practice
- Find the minutes first. Under about 900 minutes (10 matches of 90) do not trust any rate.
- Put goals per 90 and xG per 90 side by side. If they are far apart, suspect luck.
- Look at the last two seasons, not only the current one.
- Check whether the minutes came off the bench. Substitutes often play in open games with more space.
In the example, B looks better at first glance, and he may be. But 15 matches is small and xG is not an infallible model. We use it as a signal, not as proof.