What each measure adds
| Measure | Captures | Misses |
|---|---|---|
| Shots | Attempt volume | Location, placement, defensive pressure |
| Shots on target | Attempts requiring a save or scoring | Off-target chance quality and shot placement detail |
| xG | Pre-shot scoring estimate | Actual placement after contact in many models |
| xGOT | Post-shot on-target estimate | Off-target and blocked attempts by design |
Opta's definitions provide one provider vocabulary. Its xGOT guide distinguishes pre-shot xG from an on-target model that incorporates placement. Other providers can classify blocks and targets differently.
Worked comparison
Team A records 12 shots totalling 0.90 xG; Team B records 7 shots totalling 1.35 xG. Average xG per shot is 0.90 / 12 = 0.075 for A and 1.35 / 7 = 0.193 for B. A had greater volume, while B's recorded shots had greater average modelled quality. That does not establish which team will create the same chances next match.
A repeatable match review
- Confirm one provider and match period.
- Separate penalties and set pieces where the question requires it.
- Plot shot location and time rather than relying on totals alone.
- Record score state when attempts occurred.
- Compare a rolling sample with opponent strength and home-away context.
StatsBomb's open data exposes shot locations, outcomes and xG fields for selected matches, making this workflow reproducible on a known dataset.
Why shots on target are not useless
The count answers a narrower question: how many attempts met the provider's on-target definition. It can support goalkeeper or post-shot analysis when paired with placement and quality. The error is treating it as a complete attacking measure or assuming a universal conversion rate.
A compact diagnostic sequence
Read shot measures from broad opportunity to chance quality:
- Compare total shots to understand volume.
- Compare shots from dangerous locations or xG to understand quality.
- Inspect shots on target as an observed placement outcome, while remembering that blocked shots and woodwork may be classified differently.
- Split open play, set pieces and penalties where possible.
- Add game state and minutes at each score so late pressure is not confused with stable dominance.
| Pattern | Plausible interpretation | What to inspect next |
|---|---|---|
| High shots, low xG | Many attempts from low-value situations | Distance, angle, blocks and game state |
| Low shots, high xG | Few but strong opportunities | Largest chances and repeatability |
| High xG, low shots on target | Outcome variation or shot execution | Shot placement and goalkeeper actions |
| Low xG, many goals | Finishing, goalkeeping or variance | Longer sample and shot locations |
Keep the denominator visible
Per-match rates can mislead when teams play different minutes after red cards or spend different amounts of time leading. For team comparisons, report match count and minutes; for player comparisons, report minutes and shot count. For prediction, calculate every rolling feature with data available at the cutoff and evaluate it against a simpler baseline on later matches.
Related resources
Read xG explained for probability interpretation and player performance metrics for role-aware comparisons.
Continue learning
- Next guide: Expected Threat (xT)
- Related guide: Fixture Congestion in Football
Assumptions and limitations
The example is illustrative. Event correction, blocked-shot rules, provider coverage and score state affect comparisons. Descriptive shot metrics do not become calibrated future probabilities without a separately evaluated forecasting method.
GambleAware
GamCare
Gambling Therapy
GAMSTOP