Understand the Measures

It is an event total multiplied by 90 and divided by minutes played. It standardizes playing time, not role, opposition or opportunity.

For example, four goals in 600 minutes is 0.60 goals per 90. Show the minutes alongside the rate. Read the player comparison guide.

There is no universal minimum. Stability depends on the event, its frequency and the question. Publish the sample and avoid treating a threshold as proof of reliability.

Rare events usually provide less evidence than frequent actions over the same period. A larger sample can still be biased if the role or opposition differs systematically.

Expected goals, or xG, assigns a modelled scoring probability to an attempt. Summed xG describes chance quality under that model, not a guaranteed score.

Different providers can use different features. Name the model or provider and avoid treating a short-term goals-minus-xG difference as a settled measure of skill.

Not necessarily. Possession depends on score state, location and tactical choices. Examine what the team did with the ball and what it allowed without it.

A leading team may accept less possession. Use a match timeline to separate those phases.

Rates align exposure but do not equalize competition strength, schedule or scoring environment. Use matching definitions and describe the remaining differences.

League-relative measures add context but do not establish how a team would perform in another competition. Review normalization choices.

A percentile locates a value within a stated comparison group. It is not a percentage of perfect performance, and it changes when the group changes.

Name the role, competition, period and minutes threshold used to form that group. A high percentile is not automatically desirable for every metric.

No. Zero means the event was observed not to occur; missing means its value is unavailable. State how unavailable records affect totals and rates.

Dropping missing matches can change the sample in a systematic way. Report coverage rather than quietly treating the available records as a complete season.

They may use different event definitions, collection methods, model features or revisions. Check documentation before combining their figures.

Preserve source names and retrieval dates in your own working notes. A corrected event record may legitimately differ from an earlier report.

No. This is a static editorial resource. Historical snapshots identify their period, and invented examples are labelled.

We do not offer live odds, automated scouting rankings or a public discussion feed. External source pages may change independently of this site.

Use the contact tool to draft a question containing the page, competition, season, source and disputed calculation. The tool does not send messages automatically.

Include your proposed correction and the steps needed to reproduce it. Prepare a data question.