Where to Find Expected Goals (xG) Stats

Several free websites publish expected goals for major leagues. FBref offers extensive team and player tables, including xG, non-penalty xG, and expected assists. Understat covers a set of top European leagues with shot maps and per-match xG. These sites are the most common starting point for fans and analysts.

Many popular football apps display xG for each match alongside traditional stats, sometimes with shot maps. They are convenient for following games but usually not designed for bulk data export or historical analysis.

Some data providers release free event data for selected competitions and seasons, intended for education and research. These datasets include shot locations and context, allowing you to compute your own xG and test models. Check licence terms and attribution requirements.

Commercial providers offer comprehensive event and tracking data across many leagues, including proprietary xG models, freeze-frame data showing player positions at the time of shots, and APIs for automated access. Clubs, media, and betting companies use these services; costs reflect the depth of coverage.

Each provider trains its own model with different inputs and data. One may include defender positions, another only shot location and type. Penalty values and treatment of own goals can differ. As a result, the same match can show noticeably different xG values across sources.

Pick one source and stick with it for a given analysis. Check which leagues and seasons are covered, whether penalties are separated, how frequently data updates, and whether you can access it programmatically. Respect each site's terms, especially around scraping and redistribution.

Useful team stats include xG for, xG against, xG difference, and non-penalty versions, plus xG per shot to separate chance volume from quality. Useful player stats include non-penalty xG per 90 minutes, expected assists per 90, and goals minus xG over large samples.

Normalise player stats per 90 minutes so that substitutes and starters can be compared, but set a minimum minutes threshold, since per-90 numbers from small samples are unstable.

If you want full control, compute xG from event data using your own model. This lets you choose features, handle edge cases consistently, and understand exactly what the numbers mean. It also helps when public sources do not cover the competition you care about.

For forecasting, xG is often combined with betting market data. Market prices summarise many inputs, including squad news, and implied goal expectations from odds can be compared with xG trends. Normalised odds histories make it easier to test whether xG adds information beyond the market.

Mixing providers in one comparison, judging players on a handful of matches, ignoring penalties, and treating xG as a complete measure of performance. xG measures chance quality, not everything that happens on the pitch.

Coverage of women's competitions and smaller leagues has improved but remains thinner than for top men's leagues. Free research datasets sometimes include women's tournaments, and some public sites cover major women's leagues. For lower divisions, building your own xG from event data may be the only option.

Shot maps plot every shot by location, sized by xG and coloured by outcome. They make patterns visible quickly, such as a team relying on long-range shots or a striker getting into prime positions. Many public sites provide shot maps, and plotting libraries make them easy to build from event data.

Spot-check a few matches against video or official reports: missing shots, wrong body parts, or misplaced coordinates all distort xG. Watch for changes in a provider's model between seasons, which can shift values without any change in performance.

Expected assists, or xA, credits passers with the xG of shots they set up, and is often more stable than actual assists, which depend on teammates finishing. Combining xA with key passes and progressive passes gives a fuller picture of creative players than goals and assists alone.

Use public sites for quick research, free event data to build your own model, and paid providers when you need coverage or depth. Stay consistent within one source, and note its model version carefully.

This page describes where the data comes from and is not betting advice.