Expected Goals in the Premier League: Where to Get xG in 2026 and How to Read It

Premier League expected goals in 2026 come mainly from Opta, through licensed outlets such as Opta Analyst, Fantasy Football Scout, Squawka, and the FotMob app. FBref stopped carrying xG on 20 January 2026. Read team tables by xG difference per match, player tables by non-penalty xG per 90, and expect large gaps to regress.

What changed in 2026

For several seasons the default free source for Premier League xG was FBref, which carried Opta's model. On 20 January 2026, Sports Reference announced that FBref's advanced statistics, including xG, progressive passes, and shot-creating actions, were removed effective immediately. Stats Perform, Opta's owner, had terminated access and demanded that the data be deleted, asserting that Sports Reference's use breached their agreement, which Sports Reference denies; Sports Reference said that contesting it would have been costly and uncertain. The removal covered every competition on the site, men's and women's, not only the Premier League. FBref now works as a historical database for results and basic stats, not as a current xG source. Any guide, notebook, or scraper still pointing at FBref xG tables is reading frozen data.

Where Premier League xG is published now

Sources verified on 2026-11-05. Coverage and terms change, so check each one before building on it.

Opta, through licensed outlets. Opta's model remains the most widely syndicated. Opta Analyst, Stats Perform's own editorial site, runs xG features such as its xG Files series, covering xG, non-penalty xG, open-play and set-piece xG, xGOT, and expected points. Fantasy Football Scout publishes Opta-credited xG and xGC tables for 2026/27, and Squawka's 2026/27 xG league table states that its values come from Opta, though when checked it carried a last-updated date of 25 August 2026, so confirm it is current before building on it.

Match apps. FotMob shows live and post-match xG with shot maps for Premier League games, and credits Opta for the match statistics behind its numbers, so it is another Opta outlet rather than a separate model. Planet Football built its 2026/27 xG tables from FotMob data. Sofascore and WhoScored also show match-level stats, and WhoScored is likewise Opta-based. Apps are convenient for following games but are not built for bulk export, and not all of them name the underlying model.

Independent models. Understat has run its own xG model for the top European leagues, including the Premier League, with shot-level data back to 2014/15. When checked on 2026-11-05 the site displayed only a maintenance message, so its current availability is uncertain.

Paid feeds. Stats Perform (Opta) and Hudl StatsBomb sell event data and xG to clubs, media, and betting operators. StatsBomb's free open data covers selected competitions and past seasons, which is useful for building a model but not for tracking the current Premier League season.

Why two sites show different xG for the same match

Each provider trains its own model on its own shot data, with different features. Some include defender and goalkeeper positions, some only shot location, body part, and assist type. Penalties may be a fixed value or excluded. The result is that one match can carry noticeably different team xG on two sites, and a season total can differ by several goals.

The practical rule is one source per analysis. Never subtract an Understat figure from an Opta xG table, and never compare a player's numbers across sites. Two outlets carrying the same provider are a different case: a gap between them is snapshot timing or rounding, not a disagreement between models. If a series has to span the FBref break, keep historical Opta values from FBref and continue with another Opta-credited outlet, and flag the join.

How to read a team xG table

A typical Premier League xG table carries these columns:

  • xG (or xGF): total expected goals from the team's shots.
  • xGA (or xGC): expected goals conceded.
  • xGD: xG minus xGA. The single best summary of underlying performance.
  • npxG: non-penalty xG, which removes penalty variance.
  • xPts: expected points, usually from simulating each match many times using shot xG values.
  • Goals minus xG: over- or underperformance in finishing, or in conceding for defences.

Read xGD per match, not totals, because teams will have played different numbers of games. Separate open play from set pieces where the source allows, since set-piece xG depends heavily on routines and is less stable. Treat xPts as a rough guide: it ignores game state, and a team leading early often concedes territory, which lowers its xGD without being a sign of weakness.

How to read a player xG table

For players, the useful columns are npxG per 90, xA or xAG per 90, shots per 90, and xG per shot, which separates chance volume from chance quality. Set a minutes threshold before ranking anyone. Per-90 rates from a few hundred minutes swing wildly, and a substitute with one big chance can top a table built on tiny samples.

Goals minus xG for a single player is the noisiest number on the page. Over one season most of the gap is variance. Only over several seasons of shots does a consistent difference start to suggest real finishing skill, and even then the gap is usually smaller than early numbers suggest.

Regression to the mean

Goals in football are rare, so any short run of results contains a lot of noise. Over a season, team xG difference tends to predict future results better than goal difference, because chance creation is more repeatable than conversion. Teams whose points far exceed what their xG implies tend to fall back, and teams underperforming their xG tend to improve.

A simple way to apply this is shrinkage. Illustrative example, not real data: a team has scored 10 more goals than its xG after 12 matches. Rather than treating all 10 as skill, assume most of the gap will not persist and project future scoring mainly from xG, with only a small adjustment toward the observed finishing. The size of the adjustment should grow with the number of shots, not with how impressive the run looks.

Early-season tables deserve extra caution. After five gameweeks, a single penalty, a red card, or one match against a much stronger opponent can move a team's xGD substantially. Opta Analyst's own early-season xG features carry the same warning about small samples.

Using Premier League xG in a model

For forecasting, xG is an input, not an answer. Typical uses are rolling xG for and against as team strength features, npxG to reduce penalty noise, and shrunk finishing adjustments for players. Betting markets already incorporate xG and squad news, so the test of whether a model adds anything is whether it improves on market-implied probabilities out of sample, not whether it fits last season.

This page describes data sources and method and is not betting advice.

Frequently asked questions

Where can I find Premier League xG stats in 2026?
Opta xG is published through licensed outlets such as Opta Analyst, Fantasy Football Scout, and Squawka. FotMob shows live and post-match xG in its app. Understat has historically run its own model but showed a maintenance page on 5 November 2026. FBref no longer carries xG after losing Opta data on 20 January 2026.
Does FBref still have xG?
No. On 20 January 2026, Sports Reference removed FBref's advanced statistics, including xG, after Stats Perform terminated its data access over an alleged breach of agreement that Sports Reference denies. FBref still carries results and basic statistics, so it remains useful as a historical database, but it is no longer a current source for Premier League expected goals or related advanced metrics.
What is a good xG difference in the Premier League?
There is no fixed threshold, because xG levels vary by provider and season. The useful reading is relative: compare a team's xG difference per match with the rest of the league in the same source. A positive figure means the team creates better chances than it concedes, and a sustained positive figure usually goes with finishing high in the table.
Why does my team's xG differ between two sites?
If the two sites use different providers, each model is trained on its own shot data with different inputs, such as whether defender and goalkeeper positions are included, and penalties may be valued differently, so the same shots receive different probabilities. Neither figure is wrong; they are estimates from different models. If both sites carry the same provider, as FotMob and Opta Analyst both carry Opta, a gap is almost always snapshot timing or rounding instead. Either way, compare teams within a single source.