FPL expected goals: how xG works and what it means
By Arasi Rex · Updated 11 August 2026 · 8 min read
FPL expected goals, or xG, is a number attached to every shot that says how likely it was to go in, and the whole stat is worth less than the hype around it. The model behind it rates a typical Premier League shot at 0.11 xG, a fast-break shot at 0.17 and an open-play shot at 0.12, so it is a measure of chance quality, not a forecast of FPL points. The claim worth arguing over is that expected goals is the most overrated number in the game: it describes chances that have already happened, and the gap between a player's xG and his actual goals says more about last season than the next Gameweek.
What expected goals actually is
Expected goals measures the quality of a chance by calculating the likelihood that it will be scored, using information on similar shots in the past. Opta's model draws on nearly one million shots from its historical database and rates every chance on a scale between zero and one, where zero is impossible and one is a goal you would expect every single time. A chance assigned 0.1 xG would, on average, be scored once from every ten shots in that situation, or 10 per cent of the time.
That is the whole idea, and it is deliberately simple. The metric was introduced in 2012 by Opta's Sam Green, and it spread through football because it solved a real problem: the shot count lies. Two teams can take the same number of shots and one can be creating almost nothing, because a shot from 25 yards and a one-on-one are not remotely equal. xG turns that difference into numbers.
The model itself is a machine-learning system called XGBoost, trained on the nearly one million shots mentioned above, taken from 40 competitions between 2018-19 and 2021-22. It evaluates more than 20 variables at the moment of the shot, including distance to the goal, angle to the goal, goalkeeper position, the clarity of the goal mouth, the pressure the shooter is under, shot type, and the pattern of play that produced the chance. The goalkeeper position feature is the clever part: it estimates the probability of a save, using the keeper's distance from the shot and his position relative to the line of sight.
How expected goals differs by shot type
The pattern of play variable is where the numbers get interesting for FPL managers. Shots do not all carry the same xG, and the difference is bigger than most people assume. Premier League data shows the average fast-break shot is worth 0.17 xG, against 0.12 for a shot from open play and 0.09 for one from a set piece, with the overall average of all shots at 0.11. A counter-attack shot is worth nearly half again as much as a routine open-play shot, which is why transition-heavy sides keep overperforming their shot counts.
That ordering matters for player selection, because FPL pays for goals, not for xG. The points table is blunt about it: a goal is worth 6 for a goalkeeper or defender, 5 for a midfielder, and 4 for a forward. The same 0.17 xG fast-break chance scores 6 points for a centre-back arriving at the far post and 4 for the striker who buried it, which is why the best-value xG in the game belongs to defenders who attack set pieces and transitions, not the forwards who take most of the shots.
What the player snapshot shows
The site database holds the completed 2025/26 FPL snapshot, and it is the cleanest illustration of why xG needs a health warning. Here is how last season's actual goals compare with each player's expected goals:
| Player | Position | Goals | xG | Difference |
|---|---|---|---|---|
| Haaland | Forward | 27 | 25.50 | +1.50 |
| B.Fernandes | Midfielder | 9 | 10.79 | -1.79 |
| Saka | Midfielder | 7 | 7.57 | -0.57 |
| Gibbs-White | Midfielder | 15 | 10.81 | +4.19 |
| Watkins | Forward | 16 | 15.40 | +0.60 |
| Gabriel | Defender | 3 | 2.94 | +0.06 |
The same number that says Haaland overperformed his xG by 1.50 also says Gibbs-White outscored his by 4.19 while B.Fernandes underperformed his by 1.79, and the three midfielders are on the same pitch. If xG were a prediction of goals, these gaps should be small and random. Instead they run to four goals in one direction, which is the tell: xG describes the chances a player got, not the finishing that turned some of them into goals.
Where expected goals misleads FPL managers
Here is the disagreement. The most popular FPL use of expected goals is to buy the player who is "due", meaning the one whose xG is well ahead of his goals, on the theory that the goals will catch up. That logic is backwards, because the model deliberately leaves out the one thing that decides the gap: the shooter. Distance, angle, goalkeeper position, pressure, none of these variables know who is taking the shot, so a striker who constantly beats his xG is not running hot, he is doing something the model cannot see. Gibbs-White's +4.19 does not mean he was lucky; it means his finishing was worth four goals the model could not price.
The mirror image is just as costly. Buying the midfielder whose xG is miles ahead of his points, on the theory that the regression is coming, ignores that expected assists work the same way. B.Fernandes delivered 24 assists in the snapshot against 12.28 expected assists, an overperformance of more than 11 that no xG-based model flagged in advance. The players who beat their numbers do it because they are good at the things the model cannot measure, and the players who lag them are usually not unlucky.
That is not to say xG is worthless in FPL. It is the best single measure of which teams and players create high-quality chances, and the fast-break numbers prove it: the sides that generate 0.17 xG chances on the break create better football than the sides posting 0.12 from open play. Use it to find the teams whose chance quality is rising, then look at who takes those chances. What you should not do is treat a personal xG deficit as a promise that points are coming, because over the 38-game season the gap mostly stays exactly where the model put it.
What the numbers do not tell you
Every player figure in this article comes from the completed 2025/26 snapshot, the only season with full rows in the site database. The 2026/27 season starts on Friday 21 August 2026, squads change, roles change, and last season's xG tells you nothing about a player's minutes this year. A 0.11 xG per shot is a league average, not a player's rate, and xG says nothing about fixtures, form, or whether the striker will even be on the pitch. It is a description of what already happened, not a forecast of what will, and the manager who remembers that is the one who uses it correctly.
FAQ
What is a good xG in FPL? There is no single good number. The league-average shot is worth 0.11 xG, a fast-break shot 0.17, and an open-play shot 0.12, so what matters is volume and shot type: the players worth owning are the ones taking many high-xG chances, not the ones with one lucky number.
Is a player who outscores his xG lucky? Not necessarily. The model does not include finishing ability, so a player who consistently beats his xG, like Gibbs-White's +4.19 in the snapshot, may simply be a better finisher than the model assumes. The gap is not automatically luck.
Does expected goals help with captaincy? Indirectly. Captain scores double, so the armband belongs to the player with the best combination of chance volume and shot quality, which is where xG helps. But xG cannot see minutes, fixtures, or form, so it is one input, not the answer.
What is expected assists, or xA? The same idea applied to passes: the likelihood a pass becomes an assist. B.Fernandes delivered 24 assists against 12.28 xA in the snapshot, which shows assists can overperform expectations by a wide margin, and the same caution applies.
Key takeaways
- Expected goals measures the likelihood a shot is scored, on a 0 to 1 scale, using nearly one million historical shots from 40 competitions between 2018-19 and 2021-22.
- Shot types are not equal: a fast-break shot averages 0.17 xG, an open-play shot 0.12, a set-piece shot 0.09, and all shots 0.11.
- The model's variables include distance, angle, goalkeeper position, pressure and pattern of play, but not the identity of the shooter, so finishing skill is invisible to it.
- The 2025/26 snapshot shows gaps in both directions: Gibbs-White +4.19, Haaland +1.50, B.Fernandes -1.79, so personal xG is not a reliable forecast.
- Use xG to find teams creating high-quality chances, not to buy players who are "due", and treat every figure as last season's description, not next week's prediction.
Expected goals is a tool for finding chance quality, and it is a good one, but it is not a prophecy. The fastest-break sides create the best shots, the scoring guide sets what each goal is actually worth, and the differentials guide explains why low-ownership picks built on real chance volume beat template players who are merely popular. When a striker's xG is running hot, the honest move is to check the fixture list and the captaincy guide, because the next Gameweek is decided by who plays and who scores, not by a number describing shots that are already in the past. The counter-attacking football guide is where the 0.17 xG fast-break game is explained in full, and the FPL ranks guide is where all of it shows up in your rank. Read the xG, respect it, and never trust it to tell you what happens next.