Non-Penalty xG and Why It Is Preferred — explained

Penalties are the most valuable and least informative chance in football. Stripping them out produces the non-penalty variant, which is the version most analysts now quote by default for a reason that is worth spelling out.

The penalty is a different event

A penalty carries a value near three-quarters of a goal and is almost independent of the quality of the attacking move that produced it. Whether a team earns penalties is a real and repeatable skill of sorts, but it is a different skill from the ability to create chances from open play, and folding the two together makes a blended number that describes neither.

Removing penalties therefore produces a cleaner description of open-play creation. It also removes the single largest source of match-to-match volatility in a team's expected-goals total, because a single incident can add most of a goal to a figure that would otherwise sit near one.

What changes when you subtract it

Subtracting penalties usually lowers a total by a few tenths and occasionally reshuffles the ranking of a match or a season. Sides that attack through crosses and set pieces lose less than sides that attack through dribbling into the box, because the second group earns more spot kicks. That shift is information about style, not about quality.

The variant also changes how a striker's season is read. A forward who takes his club's penalties carries a scoring load that is largely independent of his movement, so comparing his goals with his non-penalty expected goals measures his open-play contribution while quietly setting his penalty duty aside.

Why analysts default to it

The non-penalty figure is the better predictor of future open-play output for a simple reason: penalties are awarded by referees and are not reliably created from one season to the next. A model that predicts next season's goals from this season's penalty-inclusive expected goals will overweight sides that happened to win spot kicks, and the correction is to remove the incident class entirely.

It also makes cross-competition comparison fairer. Competitions differ in how often penalties are awarded, and a metric that excludes them removes a league-specific constant from the comparison. Where penalties are the subject, they are counted separately and openly rather than hidden inside a blended total.

The information you lose

Stripping penalties does discard something real. Winning and converting penalties is part of how matches are decided, and a side that reliably earns them is genuinely more dangerous. The right response is not to ignore penalties but to count them beside the non-penalty figure, so the pair describes both open-play creation and set-play value.

The same logic applies to penalty shoot-outs in knockout competition, which sit outside both metrics because they are not attempts to score during play. A model that quietly folded shoot-out conversion into a scoring rate would be describing something other than football.

Where the variant still misleads

Non-penalty expected goals inherits every limit of the underlying model. It still treats a shot as a shot regardless of who took it, still cannot see defensive pressure on a video-only feed, and still aggregates a chasing team's desperation efforts into an apparently respectable attacking performance. The prefix removes one distortion and leaves the rest in place.

It is also sometimes quoted as though it were a prediction of open-play goals next season. It is not; it is an estimate of the chance quality a team generated in the season just played. The distinction matters whenever the number is used to argue about what is coming rather than about what happened.

Quoting it consistently

Whenever a non-penalty figure appears beside a goals tally, the two must be counted under the same rule, which means the goals must also be non-penalty goals. Comparing a player's total goals with his non-penalty expected goals is a genuinely common error and it inflates his apparent finishing by the value of his spot kicks.

The disciplined phrasing is to name the class at every use: non-penalty goals against non-penalty expected goals, and penalties reported separately as their own line. That habit costs a few words and removes an entire category of spurious conclusion.

Key reference points

• A penalty is worth about three-quarters of a goal regardless of the move that earned it.
• Removing penalties isolates open-play creation and cuts match-to-match volatility.
• Penalties are not reliably repeatable, which is why predictive work prefers the variant.
• Sides that dribble into the box lose more value than cross-heavy sides.
• Quoting non-penalty expected goals against total goals is a counting error.
• Always report penalties beside the variant rather than hiding them inside a total.

The non-penalty prefix does not make a metric more accurate; it makes it more specific, which is the property that actually matters for comparison.