Game-State Effects on Stats: How the Scoreline Bends Possession, Shots, and xG

Pull up the stat sheet from almost any match that ended in an upset and you will find a strange little lie hiding in it: the beaten favourite often "won" the possession, the shot count, sometimes even the expected goals. I used to read those numbers as evidence the result was a fluke. I no longer do, because in most cases the numbers are not describing dominance at all — they are describing a team that fell behind and spent an hour chasing the game. The losing is what produced the gaudy totals, not the other way round.

That inversion is the whole subject of this piece. A match statistic compiled while a team is two goals down is a fundamentally different animal from the same statistic compiled at 0–0, and pretending they are interchangeable is how you end up trusting a box score that flatly contradicts who was better. The scoreline a number was built under has a name among analysts: game state.

What game state is

Game state is the scoreline a statistic was accumulated in — whether the team was winning, level, or losing at the instant each pass, shot, or tackle happened, and often by how much and with how much time remaining. It sounds like a footnote on a box score. It behaves more like a hidden variable that quietly bends almost every number on the match sheet, because the scoreline changes how both teams choose to play.

The mechanism is incentives. A side that goes a goal up has its objective inverted: the most valuable thing it can now do is protect the lead, so it drops deeper, keeps its shape, stops sending bodies forward, and is content to let the opponent hold the ball in front of it. The side that goes a goal down has the opposite incentive. It has to score, so it pushes players up, presses higher, takes more risks, and shoots from worse positions because a low-percentage shot beats no shot at all. The football you watch after a goal is partly a product of the goal, not of the two teams' underlying quality. They are, in effect, now playing two different games, and the final box score averages them into one misleading line.

How it distorts the headline numbers

Once you start looking for game-state effects, you cannot stop seeing them — their fingerprints are on every aggregate stat. Three matter most.

Possession. This is the clearest victim. A team protecting a lead concedes the ball deliberately; a team chasing the game hoards it because it must. So high possession is as much a symptom of being behind as a sign of control. "Dominating possession" and "losing" are not opposites — they routinely arrive together, and a possession figure read without the scoreline can tell you almost the reverse of who controlled the match.

Shot volume. The trailing team throws bodies and shots forward; the leading team is happy to defend and break. The result is the most robust regularity in this whole area: losing teams shoot more. A team that falls behind early will usually out-shoot the opponent over ninety minutes, which flatly inverts the intuition that out-shooting a team means you were on top. A raw shot count is one of the most scoreline-contaminated numbers on the sheet.

Expected goals. Even xG — the metric built specifically to be more honest than raw shots — inherits the bias, because xG is summed over shots, and the trailing team takes more of them. A side can lose the match and "win" the xG simply because it spent an hour pouring forward in search of an equaliser, accumulating a stack of half-chances. xG totals are more trustworthy than shot counts, but they are not immune to game state. (For what xG does and does not claim in the first place, see expected goals explained.)

Defensive and pressing numbers swing the same way. Tackles, interceptions, and pressing intensity all move with the scoreline — a team defending a lead defends more and presses less, none of which makes it the better defensive side. The same contamination is why a full-match PPDA blends a desperate late press and a comfortable early stand-off into one uninformative average.

A worked illustration

Consider two accounts of the same fictional match. The numbers below are made up purely to demonstrate the effect, not drawn from any real game.

Read only the full-match line and Team B looks dominant: more of the ball late and sixteen shots to eight. Read the split and the story flips on its head. While the score was level — the fairest window onto the two sides' true strengths — they were near-even, with Team A marginally ahead. Everything that makes Team B's totals look impressive happened after they fell behind and were forced to chase. The single goal did not just change the score; it manufactured the statistics that appear to contradict it.

How analysts adjust for it

The fix is not to throw the numbers away but to condition them on the scoreline. A few standard moves do most of the work.

Split the stats by game state. The simplest and most powerful step: report numbers separately for winning, level, and losing, instead of mashing them together. The three splits often tell three different stories, and the contrast between them is itself informative — a team whose performance falls apart when ahead has a problem the full-match average would hide entirely.

Lean on the "while level" window. Performance recorded when the score is level is the closest thing to a neutral baseline, because neither team has yet had its incentives distorted. For the same reason, first-half numbers tend to be less score-affected than second-half ones, since more of the first half is typically played at 0–0. Many analysts treat "while level" xG as a cleaner read on which side was actually better than the full-time total.

Use score-state-adjusted models. The most rigorous approach builds the scoreline directly into the model, weighting or correcting each event for the game state it occurred in, so a shot taken at 2–0 down is not counted the same as one taken at 0–0. This is close kin to possession-adjusting defensive stats so that sheer volume of opportunity does not masquerade as quality; we cover that machinery in possession-adjusted stats.

Why it matters beyond one match

Game state is not only a single-match trap; it feeds straight into the models that project whole seasons and tournaments. A model that ingests raw shot or possession totals without accounting for the scorelines they were built under will systematically misjudge teams — over-rating sides that spend a lot of time behind and accumulate empty volume, under-rating efficient front-runners who score early and then defend a lead. Differences in exactly how forecasters handle this context are one reason two reputable projections of the same competition can disagree, a theme we pick up in why league projection models disagree.

This will matter throughout the World Cup. A favourite that scores first and then sees out the game in second gear can finish with modest possession and few shots and still have been comfortably the better team; an eliminated side that chased three group games from behind can leave with flattering aggregate numbers. Reading those box scores without the scoreline attached would invert several of the tournament's real stories.

The takeaway is a habit, not a formula. Before you let any match statistic settle an argument, ask the first question: what was the score while this number was being made? A possession figure, a shot count, even an xG total means something quite different at 0–0 than it does at 2–0 down with twenty minutes left. The scoreline is not context you can bolt on afterwards. It is the lens the entire box score was shot through. Where a live, game-state-split table for the tournament would sit, we will add one once matches have actually been played.

Sources & further reading

• Free textbook: Chapter 16: Team Performance Analysis — the theory behind this, at DataField.dev.
• StatsBomb — analysis and methodology on score effects and game-state splits.
• StatsBomb open data — event data timestamped and scored so events can be bucketed by game state.
• Understat — match and season xG data for the major European leagues.
• FBref — match logs and advanced stats useful for splitting performance by scoreline.