The 90th minute of a football match feels decisive. The broadcast cuts to the manager. The crowd rises. The stoppage-time board goes up. The next goal, surely, decides the match.
It does not. Across 1,751 matches in the 2024-25 Big Five, 848 goals were scored from the 80th minute onwards. Of those, 19% were winners (the goal that put the scoring side ahead for the final time), 22% were equalisers (the goal that tied the match), and 16% were consolations (the goal that reduced a losing margin without leveling it). The remaining 44% were the goal that followed a goal that followed a goal: lead-extenders in a match whose outcome was already decided. The broadcast cuts away. The score keeps moving.
What we measured
We pulled every match in the 2024-25 season across the Premier League, La Liga, Serie A, the Bundesliga, and Ligue 1 from Understat, which carries per-shot minute data and a per-shot score reconstruction. The dataset is 1,751 matches and 4,949 goals. For each goal scored from the 80th minute onwards, we walked the in-match goal sequence to reconstruct the score at the moment of the goal, then classified the goal.
The windows are minutes 80 to 90 inclusive and 90+ (90+1, 90+2, and onwards as recorded in the official minute column). Understat records the official minute, not the elapsed clock minute, so a stoppage-time goal in the third added minute is 90+3, not 93.
The classification has four buckets. Equaliser: pre-goal was untied and post-goal is tied (e.g. 0-1 to 1-1). Winner: the goal puts the scoring side ahead for the first time and the scoring side holds that lead to full time. This is the match-winning goal specifically. Consolation: the scoring side was behind pre-goal and is still behind post-goal (the margin is reduced but not leveled). Other: lead-extenders after the match is already decided, plus the rare case of a goal that briefly led and was later caught. We flag rather than guess in this bucket.
The own goal in Understat’s data is credited to the team that benefited (so an own goal by a defender at 0-0 is a 1-0 to the opponent). The full classification logic and the running-score reconstruction are in the methodology note at the end.
The data
Late goals as a share of all goals
| League | 2024-25 goals | Late goals (80+) | Late share |
|---|---|---|---|
| Premier League | 1,115 | 201 | 18.0% |
| La Liga | 995 | 186 | 18.7% |
| Serie A | 973 | 149 | 15.3% |
| Bundesliga* | 955 | 151 | 15.8% |
| Ligue 1 | 911 | 161 | 17.7% |
| Big Five | 4,949 | 848 | 17.1% |
* Bundesliga 2024-25 reconciled to 305 of 306 matches (one match missing from the source feed); the published league total is 959 goals. The four-goal gap is within the rounding noise of the Wikipedia total.
Late goals are 17% of all goals across the Big Five. That is roughly one late goal per match. The shares cluster between 15% and 19% across the five leagues. Serie A has the lowest late-goal share (15.3%) and La Liga the highest (18.7%).
Classification of late goals, by window
The 80 to 90 window:
| League | n | Equaliser | Winner | Consolation | Other |
|---|---|---|---|---|---|
| Premier League | 130 | 22.3% | 11.5% | 16.2% | 50.0% |
| La Liga | 118 | 20.3% | 19.5% | 16.1% | 44.1% |
| Serie A | 103 | 26.2% | 25.2% | 10.7% | 37.9% |
| Bundesliga | 108 | 25.0% | 13.9% | 23.1% | 38.0% |
| Ligue 1 | 103 | 12.6% | 21.4% | 8.7% | 57.3% |
| Big Five | 562 | 21.4% | 18.0% | 15.1% | 45.6% |
The 90+ window:
| League | n | Equaliser | Winner | Consolation | Other |
|---|---|---|---|---|---|
| Premier League | 71 | 23.9% | 23.9% | 16.9% | 35.2% |
| La Liga | 68 | 19.1% | 29.4% | 22.1% | 29.4% |
| Serie A | 46 | 34.8% | 13.0% | 15.2% | 37.0% |
| Bundesliga | 43 | 18.6% | 20.9% | 9.3% | 51.2% |
| Ligue 1 | 58 | 13.8% | 15.5% | 15.5% | 55.2% |
| Big Five | 286 | 21.7% | 21.3% | 16.4% | 40.6% |
The two windows stacked (all late goals, 80+):
| League | n | Equaliser | Winner | Consolation | Other |
|---|---|---|---|---|---|
| Premier League | 201 | 22.9% | 15.9% | 16.4% | 44.8% |
| La Liga | 186 | 19.9% | 23.1% | 18.3% | 38.7% |
| Serie A | 149 | 28.9% | 21.5% | 12.1% | 37.6% |
| Bundesliga | 151 | 23.2% | 15.9% | 19.2% | 41.7% |
| Ligue 1 | 161 | 13.0% | 19.3% | 11.2% | 56.5% |
| Big Five | 848 | 21.5% | 19.1% | 15.6% | 43.9% |
What the table shows
The match is usually decided before the 80th minute
The 80th minute cuts ten minutes from full time plus stoppage. The natural question is: how often is the match still in play by then? Roughly one in two matches is, by some measure. The “other” bucket, dominated by lead-extenders, sits at 44% of late goals across the Big Five. That is the share of late goals that come in a match whose outcome has already been settled. They move the score. They do not move the result.
The cleanest read is that 80+ goals split into two regimes. The first is the match in motion: a side is trailing or tied and a late goal is going to either rescue a point or seal three. The second is the match over: a side is already ahead by a margin the opponent cannot recover, and the goals that follow are making the scoreline uglier. The “other” bucket is the second regime. The other three buckets together are the first.
Winners concentrate in the 90+ window
Across the Big Five, the winner share rises from 18% in the 80 to 90 window to 21% in the 90+ window. The “other” share drops from 46% to 41% over the same cut. The shift makes sense: by the 90+ window, the matches that are still in motion are concentrated in the ones that are genuinely tight. Matches that are 3-0 or 4-1 at the 88th minute do not often produce winners at 90+3.
La Liga shows the cleanest version of this pattern. The La Liga winner share jumps from 19.5% in 80 to 90 to 29.4% in 90+. The 90+ window is where La Liga matches are most often decided by a single goal. The Premier League shows the same direction with less amplitude (11.5% to 23.9%).
Serie A is the late-equaliser league
Serie A’s late equaliser share of 28.9% (all-late) is the highest of the five leagues. In the 90+ window alone, Serie A late equalisers rise to 34.8% of all Serie A late goals, again the highest of the five leagues. This fits the broader Serie A pattern of late comebacks and long, drawn-out tactical matches where the trailing side finds an equaliser deep into stoppage time. The half-time lead conversion data and the 1-0 half-time lead post both point in the same direction: Serie A matches are the most recoverable in the Big Five.
Ligue 1 is the inverse. Ligue 1’s late equaliser share of 13.0% is the lowest. Ligue 1 late goals are dominated by the “other” bucket at 56.5%, the highest of the five leagues. The reading is structural: Ligue 1 matches in 2024-25 produced the largest average margin in the Big Five, and once a Ligue 1 side is ahead by two or more, the opponent rarely finds a way back.
The Premier League 80 to 90 window is the noisiest
The Premier League’s 80 to 90 winner share of 11.5% is the lowest in the dataset. The Premier League 80 to 90 window produces more “other” goals (50%) and fewer winners than any other league’s 80 to 90 window. The shape flips at 90+, where the Premier League’s winner share jumps to 23.9%, the joint highest of the five leagues.
The interpretation is that the Premier League’s late-window structure is more compressed than the other leagues’ matches. The first ten minutes of stoppage time (80 to 90) are when Premier League sides are still making tactical changes, still running, still probing. The actual decision comes in the official stoppage time, when the substitutions have settled and the tired legs show. The Premier League’s 80 to 90 window is the substitute-and-press window. The 90+ window is the goal window.
What we tried that did not work
We first classified late goals using only the pre-goal and post-goal scores, without checking whether the goal was the first time the scoring side took the lead. That classified every lead-extending goal as a winner. The 2-0 in a 2-0 win where the 1-0 was the actual match-winning goal was counted twice. The “other” bucket was empty, and the winner share inflated to 60%. The reading was unusable: nearly two thirds of late goals are not the goal that decided the match.
We then split the classification into “first time the side took the lead” versus “extending a lead already held.” That fixed the winner share to 19% and surfaced the 44% lead-extender bucket. The two-pass walk through the goal sequence (reconstruct the running score, then classify each goal) is the only version of the analysis that produces the right shape.
Known limitations
The dataset is one season. The 2024-25 Big Five figures are point estimates for a single season’s worth of data. Differences between leagues of less than 5 percentage points should be read as suggestive, not significant. The Bundesliga row reconciles to 305 of 306 matches in the source feed; the four-goal gap to the published Wikipedia total of 959 is within rounding noise.
The classification of “other” is the residual. A lead-extender in a 4-1 win and a goal in a 1-1 draw that gets overturned in stoppage time are both “other” by the working definition. The first case is the dominant one; the second is rare at minute 80+. We have not broken the “other” bucket out further in this post.
A goal that ties the match in the 90th minute, then concedes a winner in the 92nd, is counted as a winner in our data, not as an equaliser. The winner is the goal that put the scoring side ahead for the final time. This matches the match-winning goal definition used in the half-time lead post.
What this means for an analysis
A pre-game analysis that treats late goals as a uniform signal is wrong in two directions. It over-counts the importance of the late window because the broadcast cuts to the late-window goals. It under-counts the share of late goals that come in already-decided matches. The 80+ window is real, but it is noisier than it looks. About one in five late goals is a winner. About one in five is an equaliser. The remaining three in five are the score moving after the match is over.
The right read is a per-window calibration. The 80 to 90 window is closer to a continuation of the second half, with similar goal-mix proportions. The 90+ window is where winners concentrate and where the late-window tactical substitutions actually pay off. A model that treats the two windows identically will mis-price the 90+ window. The Bundesliga and La Liga 90+ numbers are the cleanest examples: 29% winners in 90+ is meaningfully different from 19% in 80-90.
For a reader following coverage across leagues, the late-window split is one of the cleanest examples of why league-specific calibration matters. The model treats Serie A and Ligue 1 as different competitions because the underlying late-window distributions are different, and the share of late goals that are lead-extenders (37.6% in Serie A, 56.5% in Ligue 1) is part of that difference.
A read is a lens for separating signal from noise. The 90th minute is signal, but only about a fifth of the time. The other four fifths are the scoreboard catching up to a match that ended ten minutes earlier.