Football is finding more of its goals after the 90th minute, and nobody can prove why

Start with the number that should not, on its own, mean very much: 9.41. That is the percentage of all goals scored in the big five leagues last season that arrived from the 90th minute or later. In isolation it is trivia. Set against the other nine seasons this reporting spans, it is the highest late-goal share on record, and it did not arrive alone.

Go back to 2016 and the figure sat at 7.13%. It fell as low as 6.13% in 2020, the empty-stadium season, which belongs in this account but not as a like-for-like year against the ones either side of it: a compressed calendar and grounds without crowds are not conditions football has otherwise had to reckon with, and a single low point inside a long run deserves to be flagged rather than folded in as if it were normal. What matters more is what happened after it. The climb from there was not a straight line: 7.67% in 2021 gave way to a fall to 6.93% in 2022, before a jump to 8.79% in 2023, a second dip to 8.10% in 2024, and then the series high of 9.41% last season. Two dips sit inside that run, but the destination is unmistakable: two of the three highest readings in the decade have come in the last two years.

That rise happens to track, almost minute for minute, a different chart entirely: how many substitutions get made in a match. From 2016 to 2018, every one of the five leagues sat in a tight band, roughly 5.6 to 5.9 substitutions per match. From 2019 onward the numbers began to climb, and by 2021 Bundesliga clubs were making 8.68 changes a match, La Liga 9.07, Serie A 8.83, Ligue 1 8.34. The Premier League lagged behind, still at 5.59 that season while the rest of Europe had already jumped, before it converged with the others: 7.86 in 2022, 8.44 by 2024, and 8.24 last season. By 2025 every one of the five leagues sits between 8.24 and 9.46 substitutions per match, roughly half as many again as the pre-2019 norm.

Put the two charts next to each other and the temptation is obvious. More substitutions means fresher legs late on, and fresher legs, the story goes, should mean more goals when the pressing intensity of a tired defence starts to fail. It is a clean mechanism, and it might even be the right one. But it is not what these figures show. The two series were counted separately: one is a share of goals by minute, the other is a count of substitutions by match. Nothing in either query traces a goal back to the changes that preceded it, and no amount of eyeballing two lines moving in the same direction turns a coincidence into a cause. What can be said honestly is narrower and still worth saying: the timing lines up. The late-goal share was essentially flat through the years when substitutions per match sat in the 5.6 to 5.9 band, it dipped in the outlier 2020 season, the same season in which substitution counts jumped sharply for four of the five leagues, a year this reporting treats as non-comparable rather than as a clean data point, and then both climbed together from 2021 onward as substitution totals settled at their new, higher level across all five leagues. That is a pattern worth watching, not a mechanism proven.

There is a second reason for caution, and it sits inside the substitution numbers themselves rather than in the relationship between the two series. Several season-competition rows in this window carry reduced match counts: Ligue 1 played only 279 matches in 2019 rather than a full 380, and its 2023, 2024 and 2025 seasons ran to 308 or 310 matches rather than the standard round number. Serie A's 2022 and 2023 counts drift slightly too. None of this touches the late-goal share, which is a ratio of goals to goals and carries no denominator to distort. But the substitutions-per-match figure is exactly the kind of per-match rate that a reduced fixture list can quietly bend, and treating every season's match count as equivalent when some of them are not is the kind of error a tidy chart hides rather than reveals.

What the two series do support, taken on their own separate terms, is a clearer statement than a causal one: football in the big five leagues now records far more changes made in a match than it did a decade ago, and it has, in the same recent years, produced a larger share of its goals after the 90th minute than at any other point in this data. Whether one is feeding the other, whether both are downstream of something else entirely, or whether they simply share a decade and nothing more, is not a question minute-stamped goal counts and substitution totals can settle. That is not a failure of the numbers. It is what an honest reading of them looks like: two trends, both real, laid next to each other rather than welded together.

The more interesting question, and one these counts cannot answer, is why the Premier League's substitution rate sat so far below its continental peers through the early years of this rise, at 5.59 in 2021 against figures above 8.3 elsewhere, before closing most of that gap by 2025. What was different about how the Premier League made its changes in those years, and why it converged with the rest of Europe when it did, is a question for a different set of queries.