The impact substitute is a sample artifact
Every season a substitute scores four goals in a hundred and twenty minutes and the per-90 table turns him into a phenomenon. The framing writes itself: a player who changes games, a finisher the manager is wasting, a super-sub with some unquantifiable knack for arriving late. The rate is real arithmetic on real goals. What it is not is a comparable number, because the minutes a substitute plays are the least representative minutes in football, and the per-90 convention that makes the comparison possible is precisely what makes it wrong.
The last twenty minutes are a different game
Matches open up late. Players are tired, defensive shape degrades, and at least one team is usually chasing a result and committing bodies forward. Goals per minute rise measurably in the final fifteen minutes relative to the first fifteen, and that rise belongs to the phase of play rather than to the players on the pitch for it. A substitute is dropped into the highest-scoring segment of the match by definition. Scaling his output to ninety minutes assumes the rest of the game would have looked like the part he played, and it would not.
Fresh legs against tired ones
The physical mismatch is the second half of the same effect. A forward entering in the seventieth minute runs at defenders who have covered ten kilometres. The advantage is genuine, it is the reason substitutes exist, and it is an advantage of circumstance rather than quality. The same player starting would face organized defenders for an hour before inheriting the tired ones, and his rate would fall accordingly. Managers understand this perfectly, which is why the player kept in that role is frequently kept there on purpose.
Added time compounds it. Stoppage-time minutes are the most chaotic in the match, and substitutes are on the pitch for nearly all of them while the starters they replaced are on the bench. Those minutes also happen to be the ones most likely to produce dramatic goals, which is how the super-sub reputation gets its highlight reel as well as its arithmetic. Memory and the per-90 table reinforce each other, and both are sampling the same unrepresentative window.
Selection effects on both sides
Substitutions are not random. Attacking substitutes come on when a team needs a goal, which means they enter matches where their side is pressing, holding territory, and generating chances — and often against an opponent sitting deep with a lead, which creates a specific kind of chance. Defensive substitutes enter to protect leads and will never post attacking numbers. The pool of minutes is sorted by game state before anyone touches the ball.
Opposition shape matters as much as opposition fitness. A substitute entering against a team defending a one-goal lead meets a low block, which suits some profiles and ruins others; one entering against a side chasing meets acres of space behind a high line. Neither is the balanced game a starter plays. The minutes are not merely fewer, they are drawn from a different distribution of match states.
The denominator is tiny
This is the plainest problem. A season of cameo appearances might total four hundred minutes, which is four and a half matches. Three goals in that span produces a per-90 figure that would lead Europe over a full season, and four hundred minutes is nowhere near enough to distinguish a good finisher from a lucky one. Goal-scoring rates need something like a thousand minutes before they begin to settle, and even then they carry substantial noise. Published per-90 tables with minimum-minutes filters exist exactly to keep these players off the list, and the filters get ignored whenever the story is good.
There is also a quieter sorting at the squad level. The players who spend a season on the bench are, on average, the ones a manager rates below his starters — that is what being a substitute means. So any pool of substitute minutes is pre-filtered toward lesser players, and the handful who post gaudy rates within it are the right tail of a weaker distribution. Picking the extreme value out of a small, pre-selected sample is the oldest way to be fooled by data.
Does the knack exist at all
The honest answer is that it is hard to find in the data. Attempts to identify players who are reliably better as substitutes than as starters mostly turn up noise: the players who post extreme sub rates one season generally do not repeat them, and those promoted to the starting eleven on the strength of a sub record typically regress to something near their underlying level. There may be a small real effect for certain profiles — a direct runner suits an open, stretched game — but it is a tactical fit rather than a temperament, and it is far smaller than the per-90 table implies.
Managers price all of this correctly, which is worth noting given how often the data is used to argue they have not. A coach who keeps a productive substitute on the bench is usually not failing to notice the goals; he is weighing them against what the player does for seventy minutes against an organised defence, how he presses, and whether the shape works with him in it. The per-90 table sees one of those things.
Five substitutions changed the arithmetic
The permanent move to five substitutions expanded the minutes in question considerably, which cuts both ways. There are more sub minutes to accumulate, so the samples are less absurd than they were under three. But teams now also rotate more aggressively and make tactical changes earlier, which means the substitute pool includes more players entering at the hour mark into matches that are still competitive and organized. The composition of those minutes has shifted, so comparisons to the historical super-sub era are not clean either.
Assists behave the same way and draw less scrutiny. A substitute winger who comes on and crosses for two late goals posts a creative rate that no starter can approach, generated against a defence that has stopped tracking runners. The mechanism is identical to the scoring case, and so is the correction: compare him to other players in similar minutes, or do not compare him at all.
How to read it on StatLine
StatLine’s Premier League player tables carry appearances, minutes, goals, assists, and shots, which is everything needed to deflate or confirm a super-sub story. Start with minutes: divide them by appearances and you can see immediately whether you are looking at a starter or a cameo player. Then compare shots to goals. A substitute scoring at a high rate on very few attempts is riding conversion that will not last; one generating shots at a strong rate per minute has at least a plausible case for more responsibility.
The honest read
The impact substitute is mostly an artifact of when substitutes play and how few minutes they play. Late-match football is more open, tired defenders are easier to beat, game state sorts the minutes before kickoff of the second half, and the samples are small enough to produce any rate you like. That does not mean substitutes are unimportant — changing a match from the bench is a real contribution and managers win points with it. It means the per-90 figure is not evidence a player deserves to start, and clubs that have promoted players on that basis have a long and consistent record of disappointment. The goals happened. The rate they imply never existed.