When do goals come? The last fifteen minutes, and what they do to our predictions
Almost a quarter of all goals come after the 75th minute: 23.3% of the 14,552 we counted, from 5,051 matches between 17 July and 31 October 2026. Those closing minutes changed the result of 22.9% of matches. We put that next to the call our model made before kick-off: late goals turned a wrong call into a right one 449 times, and a right one into a wrong one 242 times. The reason is not the one we expected.
Fifteen minutes at a time
Goals are scarce early on and come faster as the match goes on. The first fifteen minutes hold 12.9% of them; the last fifteen, with stoppage time, 23.3%. The first half as a whole gives 44.0% and the second 56.0%. Stoppage time alone produced 1,024 goals, 704 of them after the ninetieth minute.
Penalties lean late even more than goals do: 27.7% of the 1,110 scored penalties came after the 75th minute.
| Minutes | Goals | Share | Penalties |
|---|---|---|---|
| 1–15 | 1,884 | 12.9% | 118 |
| 16–30 | 2,021 | 13.9% | 153 |
| 31–45+ | 2,491 | 17.1% | 209 |
| 46–60 | 2,428 | 16.7% | 137 |
| 61–75 | 2,341 | 16.1% | 186 |
| 76–90+ | 3,387 | 23.3% | 307 |
How safe is a lead with fifteen to go?
The busiest spell of the match is also the one that decides it. 22.9% of results changed after the 75th minute, 14.5% after the 85th and 6.0% in stoppage time at the end. Where they change depends on the score at that point:
| Score at 75' | Matches | Same result | Changed |
|---|---|---|---|
| Ahead by 2 or more | 1,541 | 97.9% lead held | 2 lead lost |
| Ahead by 1 | 2,039 | 76.6% lead held | 93 lead lost |
| Level | 1,471 | 55.9% stayed level | 648 found a winner |
A two-goal lead is close to safe: 2 of 1,541 were lost outright. A one-goal lead is not. About one in four slipped, mostly to a draw; 93 turned into a defeat. A level match with fifteen to go is the least settled of all: 44.1% of them found a winner. The side behind does not get most of the late goals, either. It scored 1,170; the side in front scored 1,274.
What the closing minutes do to our calls
Every match here has a call our model published before kick-off: home win, draw or away win, whichever it gave the highest chance. We scored those 4,857 club calls twice: against the score at 75 minutes and against the final score.
- At 75 minutes the call was right in 44.7% of matches.
- At full time it was right in 48.9%.
- In 691 matches a late goal changed the verdict. 449 went our way and 242 went against us: 65.0% in our favour, with a margin of ±3.6 points.
So the closing minutes are worth about 4 points of our hit rate. A model judged at the 75th minute would look worse than it is. How well those probabilities match what happens is a separate question, which our last check of the favourites answers.
Do favourites finish stronger?
Our first guess was fitness and depth: the stronger side brings better players off the bench and wears the other down. If that were true, favourites would score a bigger share of the late goals than of the earlier ones. They do not. The side we made favourite scored 59.3% of the goals up to the 75th minute and 60.0% after it. The gap, 0.7 points, is well inside its margin of ±1.9. In this data the stronger side is as much stronger in the last fifteen minutes as in the first seventy-five, and no more.
The draw that breaks
The real reason is in what we predict. A draw is almost never the most likely of the three outcomes, so our model picked one in 40 of 4,857 matches. Almost every match that is level at the 75th minute therefore counts as a wrong call at that moment, and a level match is the one most likely to change.
When a level match did find a late winner, it was the side we had picked in 387 cases and the other side in 232: 62.5% our way, close to the share of all goals our favourites score. Nothing special happens late. The favourite simply keeps being the favourite, and the late goal breaks a draw we had never called.
Reading a prediction during a match
Our win, draw and win percentages are statements about ninety minutes, made before the first one. A favourite drawing with fifteen to go has not proved the number wrong: in this data 44.1% of matches level at that point still found a winner, and when they did it was our pick 62.5% of the time. A favourite one goal up has not proved it right either; a quarter of those leads did not survive. Each finished call is scored on the final score at ninety minutes, never at an earlier moment, and the track record shows the result for every one.
The matches behind this
Every match from 17 July to 31 October 2026 that finished in ninety minutes and whose goal list adds up to its final score: 5,051 matches across club and national-team football. We left out 45 more whose goal list did not match the score. Own goals count for the side that benefited. A goal in stoppage time belongs to the half it was scored in, so a goal at 90+3 is a late goal, not an early one.
The parts about our calls use the club matches only, 4,857 of them, because our national-team model is set up differently. Each call was logged before kick-off and none has been changed since. The margins are 95% intervals.
Published by Fulltime Stats, an independent site run by Leo Tempo. Who we are.