Do teams really concede late goals? We checked 37 leagues
"This team always concedes late" is one of the most common lines in match previews. We tested it on data from 37 leagues, and the answer is clear: the minutes in which a team scored and conceded over its last 38 matches tell you practically nothing about when the goals will come in its next 10. The correlation was −0.02 for goals scored and 0.00 for goals conceded. This article shows how we tested it, why such patterns look so convincing and what it means for late-goal bets.
Key takeaways
- A team conceding 1.3 goals a game lets in about 8 goals per 15-minute spell over 38 matches, so 11 instead of 8 is ordinary chance.
- Patterns such as "sits back when leading" did not repeat either: none of the correlations was more than 0.05 away from zero.
- How many goals come late depends mainly on how well the teams attack and defend, not on how their goals were spread across the minutes.
- Late-goal markets usually carry a higher margin than the main markets on the result or the number of goals.
Where the myth comes from
A chart of goals by 15-minute spell is very persuasive. If a team has conceded 12 of its 30 goals after the 75th minute, it is hard not to think of poor fitness, bad substitutions or lapses in concentration. Commentators, stats in apps and match previews repeat such numbers as if they were a trait of the team.
The trouble is that the number cannot tell a trait from chance. To settle it, you have to check whether a pattern seen in the past repeats in the future.
How we tested it
The test was simple and fair to the myth:
- For every team in 37 leagues we picked a point in the season (a checkpoint) and took its last 38 league matches.
- We split the goals scored and conceded into six 15-minute spells: minutes 1–15, 16–30, 31–45, 46–60, 61–75 and 76–90.
- We compared each spell with what was expected. The expectation allows for how many goals the team should have scored in each match and for how goals are usually spread over time in that league. In most leagues more goals fall in the last 15 minutes than in the first, so "lots of goals after the 75th minute" is often just the norm.
- We did the same for the team's next 10 matches and checked whether a surplus or a shortfall in a given spell came back.
That gave 11,142 checkpoints in all. If "conceding late" were a lasting trait, teams with a surplus of goals conceded after the 75th minute should have kept that surplus in their next matches.
The result: the pattern does not repeat
The Pearson correlation between past and future was:
- −0.02 for goals scored in each spell,
- 0.00 for goals conceded.
A correlation of 0 means no link at all, while 1 would mean a perfect repeat. Values this close to zero say that knowing when a team has conceded so far does not help predict when it will concede in its next matches.
We also tested the second popular myth: that some teams "sit back when they lead" or "push hard when they are behind". Here we used xG rather than goals, because there is more of it and it is less random. We measured how a team's xG and its opponent's change with the game state. These patterns barely repeated either: none of the correlations was above 0.05 in absolute value.
That does not mean teams ignore the score. Almost all of them attack a little less when they lead and a little more after going behind. The point is that the differences between teams in that reaction are so small and so unstable in the data that they cannot be told apart from noise.
Why the patterns look so convincing
A small sample
38 matches sounds like a lot, but goals are rare. A team that concedes 1.3 goals a game on average lets in about 50 goals in 38 matches. Split into six spells, that is about 8 goals per spell.
With numbers this small, chance does its work. If you expect 8 goals, 5 or 11 is perfectly normal, even for a team with nothing special about it. Yet 11 instead of 8 looks on a chart like "concedes 40% more late on". Look at the 20 teams of a league and a few of them will show deviations like that by pure chance.
Regression to the mean
The teams with the most extreme charts are largely the ones whose goals happened to fall in an unusual way. In the next matches chance spreads them differently, so the numbers drift back towards the average. This is called regression to the mean, and it affects everything measured on small samples, from strikers' finishing to winning streaks.
We remember stories, not numbers
A goal conceded in the 89th minute sticks in the mind; one conceded in the 23rd does not. Late goals also change results, so they make the highlights and the commentary more often. It is easy to build a picture of a team that "always concedes at the end" from a handful of memorable matches.
What this means for late-goal bets
- A chart of 15-minute spells is not a reason to bet. If the only reason to back a goal after the 75th minute, or "team X to concede in the second half", is the record of recent matches, our data gives no grounds for it.
- What counts is the teams' strength, not their "character" over time. How many goals a match produces, late on included, depends mainly on how well the teams attack and defend in general. Attack and defence ratings describe that better than the minute-by-minute spread of goals.
- More goals after the break, but no HT/FT reversals to order. In 281,843 matches from 37 leagues the second half averaged 1.51 goals against 1.18 in the first, yet the half-time result was turned around in only 4.6% of matches (see our article on HT/FT reversals).
- Late-goal markets usually carry a higher margin than the main markets on the match result or total goals. Without a real information edge, you pay more for the same risk.
If you still enjoy these bets, treat them as entertainment with a budget set in advance. Our bankroll rules will help.
How we show this data on SharkBetz
On each club's page (you reach them from the club rankings) you will find the "When the goals fall" chart and the "Game state" table. We show them because they describe well what happened in recent matches. Each of them, though, carries a clear note: this describes the past, it is not a forecast.
That is why next to the actual goals we also show the goals expected in each spell. A large gap between a bar and the expectation usually means chance rather than a trait of the team, especially when the spell had only a few goals.
FAQ
Do some teams really concede more goals late on?
Individual teams can have such a tendency for a while, for example because of a thin bench. But in the data from 37 leagues there is no sign that such patterns hold from match to match firmly enough to be predicted.
Why are the most goals scored in the last 15 minutes?
It is a general effect, seen in almost every league: fatigue, substitutions, stoppage time and teams taking risks to turn the result around. It applies to every team, though, so it sets none of them apart.
How many matches does a pattern need to be reliable?
More than a team plays with an unchanged side. Even 38 matches give only a few goals per spell, and over that time players and managers change.
So are these goal timing charts useless?
No. They describe past matches well and help you understand how a team's season went. They just should not be treated as a forecast.