Team Form and Regression to the Mean

“In form” is one of the most used phrases in football and one of the most misleading. A team scores twelve goals in five games and is suddenly a must-back; a side goes five without a win and the manager is under pressure. Some of those runs reflect real change. Many are just the normal ups and downs of a low-scoring sport. This guide shows how to tell the difference.

How much do results vary by chance?

Take a team whose true scoring rate is 1.4 goals per game, about average for a decent side, and assume nothing about it changes. Over any five-match stretch, a Poisson model says:

• about 17% of the time it will score 10 or more, two a game, and look red hot;
• about 8% of the time it will score 3 or fewer and look like it has forgotten how to finish.

So roughly one five-game stretch in four looks like a hot or cold streak, even though the team is exactly the same throughout. Results run the same way. A side that wins half its matches has about a 45% chance of going at least five league games without a win at some point in a 38-game season, purely by chance.

Regression to the mean

When a result is partly skill and partly luck, extreme runs tend to be followed by more ordinary ones. This is regression to the mean. A team that scored from almost every chance for a month will usually convert fewer of them next month, not because it has got worse, but because part of the hot run was luck that does not repeat.

The practical consequence: the more extreme a run, the more of it is likely to be noise. A side that has won eight in a row is probably good, but probably not as good as eight wins in a row suggests.

Signs that form is real

Some changes do persist. Form is more likely to mean something when:

• The underlying numbers moved too. If a team is creating more and better chances, not just finishing the same chances better, the improvement is more likely to last. Expected goals are the standard way to check this.
• There is a reason. A new manager, a key signing, an injury crisis ending, a change of system.
• The opposition was ordinary. Five wins against the bottom five is less impressive than it looks; five draws against the top five may be better than it looks.
• It lasts. Ten or fifteen matches say far more than three or five.

How models treat form

A good model neither ignores form nor chases it. Most give recent matches more weight than older ones, but still let a whole season or more of results anchor the estimate. That way a genuine improvement shows up gradually, and a lucky week doesn’t swing the numbers.

That is also why a model will sometimes look slow: it keeps rating a struggling big club as a favourite for a few weeks after a bad run. Most of the time, that patience is rewarded.

Reading form on our match pages

• The form strip shows the last few results for each team. Use it as context, not as a forecast.
• Check the home and away splits: a run of wins at home says little about the next away game.
• Look at goals for and against as well as results. A team winning 1-0 every week is on a different kind of run from one winning 4-3.
• When form and the model disagree, ask why before siding with either.

For how few games it takes for luck to fade, see how many games before the table tells the truth.

18+. These figures are information, not betting advice. Please read our responsible gambling page.