Regression to the Mean in Betting: Why Hot and Cold Runs Fade

A goalkeeper who has not conceded in five matches is on every highlights package. The commentators call him unbeatable, the graphics count the clean sheets, and the price on his team keeping another one starts to look like a bargain to anyone who has watched the run. Then the next three games bring six goals against. Nothing obvious changed. The keeper did not suddenly get worse. What happened has a name, and it matters every time someone reads form into a price.

What regression to the mean means

Regression to the mean is the tendency for an unusually high or low result to be followed by one closer to the average. It appears whenever an outcome depends partly on skill and partly on chance. An extreme result usually means both parts pointed the same way: the performer was good, and luck helped. The skill tends to stay; the luck does not repeat on demand. So the next result, on average, sits nearer to the performer's true level.

It is not a force that pulls results back, and it does not mean a lucky team is "due" a bad game. Each new match starts fresh. The point is only that the extreme result was a poor guide to the underlying level in the first place.

A simple way to picture it

Imagine a large group of people each flipping a fair coin ten times. A handful will get eight or nine heads. Ask only those people to flip ten more times, and their second scores will cluster around five, because there was never anything special about their coins. Now replace coins with something that mixes ability and chance, such as goals scored per match. The top performers of one month will still be decent the next month, just usually less spectacular than their best stretch suggested.

Where it shows up in sport

• Shooting and scoring streaks. A player converting an unusually high share of chances over a few games tends to settle back towards their longer-term rate.
• Goalkeeping and save rates. Short runs of very high or very low save percentages are common and often fade.
• Close-game records. Teams that win most of their one-goal or one-score games over part of a season frequently see that record level out later.
• Early-season tables. Positions after a handful of rounds rest on few results, so surprising teams at the top or bottom often drift towards the middle.
• Against-the-spread records. A side that has covered the spread in many recent games has usually been beating expectations, and the line adjusts. Our explainer on covering the spread shows why those records swing so much.

Why prices often already account for it

People who set odds generally lean on long-run measures of performance rather than the last few scores. A team on a hot run will usually see its price shorten a little, but often not as much as the excitement around it would suggest. Public money, however, tends to follow the streak. That gap is the reason runs feel like easy reads and rarely are.

Live coverage adds to the pull. Graphics count streaks, panels discuss confidence and momentum, and a few vivid minutes can feel decisive. Separating the story from the signal is the theme of our piece on what a game broadcast can tell you about the odds, and regression to the mean is one of the main reasons that separation matters.

The same effect in your own betting record

Regression to the mean does not only apply to teams. A short run of winning bets is the most tempting place to draw a conclusion, and one of the least reliable. A handful of results says very little about whether a way of choosing bets has any edge after the bookmaker's margin. Over a long run of bets at prices that include a margin, results tend to settle towards that expected cost, the pattern described in our explainer on gambler's ruin and the law of large numbers.

A more stable check than wins and losses is the price itself. Comparing the odds you took with the final odds before the event, the idea behind closing line value, is less swayed by short-term luck than a profit figure is.

Cold runs work the same way

A losing streak feels like proof that something is broken. Sometimes it is, but often the run is just the downside of the same variance that produced an earlier winning spell. Raising stakes to "get back to normal" assumes the losses must be balanced out soon, which no game or match is obliged to do.

Common mistakes linked to ignoring it

1. Judging a change by what follows an extreme. A team sacks its coach after a terrible run, then improves. Some of that improvement would likely have happened anyway, so it is hard to credit the new coach alone.
2. Chasing last month's best performer. Picking whoever topped a short-term table often means buying at the peak of their luck.
3. Reading a "due" result into a run. Regression says extreme results are poor guides; it does not say the opposite extreme is coming next.
4. Treating a small sample as a verdict. Ten bets, five games or a single tournament rarely settle anything.

Questions readers ask

Is regression to the mean the same as the gambler's fallacy?

No. The gambler's fallacy expects a coin that landed heads several times to land tails next, as if the coin remembers. Regression to the mean makes no such prediction for any single event. It says that, on average, a performer whose last result was extreme will produce a next result closer to their usual level.

Does it apply to casino games?

Casino games are pure chance with fixed rules, so there is no skill component to stay behind when luck fades. Short sessions still swing widely, and over many rounds results move towards the expected cost built into the game.

How long is a long enough sample?

There is no single number. The noisier the outcome, the more results you need. Low-scoring sports and long-odds bets need far larger samples than high-scoring sports or short-priced selections before a pattern means much.

Understanding variance makes sport more interesting to follow, but it is not a route to income. Betting is for adults only, the rules differ by country and region, and any money staked should be money set aside for entertainment.