What does a 60% probability actually mean?

By Gergely Gyüre, Edgity · Published 29 October 2026 · Updated 31 October 2026

Short answer: if a model gives an outcome 60%, then across many situations like this one, the outcome should happen about 6 times in 10, and not happen about 4 times in 10. Both halves are part of the number.

One match can’t prove a probability right or wrong

Say a team is given a 60% chance to win, and the match ends in a draw. Was the 60% wrong? Not necessarily. A 60% outcome is expected to fail four times out of ten, and this match may simply have been one of those four.

It works the other way too. If the team wins, that doesn’t prove the 60% was right. A 45% outcome comes in fairly often as well.

A probability describes a type of situation, not a single result. To judge it, you need many results.

What ten 60% outcomes look like

Imagine ten separate matches, each with one outcome rated at 60%, and assume the matches don’t affect each other. On average, 6 of the 10 come in. But “on average” hides a lot of movement:

How many of ten independent 60% outcomes come in, and how often. Pure probability (the binomial formula), not match data.

Exactly 6 out of 10 happens only about a quarter of the time (25.1%).
Somewhere between 5 and 7 happens about two times in three (66.6%).
4 or fewer happens about one time in six (16.6%), and so does 8 or more (16.7%).

So a stretch where only 4 of 10 “60% outcomes” come in is not unusual, and on its own it says little about the numbers. It is simply what 60% looks like over a short run.

More results, steadier picture

The spread narrows as the sample grows. With 100 independent 60% outcomes, the count lands between 50 and 70 about 97% of the time. With 1,000, it lands between 570 and 630 about 95% of the time. The share settles close to 60%, but over hundreds of results, not over one weekend.

How to tell whether a 60% is honest

The fair test for a probability is called calibration. Collect every outcome a model rated at around 60%, wait for the results, and count how many happened. If the model is well calibrated, the answer is close to 6 in 10. If only 4 in 10 happened, the model was too confident. If 8 in 10 happened, it was too cautious.

This is how Edgity checks its own model: not by counting wins, but by comparing what the numbers said with what actually happened, across many matches and leagues.

What 60% does not mean

It is not a promise. 60% leaves a lot of room for the other result.
It is not a recommendation. A probability says how likely something is. What, if anything, to do with it is a separate question.
It is not a huge favourite. 60% is nearer to a coin flip than to a certainty. In Edgity it carries the label “Likely”; “Most Likely” starts at 65%.

Why the same match can show different numbers

A match page shows the model’s numbers before kick-off, and they can still move until the start, because the inputs move: confirmed line-ups and late absences. A change is a recalculation with new information, not a change of mind. Once the match starts, the pre-match numbers stay as they were: they describe the match as it looked beforehand.

The takeaway

Read 60% as “this happens in most situations like this one, but far from all of them.” Judge it over many matches, not one. And when a 60% outcome fails, remember that the number already said it would, about four times in ten.