An expert correct score prediction is, at heart, a calculation. Most of the predictions you’ll see online, from tipster sites to the tables on this one, start from an estimate of how many goals each side should score and turn it into a probability for every possible score. In its modern football form the method dates from the 1980s, it’s simple to explain, and it’s less accurate than its confident presentation suggests.
This page explains how the calculation works, tests it against 14,617 real matches and shows where it gets things wrong. The prediction table below comes from an outside provider and we show it as published. We don’t know exactly how that provider produces its numbers, and nothing on this page claims otherwise.
Popular predictions
More predictions
The standard method in plain words
The method most predictors build on is called a Poisson model, after a probability formula for counting events that happen at a steady rate. For football the events are goals.
You start with the number of goals each team is expected to score. Say the home side is expected to score 1.6 and the visitors 1.1. The formula then gives the chance of any number of goals for each side on its own. Multiply the two chances together and you get the chance of any exact score.
For 1.6 against 1.1, the likeliest scores come out like this:
| Score | Chance |
|---|---|
| 1-1 | 11.8% |
| 1-0 | 10.8% |
| 2-1 | 9.5% |
| 2-0 | 8.6% |
Notice that the likeliest score is a draw even though the home side is clearly stronger. That’s because three different home wins, 1-0, 2-1 and 2-0, share the home side’s chance, and together they come to 28.8%. The draw has one big score, 1-1, and a handful of smaller ones. That’s how a confident 49% chance of a home win can sit beside a predicted 1-1.
Does it fit real matches?
We tested the idea on every league match in our sample: 14,617 matches in 18 leagues between July 2024 and October 2026. For each league we took its average home and away goals, worked out what the method says about how often each score should appear, and compared that with what happened.
| Score | Actual | Method says |
|---|---|---|
| 0-0 | 7.0% | 6.7% |
| 1-0 | 10.2% | 10.0% |
| 0-1 | 7.6% | 8.1% |
| 1-1 | 12.1% | 12.0% |
| 2-0 | 6.6% | 7.4% |
| 2-1 | 8.9% | 9.0% |
| 1-2 | 7.5% | 7.3% |
| 2-2 | 6.0% | 5.5% |
| 3-1 | 4.3% | 4.5% |
It’s a remarkably close fit for a method this simple. Most scores are within half a point of the real figure. That’s why it has survived so long.
Where it goes wrong
The method misses in two main places. First, it understates draws. It says 24.2% of matches should end level, and 26.2% did. That gap comes from 2-2 and 0-0 being slightly more common than the formula predicts, and perhaps from goals not being independent. A team that’s behind tends to push forward and a team that’s ahead tends to sit back, so scores may be less random than the formula assumes. The data can’t prove that, but it fits.
Second, it overstates some clear wins. It gives 2-0 a 7.4% chance and 2-0 happened in 6.6% of matches.
Statisticians have proposed fixes. The best known, from Dixon and Coles in 1997, nudges the chances of 0-0, 1-0, 0-1 and 1-1. We tried it. It raised 0-0 to 7.3% and 1-1 to 12.6%, so it helped on 0-0 but pushed 1-1 further from the real 12.1%. Neither version is clearly better on this data.
How accurate can a correct score prediction be?
We ran the standard method on matches one at a time, as a prediction would be made in real life. For 7,333 league matches from September 2025 to October 2026, we rated each team using only the results played before the match, worked out the likeliest score and checked it afterwards. That was a test of the method, not of any table on this site.
The exact score was right in 12.3% of matches.
The reason it isn’t higher is that the likeliest score came out as 1-1 in 73% of matches. Because most matches are fairly even, the method keeps naming 1-1. Guessing 1-1 for every match in that sample would have been right 11.9% of the time. So the method with team ratings was only 0.4 points better than a blind guess. In the 1,958 matches where it named something other than 1-1, it was right 11.7% of the time.
That’s where the evidence points. A good correct score method picks the right score about one time in eight, and a thoughtful guess does nearly as well.
What experts can’t see
The calculation uses goals. It doesn’t know that a striker was injured on Thursday, that a defender is suspended, that the weather is bad, or that a team has nothing left to play for. Those things matter, and a good prediction should be checked against them before kick-off.
It also treats every league season as if the average team were the same as last year. New signings, a new coach or a promoted club can break the pattern for weeks.
Five rules we follow
- Treat a predicted score as the centre of a range. Neighbouring scores are nearly as likely.
- Expect misses. A hit rate of 12% means about seven misses in eight.
- Remember the draw. The method understates it by two points.
- Check team news. Calculations can’t.
- Judge a source over hundreds of matches. Fifty matches is too few.
Common mistakes
The commonest mistake is believing the word “expert”. A calculation isn’t an expert, and a correct score hit rate well above the 12% we found, over a large sample, deserves checking.
Another is trusting a single confident score. The likeliest score has about a 12% chance, and the next one is not far behind.
The third is treating a few good weekends as proof. At a one-in-eight hit rate, runs of good and bad luck are normal.
Questions about expert correct score predictions
How are correct score predictions calculated? Mostly with a Poisson model: estimate each side’s goals, then work out the chance of every score.
How accurate are they? In our test of the method, the exact score was right 12.3% of the time over 7,333 matches.
Is it better than guessing 1-1? Only slightly. Guessing 1-1 every time was right 11.9% of the time in the same sample.
Does the method predict draws well? It understates them. It said 24.2% and the real figure was 26.2%.
Who developed the method? The statistician M. J. Maher described it for football in 1982, and Dixon and Coles refined it in 1997.
Can I trust a site that claims 80% correct scores? No. Nothing in real data comes close to that.
Where can I see predictions? On the homepage, or by type: 1X2, over and under and BTTS.
These are estimates and nothing more. You must be 18 or over to bet, and you should only stake money you can afford to lose.