"Title probability: 72%." Anyone who sees a number like that on a team's page asks the same question: where does this number come from? An expert's opinion? A simple calculation from the league table?
Neither. The end-of-season probabilities in LigLab are calculated by playing the rest of the season's matches on a computer 10,000 times. This article walks through that calculation step by step: how it works, how reliable it is, and where it can go wrong.
Why isn't the league table enough?
The table shows the past. A leader being 5 points clear doesn't, on its own, say how close the title is. The answer depends on three things:
- How many matches are left? With 3 rounds to go, 5 points is close to certain; in the first 10 rounds it means very little.
- Who are the remaining matches against? A team that has got its hard away games behind it is not in the same position as one that still has them ahead.
- How strong are the teams? The table includes luck. A team that takes few shots and concedes a lot may be stronger than its points suggest.
The practical way to take all three into account together is to play out the rest of the season.
Step 1: score probabilities for every match
The calculation starts with every remaining match. For each match the model produces not a single result but the probability of every score: how likely 1–0 is, how likely 2–1, how likely 0–0.
Those probabilities come from the teams' attacking and defensive strength, home advantage and current form; the only source is open football data (results, shots, squads). No betting data is used. How the model is built is explained on the Method page.
Step 2: play the season once
For each remaining match, a random score is drawn according to that match's score probabilities. The favourite wins most of the time, but not always, just as in real life.
When all the matches are done, they are combined with the real results of the matches already played, and a final table comes out. Ranking follows the real league's rules: points first, then the tie-break rules.
That is one possible season. On its own it says nothing; maybe in that season a surprise team won the title.
Step 3: repeat 10,000 times
The same process is run 10,000 times. Each time the dice are rolled again and a different end of season comes out.
Then we count: in how many of these 10,000 seasons did a team finish first? If it is 7,200 of them, the title probability is 72%. The same count is made for the top four and for relegation.
So "72%" is not an opinion but a count: if we played the season 100 times from today's information, this team would win the title in about 72 of them.
Why do probabilities change every week?
Because every match played changes two things at once. The first is obvious: the table changes and the number of matches left goes down. The second is less visible: the model also updates its view of the teams' strength with the new match.
That's why a team's probability can change even when it doesn't play; its rival dropping points is enough. In big matches the jump is large: a meeting of the top two can shift both teams' probabilities noticeably in a single match.
Each team's week-by-week change is shown as a table on its own page. You can start from the league pages: Süper Lig, Premier League, La Liga, Serie A, Bundesliga.
How far can these numbers be trusted?
Whether a probability is right can't be judged from a single season. If a team we gave "72%" doesn't win the title, that doesn't show we were wrong; 28% is a probability too.
The right question is: "Of the cases where we said 70%, did about 70% actually happen?" We tested that on past seasons:
| Test | Result |
|---|---|
| Coverage | England, Spain and Italy; 14 seasons from 2012/13 to 2025/26, 840 team-seasons |
| Measurement points | Matchweeks 11, 19 and 28 of the season |
| Benchmark | Clearly beat the simple method that treats all teams as equally strong |
| Fit | Title, top-places and relegation probabilities in line with how often they happened |
A concrete example of the fit: of the teams for which the method said "84% to win the title" on average, 85% actually won it.
One note: this test was run on three leagues. We use the same method for the Süper Lig and the Bundesliga, but we haven't yet published a separately tested result for those leagues.
Limits: what can't the simulation know?
The calculation is made with today's information. It can't know some things that will happen in the future:
- Injuries and transfers. A player who arrives in January, or an injury that ends someone's season, isn't in the calculation. When it happens, it shows up in the team's strength over the following days.
- Points deductions. Not taken into account.
- Early season. In the first rounds there is little data; probabilities move fast and say less.
- Very small probabilities. Where you see "–" in the table, it means below one in a thousand. That doesn't mean "impossible"; it means it almost never happened in 10,000 tries.
In short
- A title probability is found by playing the remaining matches 10,000 times and counting the results.
- The calculation takes three things into account together: the table, the remaining fixtures and the strength of the teams.
- Probabilities are updated every day; a rival's result changes your team's number too.
- The method was tested on 14 seasons in three leagues; its probabilities matched the rates at which things actually happened.
- A probability is not a certainty: 72% also says that in 28 seasons out of a hundred someone else would be champion.
The current table for every league is on the End of season page.
Sources
- Dixon, M. J. and Coles, S. G. (1997). Modelling association football scores and inefficiencies in the football betting market. Journal of the Royal Statistical Society: Series C, 46(2).
- Maher, M. J. (1982). Modelling association football scores. Statistica Neerlandica, 36(3).
- LigLab: Method and How we work.