Supercomputer
Plays the rest of the season 2,000 times and shows how often each club wins.
Loading data…
About the football Supercomputer
The Supercomputer plays the rest of the season thousands of times over and counts how often each club ends up champion. It does not predict a single outcome: it hands out probabilities. If a club wins the title in 1,400 of 2,000 simulated seasons its title chance is 70%, and that tells you far more than one name in a prediction box.
The model is the same one behind the newspaper pieces that say the supercomputer has predicted the final table. Each club gets an attack and a defence rating from the goals it has scored and conceded this season, and the expected goals of every remaining fixture come from pitting one side's attack against the other's defence, plus home advantage. Each match is then resolved with a Poisson distribution, the one that best describes how goals actually arrive in football.
How to use the Supercomputer
- 1Pick a league: it opens on whichever one is deepest into its season, because that is where the model has the most to work with.
- 2Hit run and let the 2,000 seasons play out; the bar moves in batches so the page never freezes.
- 3Read the title and relegation columns as shares of all simulations, not as one-off forecasts.
- 4Use the average points column to see where a club lands in a normal season rather than in its best or worst case.
Frequently asked questions
What does the title percentage actually mean?
The share of simulations in which that club finished first. Twenty-five per cent means it was champion in one of every four simulated seasons, not that it is going to win the league.
Why do all the clubs sometimes show similar odds?
Because the season has just started. The model only learns from goals scored and conceded in the current campaign, so with no matches or very few it cannot tell the clubs apart and spreads the odds almost evenly. A notice appears on screen when that happens.
Why does a club sometimes show more than 99%?
Because it has few matches left and a lead bigger than the points still available. In those cases the title is settled by arithmetic rather than by the model.
What does the model not account for?
Injuries, suspensions, transfers, managerial changes, the toll of midweek European football and the motivation of a club already safe. It looks only at this season's goals, which is exactly why it is in beta.
Why do the numbers shift slightly when I run it again?
Because every run draws 2,000 fresh seasons. The differences are a few tenths of a percent: that is the normal margin of a random method, and it narrows the more simulations you run.