Who wins the Premier League?
Simulating the rest of the season…
What should count more?
| Projected finish | Wins title | Top 4 | Relegated |
|---|
Ordered by average points across 10,000 simulated seasons. Likely range is where a club finishes on points in 8 out of 10 simulated seasons. npxG is expected goals without penalties, season to date. Lines mark the top four and the bottom three.
How it works
- Rated on squad value and xG
Every club gets two ratings: its squad's market value on Transfermarkt, and its expected goals for and against per game on Understat.
- Recent form carries more weight
The rating is 70% form and 30% squad value. Early in the season, when only a few games have been played, form also uses last season's xG.
- Simulated 10,000 times
Every remaining fixture is simulated 10,000 times. Title, top-four and relegation chances are the share of simulations in which each happened.
Would it have worked?
Tested on the last 11 seasons
Each past season was predicted by a model fitted on the other seasons, using only what was known at the time.
| Predicting from | Chance given to the eventual champion | Favourite won the title |
|---|
What it doesn't know
- Injuries, suspensions, new managers and fixture congestion, except as they show up in expected goals.
- Transfers in January, until Transfermarkt's values move.
- Point deductions. The table here is points won on the pitch.
It's a toy model for exploring the question, not betting advice.
Data from Understat (fixtures, results, expected goals) and Transfermarkt (squad values), fetched once a week.