Historical reconstruction · Premier League 2025/26
How the forecast changed
Follow the predicted final points and league finish as new results and xG arrived. Every point uses only matches played by that date.
Reconstructed with today’s xG model, using revised historical data and the actual played schedule. These were not forecasts issued at the time. No injury or lineup information is assumed.
Nottingham Forest: predicted final points
Solid line: model estimate · Dashed line: eventual outcome · Shaded band: middle 80% of simulated finishes
The league schedule changes around postponements. The default shows each of this team’s 38 games; these are not official gameweek numbers. The model also changes after other clubs play.
Forecasts v final outcomes
How close was the forecast?
Compare predictions made with the same number of team games played. A smaller points miss is closer. These are reconstructed estimates, not forecasts published at the time.
Nottingham Forest · After 8 team games
Forecast dated 2025-10-18. Clubs reach this checkpoint on different dates; this uses the first end-of-day snapshot at exactly 8 team games.
Across 20 clubs at this checkpoint: 7.7 points average miss; 11/20 actual totals inside the model’s 80% range. Range coverage measures uncertainty, not just how close the central estimate was.
All teams · 2025/26
| Club | Preseason | After 8 | Halfway (19) | After 28 | Actual pts | Avg. miss (pts) |
|---|---|---|---|---|---|---|
| 68.1 | 75.6 | 80.9 | 80.3 | 85 | 8.8 | |
| 67.0 | 68.7 | 75.2 | 76.7 | 78 | 6.1 | |
| 49.3 | 54.8 | 56.8 | 65.7 | 71 | 14.3 | |
| 52.4 | 50.4 | 65.1 | 65.0 | 65 | 6.8 | |
| 72.5 | 72.1 | 67.0 | 65.8 | 60 | 9.4 | |
| 54.1 | 57.9 | 49.8 | 52.5 | 57 | 3.8 | |
| 47.8 | 47.1 | 49.6 | 48.1 | 54 | 5.8 | |
| 53.4 | 55.1 | 52.7 | 51.0 | 53 | 1.2 | |
| 52.9 | 51.6 | 53.8 | 57.5 | 53 | 1.7 | |
| 59.9 | 60.4 | 59.0 | 61.0 | 52 | 8.0 | |
| 47.5 | 45.5 | 50.9 | 53.0 | 52 | 3.2 | |
| 61.1 | 58.9 | 55.7 | 51.3 | 49 | 7.7 | |
| 47.5 | 49.2 | 51.3 | 52.0 | 49 | 1.8 | |
| 43.4 | 47.0 | 47.2 | 45.6 | 47 | 1.3 | |
| 55.1 | 59.0 | 55.0 | 48.5 | 45 | 9.4 | |
| 46.4 | 40.0 | 41.0 | 39.4 | 44 | 3.5 | |
| 48.8 | 51.4 | 50.0 | 41.2 | 41 | 6.8 | |
| 42.9 | 36.4 | 34.7 | 35.9 | 39 | 3.5 | |
| 38.9 | 32.8 | 28.8 | 27.4 | 22 | 10.0 | |
| 40.8 | 34.9 | 23.6 | 21.5 | 20 | 10.2 |
Average miss = mean absolute points error at preseason, 8, 19 and 28 team games. All four must be available. We exclude the final checkpoint, where almost no forecasting remains. This is a descriptive score for one team-season, not a percentage accuracy grade or proof of predictive skill against a benchmark. A mean finishing position is not a predicted exact place. Points exclude administrative deductions.
The league forecast at 2026-05-24
Ordered by average simulated finish. This row number is the order of clubs, not the club’s average finish.
| Club | Played | Predicted final pts | Average finish | Actual final pts |
|---|---|---|---|---|
| 38 | 85.0 | 1.0 | 85 | |
| 38 | 78.0 | 2.0 | 78 | |
| 38 | 71.0 | 3.0 | 71 | |
| 38 | 65.0 | 4.0 | 65 | |
| 38 | 60.0 | 5.0 | 60 | |
| 38 | 57.0 | 6.0 | 57 | |
| 38 | 54.0 | 7.0 | 54 | |
| 38 | 53.0 | 8.0 | 53 | |
| 38 | 53.0 | 9.0 | 53 | |
| 38 | 52.0 | 10.0 | 52 | |
| 38 | 52.0 | 11.0 | 52 | |
| 38 | 49.0 | 12.0 | 49 | |
| 38 | 49.0 | 13.0 | 49 | |
| 38 | 47.0 | 14.0 | 47 | |
| 38 | 45.0 | 15.0 | 45 | |
| 38 | 44.0 | 16.0 | 44 | |
| 38 | 41.0 | 17.0 | 41 | |
| 38 | 39.0 | 18.0 | 39 | |
| 38 | 22.0 | 19.0 | 22 | |
| 38 | 20.0 | 20.0 | 20 |
20,000 simulations per checkpoint. Bands describe the model’s range and can be too narrow. On-pitch points exclude administrative deductions. The opening years of the archive are training data; earlier seasons cannot yet be reconstructed with this validated setup.
Current season forecast →