Opposition analysis · Research preview
Team strength, season by season
How strong is the attack? How hard is the defence to break down? Compare opponent-adjusted ratings and inspect the evidence behind them.
Current Premier League ratings use the forecast's matched 90-match snapshot; statistics updated 2026-11-03T00:00:14.178Z. Historical and other-league endpoints are dated saved studies.
Nottingham Forest
100 = the average fitted team in this league. Higher is stronger for every rating. League scales are separate: 120 here is not necessarily the same absolute strength as 120 in another league.
2026/27 · through 2026-11-02 · 9 season games / 161 available team games before recency weighting · current captured endpoint
1520 reconstructed historical prediction tests in this league.
Dots are available season endpoints through the selected season, not a game-by-game path. Current-season dots are provisional. A rating carries earlier seasons with a 240-day half-life and shrinks small samples towards average.
What sits behind the rating?
The headline score uses opponent-adjusted xG. Actual goals, shot execution and action outcomes stay visible as separate evidence; the tested blends did not earn promotion.
Results and chances
6 / 12 goals for / against
14.2 / 9.2 xG for / against · 9 games
0.0 xGoT for · 0/9 games recorded
Value per 100 recorded actions
0.141 xT / 100 actions
0.113 VAEP / 100 actions
0.099 PV / 100 actions
9/9 games recorded. Pooled values ÷ pooled actions, not averages of match rates.
2026/27: within-league ratings
| Team | Overall | Attack | Defence | Games |
|---|---|---|---|---|
| 132.1 | 115.4 | 151.3 | 9 | |
| 120.3 | 128.5 | 112.6 | 9 | |
| 115.0 | 117.4 | 112.7 | 9 | |
| 108.0 | 114.6 | 101.8 | 9 | |
| 107.6 | 112.7 | 102.8 | 9 | |
| 107.2 | 113.8 | 101.0 | 9 | |
| 102.9 | 109.7 | 96.5 | 9 | |
| 101.5 | 108.3 | 95.1 | 9 | |
| 99.6 | 104.8 | 94.6 | 9 | |
| 97.5 | 93.3 | 101.9 | 9 | |
| 97.4 | 96.1 | 98.8 | 9 | |
| 97.2 | 94.6 | 99.8 | 9 | |
| 96.6 | 101.4 | 92.1 | 9 | |
| 95.4 | 94.8 | 96.1 | 9 | |
| 94.7 | 88.5 | 101.3 | 9 | |
| 93.8 | 89.6 | 98.3 | 9 | |
| 90.5 | 87.5 | 93.5 | 9 | |
| 87.3 | 81.4 | 93.7 | 9 | |
| 86.3 | 79.7 | 93.4 | 9 | |
| 81.9 | 85.6 | 78.4 | 9 |
How to read this
Attack 120 means 20% more expected scoring than the average fitted attack. Defence 120 means about 17% fewer expected goals conceded. Overall is the geometric mean of attack and defence. These are indices, not percentages of winning.
The model accounts for opponent strength and home advantage. The goal figures on the cards describe a neutral venue against the league-average opponent. A completed-season endpoint uses only data available by that endpoint, but the source files were captured retrospectively; these were not forecasts issued at the time.
What worked in testing? The xG reference beat goals-only across the Premier League development seasons. Adding action quality helped there, then worsened 2025/26 result prediction. The fixed 70/30 xG–goals blend had a small 2025/26 gain with an interval crossing zero. No challenger cleared the promotion gate. This is a useful reference, not a proven optimal rating.
Five completed Premier League seasons support the main test. Other leagues have shorter histories; some current-season-only ratings have no historical validation. Small samples shrink towards average. There are no calibrated confidence bands yet, and the scale resets relative to each league’s teams. Ratings cannot establish which league is stronger.
Action values describe modelled changes in scoring/conceding danger, not literal extra goals. xT values ball progression, VAEP balances scoring benefit and conceding risk, and PV uses a short time horizon. Their definitions and scales differ; they are not added together. The action denominator includes converted carries; it is not a count of touches. Provider metric definitions ↗