Attack, defence and overall ratings for 144 clubs across eight European leagues.
RESEARCH SNAPSHOT09 Oct 2026Match cutoff · 08 Oct 2026144 clubs · 8 leagues
HIGHEST RATED
INDEX BASELINE100 = model average
Original model benchmark retained
CROSS-LEAGUE DATA766 cross-league matches
Between the eight displayed leagues
THE RANKINGS
Club Rankings
Attack: higher is better · Defence: lower is better
Attack and defence are non-penalty expected goals (NPxG), not total-goal forecasts. Coverage is limited to the eight European leagues shown. Read how to interpret the ratings →
AVERAGE RATINGS
League Comparison
Average club strength in this snapshot
PREMIER LEAGUE · 2026/27
Premier League
Strength ratings, season performance and forecasts for the 2026/27 Premier League.
RESULTS THROUGHGameweek 550 completed matchesModel cutoff · 08 Oct 2026
PREMIER LEAGUE · 2026/27
Fixture Predictions
Goal forecasts and outcome probabilities for each gameweek, with completed results alongside the recorded forecasts.
CURRENT OUTLOOKGameweek 6Forecast cutoff · 08 Oct 2026
Goal forecasts are averages, not predicted exact scores. Earlier-round forecasts were reconstructed using data available before each round; they were not necessarily issued at that time. Future fixtures use the current GW5 strength snapshot and do not incorporate future lineup news.
PREMIER LEAGUE
Match Predictor
Calculate goal forecasts and win, draw and loss probabilities for two Premier League clubs.
MODEL SNAPSHOTAfter GW5400 strength draws included
MATCH SETTINGS
Select Clubs
The same calibrated Poisson conversion used in fixture forecasts. These are team-level estimates; no match-specific injuries, expected lineups or future red cards are imposed.
Predictions are currently available for Premier League clubs. The cross-league model reports NPxG strength and has not yet been calibrated for global win/draw/loss forecasts.
METHODOLOGY
Model Details
Methods, data sources, validation results and limitations for both models.
Global research · v0.3
Global ratings model
The website shows 144 clubs from the Premier League, La Liga, Bundesliga, Serie A, Ligue 1, Liga Portugal, Eredivisie and Scottish Premiership. Cross-league matches connect domestic NPxG strength, accounting for the clubs involved.
For the leagues shown, the five major-league Understat datasets are calibrated to the FotMob scale; Portugal, the Netherlands and Scotland use FotMob shot data directly.
Ratings retain the original equal-weight 204-club neutral benchmark. An index of 100 refers to that source-model average, not the average of the 144 clubs displayed.
Overall index = 100 + 20 × neutral NPxG difference.
These are NPxG rates. They are not the total-goal GF and GA shown in the Premier League goals-based view.
Premier League · v1.7
Premier League model
The Premier League tables retain the accepted model and its own 20-club benchmark. Its strength indices average 100 within the Premier League.
Opponent, venue, recency and timed red-card adjustments.
Promoted prior: 40% typical promoted club / 60% translated Championship performance, plus the accepted capped squad-value adjustment.
Prior influence fades through 38 observed PL matches.
Calibrated Poisson conversion for goal and fixture forecasts.
100,000 simulated seasons; Top Five remains the probability target.
VALIDATION
Cross-League Evaluation
The table shows the saved 2025/26 UEFA evaluation: 170 matches. Lower error is better. This is a retrospective check of the source model, not a fresh independent test or a refit after reducing website coverage. Both models condition on observed red cards.
What the PL views mean
Strength Ratings
Estimated current ability, informed by recent and historical evidence.
Goals-Based Ratings
Expected total goals for and against the same neutral PL benchmark, using the historically calibrated goal conversion.
Season Performance Score
This season’s opponent- and red-card-adjusted performance, estimated with additive ridge regression.
Model Expected Points
Retrospective points from a separate multinomial win/draw/loss model. Points above the model are a descriptive gap, not proof of luck.
Projected Points
Points already earned plus expected points from remaining matches.
Coverage and uncertainty
The website covers eight European leagues; it is not a ranking of every club worldwide.
League comparisons depend on the quantity and quality of cross-league evidence. Scotland has fewer links than the other leagues shown.
Generic entry priors outside the accepted Premier League model have limited lower-division evidence.
Displayed P10–P90 ranges are conditional model ranges, not guaranteed 80% coverage.
Provider uncertainty and transferring PL red-card effects to other leagues are not fully reflected in those ranges.
Website ranks run from 1 to 144. The ratings and their uncertainty ranges retain the source-model values; reducing website coverage has not refitted league strength or changed the benchmark. Premier League coefficients and forecasts are unchanged.
Data sources and update timing
For the eight leagues displayed, Understat supplies domestic NPxG for England, Spain, Germany, Italy and France; FotMob supplies domestic shots, lineups and cards for Portugal, the Netherlands and Scotland, plus cross-league data. ESPN supplies the original dismissal timelines. ClubElo is an external comparison, not a training input. The displayed leagues account for 16,191 usable domestic NPxG matches, including 16,147 with dismissal context, and 766 matches between those leagues. The underlying source model was fitted using the broader archive of 22,655 domestic and 795 cross-league matches. Missing xG is excluded rather than replaced with goals or zero.
This is a dated snapshot created on 9 October 2026, using the model cutoff of 8 October 2026. The latest completed PL match in the source is 20 September 2026. Ratings do not update live. Published data changes when a new model update is prepared.
How are the leagues connected?
All eight displayed leagues are connected through UEFA club competitions. Those games allow the model to compare domestic strength across leagues while accounting for the specific clubs involved. The source model also includes cup and intercontinental evidence and fits separate scoring intercepts for those contexts.
Why use ridge regression and Poisson?
Ridge is the squared-coefficient penalty that stabilises strength estimates. The strength estimator uses a Poisson-style mean-fitting objective for continuous NPxG. Separately, the forecast model uses Poisson goal distributions to calculate scoreline probabilities and simulate matches. The season-only performance score uses additive ridge; retrospective model points use a multinomial conversion.
How are the two datasets kept consistent?
The global model copies the accepted current English strength coefficients. It then applies its provider and league calibration and compares every club against the original 204-club reference. The website displays 144 of those clubs, with consecutive ranks and unchanged rating values, alongside the accepted Premier League outputs. It does not refit the models in your browser.