Model Accuracy
Strength Score model performance across 503 predictions, updated3 Jun 2026
Model Accuracy
Total Predictions3 verified backtests
Avg Error Margin
Last Updated
Based on 503 predictions
How Close Are Predictions
Accuracy by Signal Type
How each signal type performs against actual outcomes.
Sample Predictions
Top predictions and largest deviations from the Strength Score model.
Top 10 Most Accurate
| Player | Type | Predicted | Actual | Accuracy | Delta |
|---|---|---|---|---|---|
Unknown Unknown | Market Value | €250K | €250K | 100.0% | 0.0% |
Unknown Unknown | Market Value | €250K | €250K | 99.9% | +0.1% |
Unknown Unknown | Market Value | €1.8M | €1.8M | 99.9% | -0.1% |
Unknown Unknown | Market Value | €50K | €50K | 99.9% | -0.1% |
Unknown Unknown | Market Value | €751K | €750K | 99.9% | +0.1% |
Unknown Unknown | Market Value | €799K | €800K | 99.9% | -0.1% |
Unknown Unknown | Market Value | €1.8M | €1.8M | 99.9% | -0.1% |
Unknown Unknown | Market Value | €3.5M | €3.5M | 99.9% | -0.1% |
Unknown Unknown | Market Value | €299K | €300K | 99.8% | -0.2% |
Unknown Unknown | Market Value | €1.8M | €1.8M | 99.8% | -0.2% |
Largest Deviations
| Player | Type | Predicted | Actual | Accuracy | Delta |
|---|---|---|---|---|---|
Unknown Unknown | Market Value | €294K | €200K | 53.2% | +46.8% |
Unknown Unknown | Market Value | €195K | €150K | 70.2% | +29.8% |
Unknown Unknown | Market Value | €141K | €200K | 70.7% | -29.3% |
Unknown Unknown | Market Value | €290K | €400K | 72.4% | -27.6% |
Unknown Unknown | Market Value | €297K | €400K | 74.3% | -25.7% |
No model is perfect. Understanding where predictions deviate helps contextualize confidence intervals.
Prediction Error Distribution
How far off predictions are from actual values. Centered at 0% = perfect prediction.
Accuracy Over Time
Rolling accuracy by month. Our models improve as more transfer windows complete.
How We Measure Accuracy
Holdout validation. Verified backtests use strict train-test splits. The model never sees the outcome data during training. These are labelled as"verified" throughout.
Model estimates. Where formal backtest data is not yet available, we compare our current Strength Score predictions against the latest Transfermarkt market values. These are marked as"estimates" and treated as a proxy for accuracy.
Continuous improvement. Models are retrained after each transfer window. Accuracy metrics refresh automatically as new outcome data becomes available.
Transparency by design. We show our largest deviations alongside our best predictions. No model is perfect, and understanding failure modes is as important as celebrating accuracy.
Data sources:player_forecasts (verified backtests),player_valuations_ai (model estimates)
Accuracy metrics are computed from historical backtest data and model estimates. Verified backtests use holdout data the model never saw during training. Model estimates compare current predictions against latest market values.
Accuracy data in every report
Every report includes per-league accuracy evidence with verifiable predictions and error margins.