Live Model
2026-W43-snapshots
Weekly tuning from pre-gameweek snapshots for Gameweek 4, Gameweek 5, Gameweek 6, Gameweek 7.
21 Oct 2026, 01:05Last live update11 x 4Factors x position groups
FPL Model Transparency
FPL Model: Updates and Prediction Accuracy
See what TallyKind's Fantasy Premier League model uses, how weekly evidence is reviewed, and how player projections compare with actual points from the latest validation run.
Latest Validation
2026-W43
Latest validation was applied to the live model.
21 Oct 2026, 01:05Report generated209,704Player/GW rows
Model Inputs
Advanced signals, kept explainable
The player score combines a prior points baseline with expected minutes, attacking involvement, recent form, club table context, clean-sheet context, defensive/save contribution, bonus-points signal, set pieces, market signal and conceded-risk adjustment.
Goalkeepers, defenders, midfielders and forwards now use separate fitted weight profiles, so defensive, clean-sheet, creative and goal-threat signals can matter differently by role.
Learning Loop
Bayesian-style cautious updating
The live weights act as the prior. Weekly evidence comes from public FPL outcomes. Suggested changes are treated as evidence, not automatically trusted, and only pass if validation improves enough.
This is Bayesian-style model discipline rather than a claim that every player projection is a full posterior probability distribution.
Validation
Latest actual-vs-predicted plot
This plots actual player points against the prediction made by the current model for the latest validation GW. If the latest run produced a suggested model, use the toggle to compare that view too.
Gameweek 7: actual vs predicted points
Dots closer to the diagonal were closer to the real GW score.
2.5Avg absolute miss48.1%Within 2 pts160Player rows shown
Overall Result
Suggested weights improved this test
+0.055RMSE improvement+0.092MAE improvement+0.043Correlation change
The latest report was based on pre gameweek snapshots from the Fantasy Premier League public API. It was not applied to the live model unless all quality gates passed.
Accuracy Stats
Current versus suggested weights
| Group | Rows | Live RMSE | Suggested RMSE | Live MAE | Suggested MAE | Correlation |
|---|---|---|---|---|---|---|
| Overall | 209,704 | 3.807 | 3.752 | 3.068 | 2.976 | 0.333 |
| DEF | 62,924 | 3.530 | 3.457 | 2.866 | 2.763 | 0.320 |
| FWD | 46,750 | 3.820 | 3.717 | 3.161 | 3.218 | 0.515 |
| GKP | 19,064 | 3.490 | 3.441 | 2.954 | 2.926 | -0.057 |
| MID | 80,966 | 4.069 | 4.050 | 3.199 | 3.015 | 0.196 |
Model Notes
What this page means
Why was the latest tuning not applied?
Quality gates passed using clean pre-gameweek snapshot evidence.
What will improve over time?
The weekly process captures a pre-GW snapshot, then compares it with actual points once the GW is complete. More clean snapshots should make the validation history more useful than a one-off report.
What data is stored?
The report stores public player IDs, feature vectors and aggregate metrics. Sampled manager IDs are used for evaluation but are not written to the report or snapshot.
Use the model
Run the projections with context
Use the live FPL tools, then open player score breakdowns when you want to see why a player scored well.