This is a local FPL (Fantasy Premier League) points prediction pipeline.
The goal: predict each player's FPL points for the next gameweek, then rank the best options overall and by position (GK / DEF / MID / FWD). I use this for transfers, captaincy, and bench decisions.
Data comes from the public FPL-Elo-Insights repo, which I have cloned locally. That dataset includes:
- per-player stats per gameweek
- team strength / Elo ratings
- fixtures by gameweek
This repo (fpl-model) does:
-
ingest.py
Combine all historical player/gameweek rows + fixtures into one table.
Output →data_processed/training_base_raw.csv -
features_with_fixture.py
Add model features:- recent form (last GW points, avg points last 3, mins last 3, % games with 60+ mins played)
- snapshot info (price, ownership%, FPL form)
- fixture difficulty for that GW (home/away, team Elo, opponent Elo, opponent defence strength)
Output →data_processed/training_table_with_fixture.csv
-
train_with_fixture.py
Train aRandomForestRegressorto predictevent_points(FPL points).
Report train/test MAE and show top predicted scorers for the last finished GW (sanity check). -
predict_next_gw.py
Train on all completed GWs so far.
Build synthetic rows for the upcoming GW using latest form + next fixtures.
Print:- Top overall projected scorers
- Top 10 GKs / DEFs / MIDs / FWDs for the upcoming GW
→ This is what I actually use before the deadline.
-
This is not a complete and fine-tuned model yet. For example: I have not included latest data on injuries so the model doesn't know if, for eg: Martinelli is injured right now (as of Nov 2, 2025). In addition, the model does not account for injuries, performance of players in other tournaments, etc.
Upcoming GW16 (model forecast)
1. Eze (ARS, MID) 10.480
2. Barnes (NEW, MID) 10.480
3. Mitchell (CRY, DEF) 9.925
4. Rogers (AVL, MID) 9.605
5. Gibbs-White (NFO, MID) 8.980
6. Nmecha (LEE, FWD) 8.945
7. Wilson (WHU, FWD) 8.935
8. N.Williams (NFO, DEF) 8.930
9. Lacroix (CRY, DEF) 8.845
10. Anderson (NFO, MID) 8.825
11. Muñoz (CRY, DEF) 8.730
12. Andersen (FUL, DEF) 8.445
13. Martinez (AVL, GK) 8.245
14. Henderson (CRY, GK) 8.210
15. Matheus N. (MCI, DEF) 7.905
16. Enzo (CHE, MID) 7.825
17. Milenković (NFO, DEF) 7.820
18. Trossard (ARS, MID) 7.725
19. Verbruggen (BHA, GK) 7.705
20. Mykolenko (EVE, DEF) 7.670
1. Martinez (AVL, GK) 8.245
2. Henderson (CRY, GK) 8.210
3. Verbruggen (BHA, GK) 7.705
4. Areola (WHU, GK) 7.230
5. Pickford (EVE, GK) 6.735
6. Sels (NFO, GK) 4.480
7. Pope (NEW, GK) 3.245
8. Leno (FUL, GK) 3.220
9. Sánchez (CHE, GK) 2.950
10. Donnarumma (MCI, GK) 2.935
1. Mitchell (CRY, DEF) 9.925
2. N.Williams (NFO, DEF) 8.930
3. Lacroix (CRY, DEF) 8.845
4. Muñoz (CRY, DEF) 8.730
5. Andersen (FUL, DEF) 8.445
6. Matheus N. (MCI, DEF) 7.905
7. Milenković (NFO, DEF) 7.820
8. Mykolenko (EVE, DEF) 7.670
9. Savona (NFO, DEF) 7.640
10. Bassey (FUL, DEF) 6.895
1. Eze (ARS, MID) 10.480
2. Barnes (NEW, MID) 10.480
3. Rogers (AVL, MID) 9.605
4. Gibbs-White (NFO, MID) 8.980
5. Anderson (NFO, MID) 8.825
6. Enzo (CHE, MID) 7.825
7. Trossard (ARS, MID) 7.725
8. Doku (MCI, MID) 7.440
9. Barkley (AVL, MID) 7.310
10. Minteh (BHA, MID) 7.030
1. Nmecha (LEE, FWD) 8.945
2. Wilson (WHU, FWD) 8.935
3. Thiago (BRE, FWD) 6.910
4. Welbeck (BHA, FWD) 4.830
5. Richarlison (TOT, FWD) 4.080
6. Raúl (FUL, FWD) 3.760
7. Marc Guiu (CHE, FWD) 3.430
8. Bowen (WHU, FWD) 3.010
9. Enes Ünal (BOU, FWD) 2.775
10. Evanilson (BOU, FWD) 2.660