A Fantasy Premier League team optimisation tool that uses Bayesian hierarchical modelling (PyMC) and linear programming (PuLP) to maximise expected points.
Lionel predicts Fantasy Premier League (FPL) points using a Bayesian hierarchical model, implemented with a No-U-Turn Sampler in PyMC. Then, it maximises the predicted points using linear programming as implemented in PuLP. More information about the model and its outputs can be found on this web app.
Either pip install from git:
pip install git+https://github.com/jth500/lionel.git
Or clone the repository.
Using FPLPointsModel to get expected points
import pandas as pd
import numpy as np
import arviz as az
from lionel.model.hierarchical import FPLPointsModel
player = ["player_1", "player_2", "player_3", "player_4", "player_5", "player_6"]
gameweek = [1] * len(player) + [2] * len(player)
season = [25] * len(player) * 2
home_team = ["team_1"] * len(player) + ["team_2"] * len(player)
away_team = ["team_2"] * len(player) + ["team_1"] * len(player)
home_goals = [1] * len(player) + [2] * len(player)
away_goals = [0] * len(player) + [1] * len(player)
position = ["FWD", "MID", "DEF", "GK", "FWD", "MID"] * 2
minutes = [90] * len(player) * 2
goals_scored = [1, 0, 0, 0, 0, 0] + [0, 0, 1, 1, 0, 1]
assists = [0, 1, 0, 0, 0, 0] + [1, 0, 0, 0, 1, 1]
is_home = [True, True, True, False, False, False] + [False, False, False, True, True, True]
points = [10, 6, 2, 2, 2, 2] + [6, 2, 10, 10, 2, 10]
df = pd.DataFrame({
'player': player + player,
'gameweek': gameweek,
'season': season,
'home_team': home_team,
'away_team': away_team,
'home_goals': home_goals,
'away_goals': away_goals,
'position': position,
'minutes': minutes,
'goals_scored': goals_scored,
'assists': assists,
'is_home': is_home
})
fplm = FPLPointsModel()
fplm.build_model(df, points)
fplm.fit(df, points)
# Get predicted points
fplm.predict(df, extend_idata=False, combined=False, predictions=True)
# array([8.1046909 , 4.69126912, 6.41898185, 8.3587504 , 2.86888174,6.79323443, 7.11748533, 3.33156234, 4.50141007, 6.13532871, 1.75916264, 5.61056724])
Using selector to select a Fantasy Team
import pandas as pd
from lionel.selector import NewXVSelector, XISelector
df = pd.DataFrame({
'player': [f"player_{i}" for i in range(20)],
'team_name': [f'team_{i}' for i in range(10)] * 2,
'position': ["FWD"] * 3 + ["MID"] * 8 + ["DEF"] * 7 + ["GK"] * 2,
'value': [100, 90, 55, 45] * 5,
'points_pred': np.random.normal(5, 2, 20),
'xv': [0] * 20,
'xi': [0] * 20,
'captain': [0] * 20,
})
xv_sel = NewXVSelector('points_pred')
xv_sel.build_problem(df, budget=1000)
xv_sel.solve()
xi_sel = XISelector("points_pred")
xi_sel.build_problem(xv_sel.data)
xi_sel.solve()
xi_sel.data[xi_sel.data.xv==1]
Returns the following, which obeys the position, budget, and team constraints:
| player | team_name | position | value | points_pred | xv | xi | captain |
|---|---|---|---|---|---|---|---|
| player_0 | team_0 | FWD | 100 | 4.550896 | 1 | 1 | 0 |
| player_1 | team_1 | FWD | 90 | 4.953460 | 1 | 1 | 0 |
| player_2 | team_2 | FWD | 55 | 7.780676 | 1 | 1 | 0 |
| player_3 | team_3 | MID | 45 | 4.961371 | 1 | 1 | 0 |
| player_5 | team_5 | MID | 90 | 4.548816 | 1 | 1 | 0 |
| player_7 | team_7 | MID | 45 | 3.105254 | 1 | 0 | 0 |
| player_9 | team_9 | MID | 90 | 4.897183 | 1 | 1 | 0 |
| player_10 | team_0 | MID | 55 | 5.297736 | 1 | 1 | 0 |
| player_11 | team_1 | DEF | 45 | 2.450868 | 1 | 0 | 0 |
| player_13 | team_3 | DEF | 90 | 6.899006 | 1 | 1 | 0 |
| player_14 | team_4 | DEF | 55 | 10.779401 | 1 | 1 | 1 |
| player_15 | team_5 | DEF | 45 | 3.833575 | 1 | 0 | 0 |
| player_17 | team_7 | DEF | 90 | 6.717436 | 1 | 1 | 0 |
| player_18 | team_8 | GK | 55 | 3.969914 | 1 | 0 | 0 |
| player_19 | team_9 | GK | 45 | 7.769195 | 1 | 1 | 0 |
FPL (Pre-2024/25): Vaastav
Betting: The Odds API
MIT