upgrade dependencies, set default python to 3.13

.github/workflows/attributes.yml

@@ -19,7 +19,7 @@ jobs:
 
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.12"
+          python-version: "3.13"
 
       - name: Install package
         run: python -m pip install .

.github/workflows/main.yml

@@ -32,7 +32,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.10", "3.12"]  # test oldest and latest supported versions
+        python-version: ["3.10", "3.13"]  # test oldest and latest supported versions
         runs-on: [ubuntu-latest, macos-latest]  # can be extended to other OSes, e.g. [ubuntu-latest, macos-latest, windows-latest]
 
     steps:

.python-version

@@ -1 +1 @@
-3.12
+3.13

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.12-slim-bookworm
+FROM python:3.13-slim-bookworm
 
 WORKDIR /airsenal
 COPY . /airsenal

README.md

@@ -18,8 +18,6 @@ Our own AIrsenal team's ID for the 2025/26 season is **[742663](https://fantasy.
 
 ## Installation
 
-⚠️ Due to a dependency that currently forces AIrsenal to use an older version of `jaxlib`, AIrsenal doesn't work on Python 3.13 or later.
-
 We recommend using [uv](https://docs.astral.sh/uv/) for managing Python versions and dependencies. For instructions on how to install uv, go to: https://docs.astral.sh/uv/getting-started/installation/.
 
 ### Installation from source [Recommended]

airsenal/framework/FPL_scoring_rules.py

@@ -16,6 +16,10 @@
 
 saves_for_point = 3
 
+def_cons_required = {"GK": 999, "DEF": 10, "MID": 12, "FWD": 12}
+
+points_for_def_cons = 2
+
 
 def get_appearance_points(minutes):
     """

airsenal/framework/prediction_utils.py

@@ -10,12 +10,15 @@
 import numpy as np
 import pandas as pd
 from scipy.stats import multinomial
+from sqlalchemy import and_
 from sqlalchemy.orm.session import Session
 
 from airsenal.framework.FPL_scoring_rules import (
+    def_cons_required,
     get_appearance_points,
     points_for_assist,
     points_for_cs,
+    points_for_def_cons,
     points_for_goal,
     points_for_red_card,
     points_for_yellow_card,
@@ -296,6 +299,33 @@ def get_bonus_points(
     return df_bonus[1].loc[player_id]
 
 
+def get_def_con_points(
+    player_id: int, minutes: int | float, df_def_con: tuple[pd.Series, pd.Series]
+) -> float:
+    """
+    Returns expected defensive contribution points scored by player_id when playing
+    minutes minutes.
+
+    df_def_con : Tuple containing df of average def_con_pts scored when playing at least
+    60 minutes in 1st index, and when playing between 30 and 60 minutes in 2nd index
+    (as calculated by fit_def_con_points()).
+
+    NOTE: Minutes values are currently hardcoded - this function and fit_def_con_points
+    must be changed together.
+    """
+    if minutes >= 60 and player_id in df_def_con[0].index:
+        print("Def cons points for", player_id, df_def_con[0].loc[player_id])  # DEBUG
+        return df_def_con[0].loc[player_id]
+    if (
+        minutes >= 60
+        or (minutes >= 30 and player_id not in df_def_con[1].index)
+        or minutes < 30
+    ):
+        return 0
+    print("Def cons points for", player_id, df_def_con[1].loc[player_id])  # DEBUG
+    return df_def_con[1].loc[player_id]
+
+
 def get_save_points(
     position: str, player_id: int, minutes: int | float, df_saves: pd.Series
 ) -> float:
@@ -331,6 +361,7 @@ def calc_predicted_points_for_player(
     df_bonus: tuple[pd.Series, pd.Series] | None,
     df_saves: pd.Series | None,
     df_cards: pd.Series | None,
+    df_def_con: tuple[pd.Series, pd.Series] | None,
     season: str,
     gw_range: list[int] | None = None,
     fixtures_behind: int | None = None,
@@ -457,6 +488,8 @@ def calc_predicted_points_for_player(
                     points += get_save_points(
                         position, player.player_id, mins, df_saves
                     )
+                if df_def_con is not None:
+                    points += get_def_con_points(player.player_id, mins, df_def_con)
 
             points /= len(recent_minutes)
 
@@ -479,6 +512,7 @@ def calc_predicted_points_for_pos(
     df_bonus: tuple[pd.Series, pd.Series] | None,
     df_saves: pd.Series | None,
     df_cards: pd.Series | None,
+    df_def_con: tuple[pd.Series, pd.Series] | None,
     season: str,
     gw_range: list[int],
     tag: str,
@@ -498,6 +532,7 @@ def calc_predicted_points_for_pos(
             df_bonus=df_bonus,
             df_saves=df_saves,
             df_cards=df_cards,
+            df_def_con=df_def_con,
             season=season,
             gw_range=gw_range,
             tag=tag,
@@ -679,6 +714,7 @@ def get_player_scores(
     gameweek: int,
     min_minutes: int = 0,
     max_minutes: int = 90,
+    position: str | None = None,
     dbsession: Session = session,
 ) -> pd.DataFrame:
     """
@@ -691,6 +727,16 @@ def get_player_scores(
         .filter(PlayerScore.minutes <= max_minutes)
         .join(Fixture)
     )
+    if position:
+        query = query.join(
+            PlayerAttributes,
+            and_(
+                PlayerAttributes.player_id == PlayerScore.player_id,
+                PlayerAttributes.season == Fixture.season,
+                PlayerAttributes.gameweek == Fixture.gameweek,
+            ),
+        ).filter(PlayerAttributes.position == position)
+
     df = pd.read_sql(query.statement, dbsession.connection())
 
     is_fut = partial(is_future_gameweek, current_season=season, next_gameweek=gameweek)
@@ -805,3 +851,47 @@ def fit_card_points(
     )
 
     return mean_group_min_count(df, "player_id", "card_pts", min_count=min_matches)
+
+
+def fit_def_con(
+    gameweek: int = NEXT_GAMEWEEK,
+    season: str = CURRENT_SEASON,
+    min_matches: int = 10,
+    dbsession: Session = session,
+) -> tuple[pd.Series, pd.Series]:
+    """
+    Calculate the average defensive contribution points scored by each player for
+    matches they play between 60 and 90 minutes, and matches they play between 30 and
+    59 minutes.
+    Mean is calculated as sum of all bonus points divided by either the number of
+    maches the player has played in or min_matches, whichever is greater.
+
+    Returns tuple of dataframes - first index bonus points for 60 to 90 mins, second
+    index bonus points for 30 to 59 mins.
+    """
+
+    def get_def_con_df(min_minutes, max_minutes):
+        dfs = []
+        for position in ["DEF", "MID", "FWD"]:
+            df = get_player_scores(
+                season,
+                gameweek,
+                min_minutes=min_minutes,
+                max_minutes=max_minutes,
+                position=position,
+                dbsession=dbsession,
+            ).dropna(subset="defensive_contribution")
+            df["def_con_pts"] = (
+                df["defensive_contribution"] / def_cons_required[position]
+            ).astype(int) * points_for_def_cons
+            dfs.append(df)
+
+        df = pd.concat(dfs)
+        return mean_group_min_count(
+            df, "player_id", "def_con_pts", min_count=min_matches
+        )
+
+    df_90 = get_def_con_df(60, 90)
+    df_60 = get_def_con_df(30, 59)
+
+    return (df_90, df_60)

airsenal/scripts/fill_predictedscore_table.py

@@ -26,6 +26,7 @@
     calc_predicted_points_for_player,
     fit_bonus_points,
     fit_card_points,
+    fit_def_con,
     fit_save_points,
     get_all_fitted_player_data,
 )
@@ -49,6 +50,7 @@ def allocate_predictions(
     df_bonus: tuple,
     df_saves: Series,
     df_cards: Series,
+    df_def_con: tuple[Series, Series],
     season: str,
     tag: str,
     dbsession: Session,
@@ -69,6 +71,7 @@ def allocate_predictions(
             df_bonus,
             df_saves,
             df_cards,
+            df_def_con,
             season,
             gw_range=gw_range,
             tag=tag,
@@ -86,6 +89,7 @@ def calc_all_predicted_points(
     include_bonus: bool = True,
     include_cards: bool = True,
     include_saves: bool = True,
+    include_def_con: bool = True,
     num_thread: int = 4,
     tag: str = "",
     player_model: NumpyroPlayerModel | ConjugatePlayerModel | None = None,
@@ -129,6 +133,10 @@ def calc_all_predicted_points(
         df_cards = fit_card_points(gameweek=gw_range[0], season=season)
     else:
         df_cards = None
+    if include_def_con:
+        df_def_con = fit_def_con(gameweek=gw_range[0], season=season)
+    else:
+        df_def_con = None
 
     players = list_players(season=season, gameweek=gw_range[0], dbsession=dbsession)
 
@@ -146,6 +154,7 @@ def calc_all_predicted_points(
                     df_bonus,
                     df_saves,
                     df_cards,
+                    df_def_con,
                     season,
                     tag,
                     dbsession,
@@ -172,6 +181,7 @@ def calc_all_predicted_points(
                 df_bonus,
                 df_saves,
                 df_cards,
+                df_def_con,
                 season,
                 gw_range=gw_range,
                 tag=tag,
@@ -190,6 +200,7 @@ def make_predictedscore_table(
     include_bonus: bool = True,
     include_cards: bool = True,
     include_saves: bool = True,
+    include_def_con: bool = True,
     tag_prefix: str | None = None,
     player_model: NumpyroPlayerModel | ConjugatePlayerModel | None = None,
     team_model: ExtendedDixonColesMatchPredictor
@@ -212,6 +223,7 @@ def make_predictedscore_table(
         include_bonus=include_bonus,
         include_cards=include_cards,
         include_saves=include_saves,
+        include_def_con=include_def_con,
         num_thread=num_thread,
         tag=tag,
         player_model=player_model,

pyproject.toml

@@ -9,26 +9,26 @@ authors = [
 ]
 readme = "README.md"
 license = {file = "LICENSE"}
-requires-python = ">=3.10,<3.13"
+requires-python = ">=3.10,<4"
 dependencies = [
+  "beautifulsoup4>=4.13.5",
   "bpl @ git+https://github.com/anguswilliams91/bpl-next",
-  "click>=8.1.7",
-  "pandas>=2.2.1",
-  "requests>=2.31.0",
-  "sqlalchemy>=2.0.27",
-  "tqdm>=4.66.2",
-  "dateparser>=1.2.0",
-  "prettytable>=3.10.0",
-  "beautifulsoup4>=4.12.3",
-  "platformdirs>=4.2.0",
-  "deap>=1.4.1",
-  "python-dateutil>=2.8.2",
-  "lxml>=5.1.0",
-  "html5lib>=1.1",
-  "jax==0.4.24",
-  "jaxlib==0.4.24",
-  "numpy>1,<2",
+  "click>=8.3.0",
   "curl-cffi>=0.13.0",
+  "dateparser>=1.2.2",
+  "deap>=1.4.3",
+  "html5lib>=1.1",
+  "jax>=0.6.2",
+  "jaxlib>=0.6.2",
+  "lxml>=6.0.1",
+  "numpy>=2.2.6",
+  "pandas>=2.3.2",
+  "platformdirs>=4.4.0",
+  "prettytable>=3.16.0",
+  "python-dateutil>=2.9.0.post0",
+  "requests>=2.32.5",
+  "sqlalchemy>=2.0.43",
+  "tqdm>=4.67.1",
 ]
 
 [tool.hatch.metadata]