generate_features

Module path

Data.generate_features

`generate_features_teamwins()`

Reads raw game data from the games table, computes rolling averages and derived features for each team, and writes the result to the team_game_stats table. This table is consumed directly by the ML training step.

Signature

def generate_features_teamwins(rolling_window: int = 5):

Parameters

rolling_window

integer

default:"5"

Number of prior games used to compute each rolling average. A min_periods=1 setting means rows are included even when fewer than rolling_window games exist for a team.

Returns: None Side effects:

Reads all rows from the games table in Data/nba_stats.db.
Drops and recreates the team_game_stats table (via if_exists='replace').
Prints the loaded row count and the generated row count to stdout.

Run fetch_games_teamwins() before this function so the games table is populated.

Usage example

from Data.generate_features import generate_features_teamwins

# Use the default 5-game rolling window
generate_features_teamwins()

# Use a 10-game rolling window
generate_features_teamwins(rolling_window=10)

`team_game_stats` output columns

The function creates or replaces the team_game_stats table with the following columns:

Column	Type	Description
`game_id`	`TEXT`	NBA game identifier. Primary key.
`team`	`TEXT`	Team abbreviation.
`opponent`	`TEXT`	Opponent team abbreviation.
`points_diff`	`REAL`	`team_points_roll - opponent_points_roll`. Positive values favour the team.
`elo_diff`	`REAL`	Elo rating difference. Currently always `0` (reserved for future use).
`home`	`INTEGER`	`1` if the team played at home, `0` if away.
`win`	`INTEGER`	`1` if the team won, `0` if the team lost.
`team_points_roll`	`REAL`	Rolling average of `team_points` over the last N games for this team.
`opponent_points_roll`	`REAL`	Rolling average of `opponent_points` over the last N games for this team.
`team_reb_roll`	`REAL`	Rolling average of `team_reb` over the last N games for this team.
`opponent_reb_roll`	`REAL`	Rolling average of `opponent_reb` over the last N games for this team.
`team_ast_roll`	`REAL`	Rolling average of `team_ast` over the last N games for this team.
`opponent_ast_roll`	`REAL`	Rolling average of `opponent_ast` over the last N games for this team.

Rolling averages are computed per team using groupby('team') before being joined back to the full DataFrame. Game ordering follows the natural order of rows in the games table.

API Reference

Database

Module path