Skip to main content

​
Module path

prediction_ai

​
File paths

ConstantPath
DBPATHData/nba_stats.db
MODELPATHmodels/nba_model.pkl

​
train_model_teamwins()

Trains a RandomForestClassifier on the team_game_stats table and saves the model to disk. Returns: None Side effects:
  • Prints test accuracy to stdout.
  • Creates models/ directory if it does not exist.
  • Writes the trained model to models/nba_model.pkl (overwrites any existing file).
You must run generate_features_teamwins() before calling this function so the team_game_stats table exists and is populated.

​
Source

prediction_ai.py
def train_model_teamwins():
    #database connection and data loading
    
    conn = sqlite3.connect(DBPATH)
    df = pd.read_sql("SELECT * FROM team_game_stats", conn)
    conn.close()

    X = df[['points_diff', 'team_reb_roll', 'opponent_reb_roll', 'team_ast_roll', 'opponent_ast_roll', 'home']]
    y = df['win']

    #split and training
    X_train, X_test, y_train, y_test = train_test_split(
        X, y, test_size=0.3, random_state=42
    )

    model = RandomForestClassifier(n_estimators=200, random_state=0)
    model.fit(X_train, y_train)
    preds = model.predict(X_test)

    print("Test Accuracy:", metrics.accuracy_score(y_test, preds))

    os.makedirs(os.path.dirname(MODELPATH), exist_ok=True)
    pickle.dump(model, open(MODELPATH, "wb"))

​
predict_game_teamwins()

Loads the saved model and returns the predicted win probability for a single game.

​
Parameters

​
points_diff
float
required
Difference between the team’s rolling average points and the opponent’s rolling average points.
​
team_reb_roll
float
required
Team’s rolling average rebounds over the last N games.
​
opponent_reb_roll
float
required
Opponent’s rolling average rebounds over the last N games.
​
team_ast_roll
float
required
Team’s rolling average assists over the last N games.
​
opponent_ast_roll
float
required
Opponent’s rolling average assists over the last N games.
​
home
integer
required
1 if the team is playing at home, 0 if away.
Returns: float — predicted probability that the team wins, in the range [0.0, 1.0].
models/nba_model.pkl must exist before calling this function. Run train_model_teamwins() to generate it.

​
Source

prediction_ai.py
def predict_game_teamwins(points_diff, team_reb_roll, opponent_reb_roll, team_ast_roll, opponent_ast_roll, home):
    model = pickle.load(open(MODELPATH, "rb"))
    features = pd.DataFrame([{
        "points_diff": points_diff,
        "team_reb_roll": team_reb_roll,
        "opponent_reb_roll": opponent_reb_roll,
        "team_ast_roll": team_ast_roll,
        "opponent_ast_roll": opponent_ast_roll,
        "home": home
    }])

    prob = model.predict_proba(features)[0][1]
    return prob

​
evaluate_bet_teamwins()

Converts a win probability into a human-readable betting recommendation.

​
Parameters

​
probability
float
required
Win probability returned by predict_game_teamwins(). Expected range: [0.0, 1.0].
Returns: str — one of the following labels:
Probability rangeReturn value
> 0.60"Good Bet"
> 0.52"Slight Edge"
≤ 0.52"Avoid"

​
Source

prediction_ai.py
def evaluate_bet_teamwins(probability):
    if probability > 0.60:
        return "Good Bet"
    elif probability > 0.52:
        return "Slight Edge"
    return "Avoid"

Build docs developers (and LLMs) love

Get started for freeTalk to us