Betting Strategy

Betting Strategy

Edge Calculation

edge = elo_probability - market_probability

A bet is recommended when:

1. elo_prob > threshold (high confidence)
2. edge > 0.05 (5%+ edge over market)

Sport Thresholds

Confidence Levels

def get_confidence(elo_prob, threshold):
    if elo_prob > threshold + 0.10:
        return "HIGH"
    elif elo_prob > threshold:
        return "MEDIUM"
    return None  # Don't bet

Lift/​Gain Analysis

Validate model quality by decile:

from lift_gain_analysis import analyze_sport

overall, current = analyze_sport('nba')

# Key metrics:
# - Lift > 1.0: Better than random
# - Top decile lift > 1.3: Strong signal
# - Gain %: Cumulative wins captured

Interpretation:

• Decile 10 (highest prob): Should have highest lift
• If lift < 1.0 in high deciles, model is miscalibrated
• Set threshold where lift starts exceeding 1.2-1.3

Kelly Criterion

Optimal bet sizing:

def kelly_fraction(win_prob, odds):
    """Calculate Kelly bet fraction."""
    # odds in decimal format (2.0 = even money)
    q = 1 - win_prob
    b = odds - 1
    return (win_prob * b - q) / b

# Use fractional Kelly (25-50%) to reduce variance
bet_fraction = kelly_fraction(prob, odds) * 0.25

Portfolio Optimization

from portfolio_optimizer import optimize_bets

# Optimize across all opportunities
optimized = optimize_bets(
    opportunities=bets,
    bankroll=1000,
    max_exposure=0.10,  # 10% max per bet
    correlation_penalty=0.5  # Reduce correlated bets
)

Bet Tracking

All bets saved to placed_bets table:

default_db.execute("""
    INSERT INTO placed_bets
    (bet_id, game_id, sport, ticker, side, amount, price, elo_prob, edge, placed_at)
    VALUES (...)
""")

CLV (Closing Line Value)

Track if we beat the closing line:

from clv_tracker import calculate_clv

clv = calculate_clv(bet_id)
# Positive CLV = consistently beating market

Files to Reference

• plugins/​portfolio_optimizer.py - Kelly and optimization
• plugins/​lift_gain_analysis.py - Model validation
• plugins/​clv_tracker.py - CLV tracking
• docs/​VALUE_BETTING_THRESHOLDS.md - Threshold analysis