Power ratings reduce team strength to a single number, letting you predict the expected spread for any matchup. This systematic approach removes emotion and bias from your betting decisions.
Step 1: Choose Your Methodology
Simple Margin-Based Ratings
The most accessible approach:
- Assign every team a starting rating of 100
- After each game, calculate the margin of victory
- Adjust both teams: winner gains points, loser loses points
- Scale the adjustment by opponent strength — beating a strong team earns more points
Adjustment formula: Rating change = (Margin of Victory / K-factor) × Opponent Strength Modifier
A typical K-factor of 20 works well for NFL; use 30-40 for football leagues with more variance.
Advanced: Regression-Based Ratings
For more sophisticated models, use linear regression with variables like:
- Offensive and defensive efficiency (points per possession)
- Turnover differential
- Third-down/set-piece conversion rates
- Pace-adjusted statistics
Step 2: Add Home Advantage
Home advantage varies by sport and league:
- NFL: Approximately 2.5-3 points (declining in recent years)
- Premier League: Approximately 0.3-0.4 goals
- NBA: Approximately 3-3.5 points
Add this factor to the home team's rating when calculating predicted spreads.
Step 3: Generate Predicted Spreads
For any matchup: Predicted Spread = Home Team Rating - Away Team Rating + Home Advantage
Example: Home team rated 108, away team rated 103, home advantage 3 points. Predicted spread = 108 - 103 + 3 = Home team -8
If the bookmaker line is Home -5.5, your model sees 2.5 points of value on the home side. A £30 bet at odds of 1.91 returns £57.30 if the home team covers.
Step 4: Validate and Refine
Track your predicted spreads against actual results over a full season. Key metrics:
- Mean Absolute Error: How far off are your predictions on average?
- Cover Rate: When your spread disagrees with the market, how often does your side cover?
- Yield: What is your return on investment when betting your model's recommendations?
Step 5: Update Consistently
Apply a recency weight so recent games matter more. A common approach weights each game's influence by 0.95^(weeks_ago), meaning last week's game counts at 100%, two weeks ago at 95%, a month ago at 81%, and so on. This keeps your ratings responsive to form changes without overreacting to single results.