AI summary by TwinMind
The Z Factor - Numberphile
Numberphile
22m 35s
Introduction to Z-Factor and Normal Distribution
- Z-Factor uses statistical distribution (normal distribution) to compare teams across different eras
- Normal distribution = bell-shaped curve (Gaussian) with center at mean (μ)
- Standard deviation (σ) measures spread of data
- Mathematical function: 1/(√2π σ) × e^(-1/2σ² × (x-μ)²)
- Area under curve = probability (total area = 1)
- CDF (Cumulative Distribution Function) = integrate to find probability up to point
- PDF (Probability Density Function) = differentiate CDF
Basketball Application and Methodology
- Apply normal distribution to NBA season win data
- Teams cluster around average wins, with champions at high end, poor teams at low end
- Normal distribution model works well for sports with many games
- Create separate distributions for different seasons (96-97 Bulls vs 2015-16 Warriors)
- Challenge: comparing teams from different eras with different competition levels
Standardized Normal Distribution (Z-Factor)
- Convert any normal distribution to standardized normal (mean=0, standard deviation=1)
- Formula: Z = (x - μ)/σ
- Z-Factor = (data point - season mean) ÷ season standard deviation
- Removes era differences by reducing everything to same distribution
- Standard normal PDF: 1/√2π × e^(-Z²/2)
Z-Factor Interpretation and Probability
- ±1σ: 34.13% probability each side
- ±2σ: additional 13.59% each side
- ±3σ: additional 2.14% each side
- Beyond 3σ: only 0.13% (extremely rare)
- Z > 2: very good teams, likely league winners
- Z > 3: generational performance (1 in 800 teams, ~1 in 40 years for 20-team league)
- Never seen Z > 4 across all sports analyzed
Basketball Results: Bulls vs Warriors
- 1996-97 Chicago Bulls: Z-factor = 2.06
- 2015-16 Golden State Warriors: Z-factor = 2.54
- Mathematical conclusion: Warriors more dominant relative to competition
- Warriors set regular season wins record but lost championship
- 2016-17 Warriors (championship winners) had lower Z-factor due to fewer regular season wins
Statistical Limitations and Considerations
- Analysis based only on number of wins, not margin of victory
- Doesn't account for playoff performance or head-to-head matchups
- Choice of data metric affects conclusions
- Need minimum 20 teams (preferably 30) for reliable analysis
- Need sufficient games (80+ for NBA, not 16 like NFL)
Other Sports Z-Factor Examples
Premier League Football
- Liverpool 2020: Z = 2.62 (highest found)
- Arsenal Invincibles (undefeated season): Z = 2.53
- Man City Centurions (100 points): Z = 2.50
- Man United 1992-93: Z = 2.59 (lost 6 games but weak competition)
La Liga Spain
- Real Madrid 2011-12: Z = 2.85 (highest for any football team, 102 points)
- Barcelona 2012-13 (Messi's record season): Z = 2.66
Baseball
- 1906 Chicago Cubs: 116 wins, 76.3% win rate, Z = 2.05
- 2001 Seattle Mariners: 116 wins, Z = 2.68
- Despite Cubs having higher win percentage, Mariners faced tougher competition
Methodology and Applications
- Can apply to any sport/statistic: goals, home runs, three-pointers, etc.
- Formula: Z = (individual performance - mean) ÷ standard deviation
- Enables cross-sport comparisons (e.g., Jordan's three-pointers vs Bradman's cricket runs)
- Measures how much of a "surprise" or outlier a performance was
- Standard deviation accounts for era-specific competition strength
Data Requirements
- Large dataset (minimum 20 data points, preferably 30+)
- Sufficient sample size for team performance evaluation
- Clear definition of measurement metric
- Historical data across multiple seasons/eras
