Metric Regression to Mean

Overview

This pillar identifies teams and players performing significantly above or below their historical baseline, flagging them as prime candidates for statistical regression. It helps you position against unsustainable 'lucky' streaks or on 'unlucky' teams due for a comeback.

What It Does

It establishes a long-term statistical mean for key performance indicators like shooting percentage or turnover margin for a team or player. Then, it compares recent short-term performance against this baseline to identify significant, unsustainable deviations. The pillar quantifies this deviation to signal the likelihood and direction of future performance correction.

Why It Matters

The betting public often overreacts to recent hot or cold streaks, creating market inefficiencies. This pillar provides a data-driven, contrarian edge by showing when a team's recent results are fueled by luck rather than a true change in skill. This allows you to fade overvalued teams and back undervalued ones before the market corrects.

How It Works

First, the system calculates a player or team's baseline performance using data from a large sample size, like the previous season or the last 50 games. It then measures the performance over a shorter, recent window, for example the last 5 games. Using standard deviation, it flags any metric that is significantly outside the established norm, generating a signal that regression to the mean is likely.

Methodology

The core calculation involves comparing a short-term moving average (e.g., 5-game SMA) to a long-term moving average (e.g., 50-game SMA) for a given metric. A Z-score is calculated: (ShortTerm_Avg - LongTerm_Avg) / StdDev(LongTerm_Data). A Z-score greater than +1.5 or less than -1.5 is typically flagged as a strong signal for negative or positive regression, respectively.

Edge & Advantage

This pillar provides a mathematical edge by systematically identifying and betting against market overreactions to statistically noisy, short-term results.

Key Indicators

• PDO (Shooting % + Save %)

  high

  In hockey, a team's combined shooting and save percentage. A value far from 1000 suggests luck.

• BABIP (Batting Avg on Balls in Play)

  high

  In baseball, measures how often non-homerun balls in play become hits. Extreme values suggest regression for pitchers or hitters.

• Turnover Margin Variance

  medium

  A team's recent turnover margin compared to their long-term average. Large swings are often unsustainable.

• Shooting Percentage Variance

  high

  Compares recent shooting efficiency (e.g., 3P% in basketball) to the established seasonal norm.

Data Sources

• Hockey-Reference

  Provides advanced hockey statistics including PDO for teams and players.

• FanGraphs

  Offers in-depth baseball analytics, including BABIP for hitters and pitchers.

• Basketball-Reference

  Source for team and player shooting percentages and historical data.

Example Questions This Pillar Answers

• → Will the Golden State Warriors shoot under 40% from three in their next game after shooting 55% in their last two?
• → Will the Boston Bruins' PDO regress towards 1000 over their next 5 games?
• → Will a pitcher with a .150 BABIP over his last three starts give up more hits in his next start?

Tags