Power of regression to the mean

Why the Dolphins will better this year and the Chiefs will be worse

Originally Published: October 4, 2012
By Peter Keating | ESPN The Magazine

NOT TO GET too cosmic on you, but there is an unwritten law that governs the sports we watch. I see it so often that I figure it's time to step forth and reveal its secrets. I'm talking about regression to the mean, the relentless tendency for any extreme result to move closer to average when repeated over time -- the force that pulls performance charts into bell curves. Really, it's just the technical term for bouncing back or falling to earth. It explains why hitting 900 out of 1,000 free throws is harder than sinking nine of 10 and why hitters tend to hit fewer long balls after they participate in the Home Run Derby.

Understanding regression to the mean makes us smarter fans and better fantasy owners, because it gives us an appreciation for truly significant achievements as opposed to statistical noise. Once you begin to see regression to the mean, you find it everywhere, especially in pro football. The NFL's small sample size -- just 16 games per season and roughly 130 plays per game -- guarantees results that seem outstanding but are often just flukes, waiting to be counteracted by regression to the mean.

For instance, it's easy to think that close games are the ultimate test of a team's character, but they're often decided by a crazy bounce or a dropped pass -- events that are a function of luck. Don't be fooled by a team that wins (or loses) a lot of close games. Underlying ability might cause a team's record to temporarily get out of whack, but regression to the mean will correct matters in the long run. Last season, for example, the Titans went 9-7 while the Seahawks went 7-9, despite the two outscoring their opponents by nearly identical tiny margins (eight points and six points). Each club is already going through a correction this season -- in the Seahawks' case, with a little help from some friends wearing black and white.