Luck Masquerading as Skill in Small Samples

Definition

Whether a game's final score reflects the better team or mere noise depends on how many independent scoring chances it contains, not on how skilled the players are. A sport with many scoring events per game — basketball, with eighty to one hundred possessions — lets skill average out over enough trials that the better team usually wins; a sport with few — football, with a handful of scoring drives — lets weather, crowd noise, and randomness produce upsets that look like meaningful outcomes but aren't.

In the Book

Chapter Six opens by pointing out that nothing important happens on any single football play, yet the game around it is genuinely complex — "a high-speed ballet of many, many thousands of moving parts." The chapter's real argument is that upsets — a heavy underdog beating a favorite — happen "often enough to illustrate the way everything from weather to crowd noise to mere randomness can confound even the best handicappers," and that this isn't a flaw unique to any one team but a structural property of how many scoring events a sport packs into a game. Basketball's high possession count lets a team's true quality show up reliably over a single game; baseball's and football's sparser scoring windows mean a single lucky bounce carries disproportionate weight, and a best-of-seven World Series can still turn on being "lucky enough."

Why It Matters

Any evaluation built on too few independent trials — a hiring interview, an A/B test with low traffic, a single sales quarter, one clinical case — will let noise pose as signal; the fix isn't a better evaluator, it's more independent scoring events before you trust the result.