Signal vs Noise

Your Team's Record in Close Games Is Luck, Not Clutch

The same roster that goes 6-1 in one-score games routinely goes 3-4 in them a year later. If you want to know who a team really is, look at its blowouts.

Every fan base runs the same argument in December. One side says the team knows how to win close games; the other says it cannot finish. Both sides believe close-game results reveal character. The record keeps a different score.

A team's win rate in games decided by one score carries over from one season to the next at a correlation of about +0.07, which is statistical language for basically not at all. The count of fourth-quarter comeback wins repeats at +0.03. For scale, a genuinely stable trait, like how often a quarterback takes off and runs, repeats at around +0.80. Clutch, as a team trait that persists, does not show up at this level of football.

The reason is arithmetic, not character. A one-score outcome turns on a handful of the highest-leverage bounces in the sport: a tipped pass, a fumble roll, a missed kick, a replay review. Those are close to coin flips even for good teams, and a season only deals eight or nine one-score games, far too few for a modest real edge to beat the noise even if one exists.

A cardiac-kids season is mostly a run of coin flips landing the same way. Enjoy it. Just do not project it.

The tell is the blowouts

The flip side is the useful half. Take a team's midseason record and throw away every close game, keeping only results decided by nine or more points. That blowout-only record, built from roughly half the games, predicts the rest of the season about as well as the full record does, at a correlation of about 0.42 versus 0.41, and it has pointed the right way in every season since 1999.

A nine-point margin is bigger than any single bounce can explain, so each decisive result is a high-purity reading of how good a team actually is. Each one-score result is roughly half luck. Half the schedule, filtered to the decisive games, carries the whole signal; the close games mostly stack a coin-flip ledger on top of it.

The caveats

Two of them. First, good teams appear in fewer close games to begin with, because their games become blowouts, and that filtering pushes the year-to-year correlation down; the data cannot prove that no clutch skill exists anywhere, only that none survives measurement at the team level. Second, close wins absolutely count. They correlate with the same season's record just fine. They just do not predict the next one.

So when your 10-3 team got there on a 6-1 run in one-score games, the projection worth trusting is the one built on its point margins, not its poise. Expect the record, not the magic, to regress. And when the neighboring fan base calls its team fragile after three narrow losses, you now know the kindest and most accurate reply: they were probably just unlucky, and next year the coin gets flipped again.