October 24, 2012
Follow me on Twitter

Football: Why Plucky Teams Win Close Games

There’s been a lot of chatter lately about character and the New England Patriots. When Tom Brady was just starting out, just as the Patriots were beginning to be a good team, they won a disproportionate number of close games. Nowadays, they lose a disproportionate number of those close games. Does this mean the Patriots lack character?

No. I think it means the Patriots are good.

Tom Brady was was a 6th-round pick, a backup pressed into service who worked out fairly well. The Patriots back then were a historically bad franchise, working its way toward respectability. Inconveniently, they’d just lost their starting quarterback, who was expected to go to the Hall of Fame.

In that era, sometimes the Patriots would get blown out, and sometimes they’d hang in and make it a close game — and, more often than not, they’d win at the end. If an inferior teams is close, that means something is wrong with the superior team’s model. Either they’ve got a mistaken assumption (e.g. they think Tom Brady is Carson Palmer) or they’ve got some injuries, or they’ve got an unexpected matchup headache. If the inferior team is already over-performing and they’re close in the fourth quarter, there’s every reason to think they’ll continue to overperform. This isn’t momentum. It’s merely the expectation that the game is likely to continue to unfold at the end as it’s been unfolding all day.

If a team is superior but the score is close, that still means something is wrong with the superior team’s model. If you’re the Patriots, you’re suppose to blow out the Jets. If it’s a one-score game in the fourth quarter, then we know that something has already gone wrong. And, if things have been going wrong, they may well continue to go wrong.

What we don’t consider here is random fluctuation. The good team dominates play, but loses three fluky turnovers in the first quarter. The football bounces oddly, and the team that should be up 35-7 at the half is, instead, down 10-28. The good team may well pull it out, but this lopsided contest is almost certain to go into the books as a "close game".

Flukes happen, but I suspect the majority of close games turn out to be close, in part, because some unexpected strength or weakness emerged. The superior team’s quarterback doesn’t feel great. The bad team has unexpected success running some sort of crazy screen. The bad team’s replacement quarterback turns out to be Tom Brady. The good team’s backup nickel back just can’t cover the bad team’s slot receiver, and there’s no one left on the bench.

Above all, selecting for close games biases your sample. If you look at “close games” when you’re not very good, you’re selecting games where things have been going unexpectedly well. If you look at “close games” when you’re a very good team, you’re selecting games when things have been going wrong. And, if you take a bunch of games where “things haven’t been going well,” it’s not surprising that you lose a disproportionate number of those games. Character has nothing to do with it.