More on momentum and the hot hand
Posted by lukegofannon on August 16, 2008
I’ve been reading posts on the DMB and SAT forums where people have suggested that the game’s AI somehow does, or should, account for momentum, or “hotness,” from game to game or within a particular game. Here are some links, taken from The Hot Hand in Sports, to short bits, articles, and interviews about momentum, streaks, and hotness:
• Clustering Illusion – from Wikipedia
The clustering illusion refers to the tendency to erroneously perceive small samples from random distributions as having significant “streaks” or “clusters,” caused by a human tendency to underpredict the amount of variability likely to appear in a small sample of random or semi-random data due to chance.
• Clustering Illusion – from Skeptic’s Dictionary
The clustering illusion is the intuition that random events which occur in clusters are not really random events. The illusion is due to selective thinking based on a counterintuitive but false assumption regarding statistical odds.
• Momentum in the Postseason – from The Hardball Times (Studeman)
Neither a team’s overall record nor its record in September matters a whole lot.
In the 35 years from 1969 through 2004, the team with best overall winning percentage won the World Series only eight times. Let me emphasize: the team with the best regular-season record has won the World Series only 23% of the time. The winners include some of the best and best-known teams of our time: the 1970 Orioles, the Big Red Machine in 1975 and 1976, the Mets in 1986 and the 1998 Yankees.
How about the teams with the best September record? The answer is exactly the same: they won eight World Series, too. Same impact. In six of the eight examples, however, the team with the best September record was also the team with the best overall record. So there’s a lot of overlap between the two groups.
In fact, in about half of the last 35 years (17, to be exact), the team with the best regular-season record was also the team with the best record in September. Of those 17 teams, six won the World Series. Even teams that were Good and had Momentum won it all only 35% of the time.
• Streak Posts – from Inside the Book Blog
Dackle gives us some great research. Suppose you have a team that went 0-10. And in the 5-games preceding that 10-game run, they were .358. What would you expect their record to be after the 10-game run? And suppose you have a team that went 10-0, and in the 5 preceding games they were .610, what’s your expectation for the games following the streak? In short, does the hot/cold hand continue, or will teams simply revert to their previously established levels?
Then, Dackle also shows the record game-by-game, leading up to the runup. In essence, when you are hot, you are hot… until you are not. There’s no runup, there’s nothing for us to see that the run would start, nor that the run would stop.
• Momentum and Winning Streaks Essay – from Idea Jungle
I’m interested in whether the “spirit” of a team has any effect on its play. In my first analysis of this, I looked at dramatic games. These are games where a home team comes from behind to win in the bottom of the ninth. It seems clear that if team mood, self confidence and belief have any effect on subsequent play, we’d expect teams to do better immediately after these wins than they do before. That study demonstrated that there isn’t any evidence of this in past history. Teams don’t seem to do any better after a dramatic win than they do before.
In this analysis, I evaluate a different aspect of this. If mood affects performance, we’d expect teams that begin winning to have a tendency to continue winning. Similarly, teams that start losing and develop a poor attitude will continue to do so. If the sole determinant of a teams ability is its annual winning percentage, P, than the expectation of a win in any particular game would be P. But if “momentum” matters, than if we pick an arbitrary game after a win , we’d expect the probability of a win to be greater than P. Similarly, in games that follow a loss, we’d expect the probability of winning to be less than 1-P.
This leads to the following principle: if “momentum” matters, we’d expect streaks to be longer than they would by chance. A team would, in general, be more “streaky” than a model of the team that was based on purely on a series of random wins and losses with the same overall win percentage.
• Online Discussion with Tom Gilovich
Cornell University social psychologist Tom Gilovich, whose 1985 Cognitive Psychology article (with Vallone and Tversky) makes him a “founding father” of hot hand research, fittingly participated as the “guest of honor” at the first-ever online chat to be hosted by the hot hand homepage.
Questions for the chat were submitted by scholars from around the world and then forwarded to Dr. Gilovich for his responses.
