Wednesday, August 24, 2005
Regression to the mean
A month or so ago, I was watching Baseball tonight, and the topic of discussion was the Nationals.
The comment was made that they could not stay in the top, due to their reliance on winning 1-run games. I guess that a 1-run win is not viewed as a ringing endorsement of a team's strength and prowess. Over any given season, it was stated, a team should expect to have a 50/50 split when it comes to 1-run games. At that time, the Nationals were winning too many 1-run games. The time would soon come when the statistical reality known as "regression to the mean" would kick-in, and the Nationals would have to pay the piper and start losing 1-run games.
The Nats have fallen from grace, whether is came at the expense of 1-run games or not, I can't say.
This idea of regression to the mean stayed with me though, and it reappeared in a book I was reading about intuition and decision-making. The book, called Intuition: it's powers and perils, has an entire chapter devoted to "sports intuition" and in it, the author goes after a bunch of sacred cows. Think players "get hot" and that the past few at-bats or games can help you figure out who is hot? Think again. Think the rousing half-time speaches in every football movie ever made have an impact on the game's outcome? Think again. Think a great player can explain the mechanics of his swing and why it works? Nope.
Players and teams *do* get hot. In any random sequence of events there will be streaks. If you flip a coin 100 times there will be maybe 8 heads in a row at times, but the overall percentage of heads and tails should be relatively equalized.
But according to the research cited in the book, a team is just as likely to lose after a win as they are to win again.
The critical truth is this: "we forget that exceptional performance tends to regress toward normality"
This truth has colored the thoughts I've been carrying around with me in light of the current bullpen situation. Is this downturn in bullpen efficacy the price to be paid for the previous bullpen dominance? Is it interesting, or utterly predictable that the bullpen has gone from winning games with questionable starts to losing games with amazing starts?
But I'm also confused a bit by all this research. Athletes, especially great ones are not average. What is this mean they are supposed to regress to? The same for teams, they make it sound as if the quality of the team has little to do with it. As if these games that are played are nothing more than the flipping of a coin.
The comment was made that they could not stay in the top, due to their reliance on winning 1-run games. I guess that a 1-run win is not viewed as a ringing endorsement of a team's strength and prowess. Over any given season, it was stated, a team should expect to have a 50/50 split when it comes to 1-run games. At that time, the Nationals were winning too many 1-run games. The time would soon come when the statistical reality known as "regression to the mean" would kick-in, and the Nationals would have to pay the piper and start losing 1-run games.
The Nats have fallen from grace, whether is came at the expense of 1-run games or not, I can't say.
This idea of regression to the mean stayed with me though, and it reappeared in a book I was reading about intuition and decision-making. The book, called Intuition: it's powers and perils, has an entire chapter devoted to "sports intuition" and in it, the author goes after a bunch of sacred cows. Think players "get hot" and that the past few at-bats or games can help you figure out who is hot? Think again. Think the rousing half-time speaches in every football movie ever made have an impact on the game's outcome? Think again. Think a great player can explain the mechanics of his swing and why it works? Nope.
Players and teams *do* get hot. In any random sequence of events there will be streaks. If you flip a coin 100 times there will be maybe 8 heads in a row at times, but the overall percentage of heads and tails should be relatively equalized.
But according to the research cited in the book, a team is just as likely to lose after a win as they are to win again.
The critical truth is this: "we forget that exceptional performance tends to regress toward normality"
This truth has colored the thoughts I've been carrying around with me in light of the current bullpen situation. Is this downturn in bullpen efficacy the price to be paid for the previous bullpen dominance? Is it interesting, or utterly predictable that the bullpen has gone from winning games with questionable starts to losing games with amazing starts?
But I'm also confused a bit by all this research. Athletes, especially great ones are not average. What is this mean they are supposed to regress to? The same for teams, they make it sound as if the quality of the team has little to do with it. As if these games that are played are nothing more than the flipping of a coin.
Comments:
Post a Comment
