from Amazon
Monday, July 02, 2007
Another article I have a problem with…
This time from a sabermetric web site. Where are their editors?
Here is a copy of an e-mail I sent to the author of this article. Actually a series of articles. My criticism applies to all of them collectively.
As with any other splits, first you have to determine whether any of the fluctuations you see among umpires are more than just random noise, then perhaps you can start talking about who you would rather see (in the future) if you are the home or visiting team. In fact, without doing such an analysis (whether there is any real bias among umpires or you are just seeing random noise), why even bother to report the data? I can also report umpire (or player) data for weekends versus weekdays, odd days versus even days, etc. What would be the point? The only thing that makes splits interesting is when they are based on a true talent or bias. Just looking at the variation among players or numbers does not tell us whether there is a true talent or bias. For that you have to compare the distribution of the numbers with that expected by chance or do a regression from one time period to another. My guess is that you will find little if any true home/road team bias among umps. But I could be wrong. Either way, you can’t tell from the data, and to report the data and imply that it is because of a bias (which clearly you are in the article) is misleading from a sabermetric point of view, with all due respect.