Friday, March 21, 2008

Charlie Manuel flouts statistics (or does he?)

From this BBTF thread about this Philadelphia Inquirer article:
[Phillies Manager Charlie Manuel] will continue to consult statistics when considering matchups, but he will trust his eyes more. So what if a hitter is 0 for 6 against a particular pitcher? Manuel saw those six outs, and they were all hard-hit line drives. Eyes win. And so does the gut.
This is totally illustrative of the inability of the mainstream to grasp what statistical analysis is. Manuel is here portrayed as flouting statistics to go with his gut feeling, but in reality he is demonstrating an implicit understanding of proper use of statistical analysis. 0 for 6 tells you absolutely nothing. No self-respecting statistician would try to draw any sort of conclusion ever on the basis of 6 trials.

Now, you shouldn't actually use the fact that your guy hit the ball hard six times either. That's just another statistic. However, I suspect that what Manuel is really saying here is "I know based on my vast experience observing baseball players that my player is well equipped to deal with this pitcher, I just don't have the numbers to back it up yet." And you know what? That's what you have to do in that situation.

Now if you find that your opinion is constantly leading you to choose hitters for specific situations who are generally inferior, it may be time to question your baseball experience. It should take a lot of experience to choose Neifi Perez over Albert Pujols, for example. Pujols should be your choice 99.9999% of the time, not because he has performed better or worse in a specific situation a small number of times, but because he has performed well in a general sense many, many times. Therefore, the number of situations in which Perez is a better hitter must be exceedingly small (or non-existent). If it were not, the general gap between them would be smaller as hitters.

People who do not understand statistical analysis are the ones who constantly abuse it. The people who think stats are garbage and say that "spreadsheets don't play baseball" or that "players aren't stat generating robots" are the same people who turn around and use small samples both as the tools of confirmation bias and as a straw man to attack those number-loving geeks.

For the love of God, people, please stop using small samples for any reason whatsoever.

** EDIT ** Changed use of the word "flaunt" to "flout" because I care about the English language.

1 comment:

rklllama said...

I agree mostly with what you have to say about sample sizes, but both Melky and Abreu were one for one with two walks yesterday, which can only mean that Melky will become a second Bobby Abreu!