<$BlogRSDURL$>
Proceeds from the ads below will be donated to the Bob Wuesthoff scholarship fund.

Tuesday, May 02, 2006

The Brutal Reality Of Sample Sizes: A's 10, Angels 3

Today's game was one I glad I missed, forgetting it was a businessman's special, complete with Lackey and Kendall getting into a benches-clearing brawl. Robb Quinlan got hit by a pitch on the elbow, a player whose bat the team can hardly afford to lose at the moment. The Angels' pitching staff fell apart, assisted by a bad call on Mike Scioscia's part by installing, in turn, J.C. Romero and Chris Bootcheck in the ninth.

What I wanted to harp on, though, was something that intrigued me when I first skimmed it in Baseball Hacks, and that is that the bare minimum number of at bats required to establish a player's true level of ability is 250. Two hundred and fifty! And that's just to get a peek with a ±10% error band, and a mere 80% of the time. Looking at the team stats page for the Angels, it's evident that we'll need to get to the end of May before we're even close to the right number of at bats. Assuming the same 250 at bats works, if converted to batters faced, on the pitching side, it would appear that we need another month; even the most advanced pitcher, Lackey, has only faced 155 batters. So we have a ways to go. Small sample sizes are a lot higher than you might think; as often as I'm tempted to (and actually do) panic, the reality is a lot different.

Recap


Comments:
I've heard a bunch of good things about Baseball Hacks. Would someone with zero database/programming experience be able to understand the book?
 
I think so, but beware that it really is a heavily geek-oriented book. There's stuff in there that I don't touch on a regular basis (their R programming, for instance, to do the heavier statistical lifting).
 
Y'know, we *do* have a spare copy...
 

Post a Comment

Note: Only a member of this blog may post a comment.



Newer›  ‹Older
This page is powered by Blogger. Isn't yours?
Google

WWW 6-4-2