Tango on Baseball Archives

© Tangotiger

Archive List

Estimating Pitch Counts (July 2, 2003)

Yet another pitch count estimator. I would offer some suggestions, such as taking out IP, and bringing in BFP. It's really silly to count a Pedro IP and a bad pitcher's IP the same. I'll be breaking out my new pitch count estimator in a few weeks I hope, so look for that...
--posted by TangoTiger at 07:23 AM EDT


Posted 8:03 a.m., July 2, 2003 (#1) - Sylvain(e-mail)
  I don't know if anybody saw this estimator, I fell on it a few weeks ago; it requires more data though (BFP), and is based on college pitchers data:

http://www.boydsworld.com/breadcrumbs/epcintro.html

Jim Boyd writes for Prospectus as well I think...

Sylvain

Posted 9:09 a.m., July 2, 2003 (#2) - Mike Moffatt (homepage)
  I had a couple problems with the analysis.

1. The author mentioned the problem with multicollinearity between Runs and Earned Runs (obviously). What about multicollinearity between Hits and Walks and Earned Runs. What about Hits, Walks, Strikeouts, and Earned Runs. The T-stats were pretty high, indicating that it's probably not a problem, but I would have liked to seen him regress each X against the other X's, so we could see the extent of their relationships.

2. The author compares R^2 values from two different functional forms (linear and non-linear). Because the functional forms are different, the comparison is meaningless.

Mike

Posted 12:51 p.m., July 2, 2003 (#3) - MAH
  Tango,

Thanks for posting the article. I very much agree that you have to measure Pedro using BFP, *not* IP. Is per-pitcher BFP data available publicly? Going how far back? Or do people try to estimate BFP? Thanks.

Posted 2:24 p.m., July 2, 2003 (#4) - bob mong
  It's really silly to count a Pedro IP and a bad pitcher's IP the same.

I don't see the problem with this. I mean, IP is just a proxy for outs, which is a possible plate appearance outcome, just like hit, walk, or strikeout. How is a groundout to 2B different, whether Pedro induces it or Lima does?

Actually, I am assuming you are right about this, since you usually are; what am I missing?

Posted 10:12 p.m., July 2, 2003 (#5) - tangotiger
  bob, you are right, and I'm wrong. He's essentially giving out 2.8 pitches / non-K out (which as we know is too low). The intercept should go through zero, though.

Posted 7:47 p.m., July 3, 2003 (#6) - bob mong
  Something else I have been wondering:

In the past, pitchers hit better, as a whole, than they do now. Should that effect our estimates of historical pitch counts?