Boyd's World-> Breadcrumbs Back to Omaha-> Predictability of Pitching Stats About the author, Boyd Nation

Predictability of Pitching Stats

Publication Date: January 18, 2005

The Flip Side

There's a saying going around baseball analysis circles -- "There's No Such Thing as a Pitching Prospect" -- mostly circulated by the Baseball Prospectus guys. They're not being literal with that, of course, but the point is well-taken -- trying to project a top 21-year-old pitcher is much tougher than trying to project a top 21-year-old hitter. Is that due to general unreliability, though, or due to the current (inherent? We hope not) high risk of injury for young pitchers?

Last week, I looked at the question of which individual offensive stats correlated best from season to season. This week, I'll do the same study for the pitchers, with two questions in mind: Which stats are the best predictors of future performance, and are pitching stats more or less predictable than hitting stats? The population for this week's study consists of pitchers who started at least eight games for the same Division I team in back-to-back seasons between 2002 and 2004. There are fewer of those than there are hitters, of course; I came up with a total of 559 pairs of seasons. Here are the results:

Stat   R     X

SO    0.64  1.10
IP    0.55  1.05
BB    0.54  1.00
AB    0.53  1.05
APP   0.52  0.98
GS    0.50  1.07
CG    0.50  1.10
W     0.49  1.04
BAA   0.47  0.99
ERA   0.47  0.97
H     0.45  1.04
HBP   0.43  1.01
L     0.40  1.06
R     0.39  1.02
HR    0.37  1.02
ER    0.35  1.02
2B    0.29  1.05
WP    0.29  0.98
SHO   0.26  1.10
3B    0.25  1.05
BK    0.26  0.84
SHA   0.18  1.08
SFA   0.15  0.95

R -- Correlation from year N to year N+1
X -- Multiplier from year N to year N+1

These results match up fairly well with what we would expect given DIPS theory and general thinking combined with the role of coaching decisions in determining workload. Strikeouts are the most predictable, something that continues on through at least the mid-thirties, it seems from other results, with walks up there as well. It's surprising that home runs allowed don't correlate well (and actually increase trivially in the second year); that may bear further investigation. It's quite encouraging that the average pitcher increases his K count by 10% from year to year; that implies a good development pattern.

A small side note would be that the balk correlation would seem to imply that those are not particularly under the pitcher's control, which matches with the sometimes-heard observation that the same move varies in legality from game to game. An interesting study would be to correlate balk rates by umpire or crew instead.

As to the second question, it appears that the unpredictability is due to the injury risk and not to anything inherently variable about pitching performance -- the only offensive stats that fall outside the general window represented in the pitching stats are the stolen base stats that are as much coaching policy related as talent-driven. The piece that doesn't show up, of course, due to the selection criteria of managing eight starts in consecutive seasons, is the injury risk which makes those consecutive seasons less common than we'd like.

Side Issues

I'm pretty thoroughly agnostic about the subject of changing the season calendar, but one piece of the current proposal to do so that really excites me is the notion of a common starting date. In case you missed it, the season starts on Friday, as San Diego travels to Cal Poly for a 1:50 PST start. Cal Poly has an audio feed listed on their schedule for all three games of the weekend, if you're jonesing the way I am.

Last weekend, I completed a major update to the historical conference standings archive in The Filing Cabinet, adding a large number of conference years and individual seasons along with corrections to old problems. I still have much that I can do there, most notably in pulling records from paper media guides, and I'll always welcome any input from anyone who wants to help out, but be sure to take a look at what's there now.

If you're interested in reprinting this or any other Boyd's World material for your publication or Web site, please read the reprint policy and contact me


Boyd's World-> Breadcrumbs Back to Omaha-> Predictability of Pitching Stats About the author, Boyd Nation