A.J. Burnett, Career: 54.6 average game score, 17.2 standard deviation Andy Pettitte, Career: 51.4,...

A.J. Burnett, Career: 54.6 average game score, 17.2 standard deviation
Andy Pettitte, Career: 51.4, 17.2

A.J. Burnett, Yankee: 53.3, 17.0
Andy Pettitte, Yankee, 2007-2009: 49.8, 16.3

And, for the hell of it:

C.C. Sabathia, Yankee: 58.1, 17.8

A.J. Burnett has a reputation for being wildly inconsistent, throwing three ugly innings in one start and then spinning a shutout the next. That angle has, naturally, been played up in the postseason, as the Yankees "can't be sure what they'll get" out of one of their more shiny investments from game to game.

The people saying that last bit are correct. The point they're missing, though, is that the same thing could be said for pretty much every other pitcher in baseball. Starters are inconsistent. When so much depends on the weather, the opposing lineup, and the defense, it isn't realistic to expect a guy to go out and have the same game every turn.

Comments

I actually think there is some merit to consistency, but Standard Deviation is not the way to look at it.

The games scores should be divided into bucks. What buckets exactly is the question. I have done a little work on it, but nothing great yet. I hope to revisit it this off season.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Oct 23, 2009 11:00 AM EDT reply actions

I am very much open to smarter people using a smarter methodology here

by Jeff Sullivan on Oct 23, 2009 11:26 AM EDT up reply actions

I used SD for consistency, but a couple horrible games could ruin a rather good season worth of data.

Let me see if I can find my work and throw out some numbers to work with.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Oct 23, 2009 12:27 PM EDT up reply actions

Graham suggested I look at the correlation of game score from one start to the next

Out of that idea, I started looking at |game+1 score – game score| and came up with the following test figures:

Burnett, 2009: 23.7 average change, 14.9 stdev
Pettitte, 2009: 18.5, 13.3
Sabathia, 2009: 20.8, 16.4
Maddux, 1995: 15.5, 12.9

by Jeff Sullivan on Oct 23, 2009 12:59 PM EDT up reply actions

Do you mean buckets of certain ranges of game score?

Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.

by SFiercex4 on Oct 23, 2009 11:44 AM EDT up reply actions

So my goal is eventually determine what chance a certain pitcher gives a team to win a game. I went the SD route and there were too many holes

So I began looking in to game score with much better results. The following numbers are determined from AL teams in 2008. (Colorado games were really messing things up so no NL. I haven’t determined how to correct Colorado games yet). With that information I determined that if a starter threw a GS of >57 the team had a 90% chance to win, between 57 to 37 the chances were 50-50 and below <37 there was a 10% chance for a win. This was for a pitcher that got 4.78 R/G of support. Now the Yankees would be giving more support (5.65R/G), so the values should be adjusted. about 7 points of GS for 30 and 50 being the cutoffs.

With all that here are the cutoffs for the 2 pitchers:

Burentt
20 61%
8 24%
5 15%
expected team record when he starts: 24-9 73%

Pettitte
18 56%
11 34%
3 9%
expected team record when he starts: 23.5-8.5 73%

These 2 were very similar this year. Burnett blew up a couple time more this season, but given the Yankees run support.

Let me know if you have any questions. I might be slow in responding as I have a full day ahead of me.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Oct 23, 2009 1:13 PM EDT reply actions