clock menu more-arrow no yes

Filed under:

Pitcher Performance with Normalized Offense

New, 20 comments

For anyone that missed the fanpost set up by elijiahdukes, LINK, Sky, Ryan and I were having fun with numbers.  This got me to thinking about probability of the Rays scoring a certain level of runs.  If we know these probabilities, then we can make a guess as to how many runs we can allow, and still get the win.

Pitcher ER IP  R/IP  Times  PROBw   Est. Wins 
Shields      2.5758      6.4945      0.3966 33      0.5082          16.772
Sonnanstine      2.9375      6.0413      0.4862 32      0.4358          13.944
Kazmir      2.1852      5.6422      0.3873          27      0.5158          13.926
Garza      2.5333      6.1560      0.4115 30      0.4632          13.895

For those that are still with me, one of the reasons I like playing with stats, is because I am intrigued by quantifying different types of pitchers.  Take the Shields v. Kazmir debate.  I love this one a) because I am a huuuge Scott Kazmir fan, and b) because they are assumed to be polar opposites of one another.  Most people look at Scotty Dangerous as the prototypical "sky is the limit" type of potential.  This is due to his good fastball, filthy slider, developing change, and it doesn't hurt that he is a lefty.  James Shields is, in a lot of ways, the opposite.  Never hyped, but always puts up consistently good numbers.  Not flashy, with his league average fastball and the guy uses changeups to strike people out, can you believe that? 

Back on subject, earlier today, I put together a chart, A. that you can find below.  Basically it is a frequency chart showing how many different games we scored a total of 0-15 runs.  The events column is 162 minus the times column to show the inverse, or how many times we scored more than that amount of runs.  Pe shows this in percent form as the probability that we would score that many runs, and R/IP breaks this down to an inning x inning basis.  Note, I did not include extra innings into my equations.  Ultimately, this tell us, using 2008 data, the probabilty of scoring a certain level of runs per game.

That gives us the offensive side, but the one or two of you that are still here are scratching your heads saying, "Well I thought this was about pitchers, you liar."  Well this is where it gets fun.  I have also gone through and collected Earned Runs, and Innings Pitched for Kazmir, Garza, Shields, and Sonnanstine on a start x start basis.  From here a R/IP column is easily created.  In fact if you go HERE you can follow along without cluttering up this limited workspace.  Now using Chart A. below we can look at certain thresholds of Runs Allowed vs. Runs Scored per inning.  For example, Kaz allowed 0 runs seven times last year.  Meanwhile, we scored more than 0 runs 95.7% of the time or 155 times.  This means that for those 7 goose eggs that Kaz put up he should have received roughly 6.7 Est. Wins.  We can do this for each threshold until we reach a point where you get to a start where there is no statistical possibility we could have won the game.  In this case it is the start where Kaz gave up 9 runs in 3 innings, incidentally, I was at this game and it was the worst experience of my life.  He allowed 3 R/IP.  The most we scored in a game all year was 1.67 R/IP.  Therefore we can now total all of our Est. Wins to reach the figure you saw on the front page, or 13.93 wins.  That is how many Scott Kazmir should have won based on his performance, normalizing our rate of scoring. 

I have done this for all the pitchers if you were smart enough to click on that link that said HERE, try that one, if you feel up to it.  Basically, this gives you yet another way to value pitchers.  It is also handy, because it shows the value of someone that is in and out of a lineup compared with a guy that is a steady performer.  I plan to work on this for the rest of the AL East tonight so if you liked this I will be attempting to post that sometime tomorrow. 



A.                               Offense
Runs Times Events Pe R/IP
0 7 155 95.7%         -  
1 16 139 85.8%    0.11
2 20 119 73.5%    0.22
3 18 101 62.3%    0.33
4 23 78 48.1%    0.44
5 24 54 33.3%    0.56
6 11 43 26.5%    0.67
7 15 28 17.3%    0.78
8 9 19 11.7%    0.89
9 3 16 9.9%    1.00
10 6 10 6.2%    1.11
11 5 5 3.1%    1.22
12 1 4 2.5%    1.33
13 2 2 1.2%    1.44
14 1 1 0.6%    1.56
15 1 0 0.0%    1.67