Baseball for the Thinking Fan

Login | Register | Feedback

You are here > Home > Primate Studies > Discussion
Primate Studies
— Where BTF's Members Investigate the Grand Old Game

Thursday, July 04, 2002

Earned Wins Percentage, A New Pitcher Rating Tool

This young man asked for some feedback.


  Although there are many different ways to evaluate a pitcher, the following
idea occurred to me after reading an article about Support-Neutral
W/L Support, by Michael Wolverton
. I read through the article, and I thought
that the idea of a stat to calculate the wins that a pitcher earned is a great
idea, and I use the SNWL numbers in evaluating pitchers. However, SNWL is extremely
complicated, and I?m not too good with math. I did get past the fact that sigma
is ?the sum of,? but higher math is not my forte.

  My name is Daniel Scotto. I am a 15 year old high school student who was interested   in creating a stat for pitcher evaluation. Last summer, I tried to work out   a formula to see which hitter is most deserving of the Most Valuable Player   award, by including a combination of OPS and winning percentage. Instead, I   was led to my   current hitter evaluation tool, XR, or extrapolated runs, by Jim Furtado.   My system was a failure, but I still wanted to try my hand at a new evaluation   of pitchers.

  The first thing that occurred to me while reading Mr. Wolverton?s article was   the fact that the wins that a pitcher gets, although important to a team, are   not exactly a very good evaluation of a pitcher?s personal performance. This   is quite obvious; it is impossible to get wins without run support. The simple   earned run average is useful as well, but it is not a good evaluation of a pitcher?s   game-to-game performance. Other stats, like walks + hits per innings pitched   (WHIP) and a K/BB ratio are extremely helpful, but they do not tell the whole   story of a pitcher.

  The main thing that I gained from Mr. Wolverton?s article was the importance   of the individual start. So the thought occurred to me that the way to determine   if a pitcher has earned himself a win would be to consider how many different   games could surmount his own. Simply put, what is the percentage of games that   a pitcher has earned a win in based on his personal innings and runs totals?

  Innings and runs are basically the only two stats which show the pitcher?s importance   to a team. In an offseason, if I was a general manager for a baseball team,   and I take a look at 3 free agent pitchers, I am interested in many different   stats: innings, ERA, W/L, K/BB, WHIP, HRs allowed, the list goes on. But instead   of examining the season as a whole, I would prefer to examine it at a game-to-game   level. I would look at every one of the starts from each of the pitchers, and   I would try to find the trend to attempt to predict his next year?s performance.

  Examining starts on a game-to-game level as described in the preceding paragraph   is truly a qualitative study. There would be no true mathematical calculation   to see which is better. If there were numbers to assign to each of these starts,   then a more accurate evaluation could be made.

  And that was the thought process that led me to my system.


  The first part of the experiment was to see the percentages of occurrence of   a particular start. This involved looking at countless old box scores and placing   only two elements of data into my Excel spreadsheet: innings, and runs. I did   this for 200 games, courtesy of the box score archives from ESPN and CBS Sportsline.   This simply converts into 400 pieces of data.

The next part of the evaluation was to group the various starts. I decided   to form a chart like the one below:

0 1 2 3 4 5 6 7+ Total
<4 0.75% 0.25% 0.50% 0.50% 1.25% 1.50% 1.00% 2.50% 8.25%
4 to 4 2/3 0.25% 0.75% 0.75% 1.00% 2.25% 1.25% 1.50% 0.25% 8.00%
5 to 5 2/3 1.75% 2.50% 3.75% 1.75% 3.75% 2.50% 2.00% 0% 18.00%
6 to 6 2/3 2.25% 6.00% 6.25% 7.50% 5.00% 3.25% 0.75% 0.25% 31.25%
7 to 7 2/3 3.50% 4.50% 6.50% 4.25% 3.75% 0.75% 0.25% 0% 23.50%
8 to 8 2/3 2.50% 1.75% 1.00% 1.75% 0.25% 0.25% 0% 0% 7.50%
9 1.75% 0.25% 0.25% 0.25% 0% 0% 0% 0% 2.50%
Total 12.75% 16% 19% 17% 16.25% 9.5% 5.5% 3.00%

The numbers in the left column are the innings; the numbers in the top row are the runs allowed.

I decided that if a pitcher gave up more than seven runs, he did not give himself any chance of victory.

The hardest part of the experiment was ordering the quality of start. Obviously, the best chance of winning is 9 innings, 0 runs. It was not as concrete for the other ones. Which was better: 6 innings, 0 runs, or 9 innings, 1 run? The thought kept occurring to me that it truly depends. I asked for various people?s opinions. The two that stood out were my sister?s opinion and my friend?s opinion: my sister said that she?d rather have 9 innings, 1 run, because there is less chance for the bullpen to blow it, but my friend said that he would rather have 6 innings, 0 runs, because he is a strong supporter of a good bullpen. So the epiphany came to me: two charts, one for innings, and one for runs. The average would be the final number, although one could look at either chart based on team need.

Below are the orders for each preference of start.

Innings Chart Runs Chart
Innings Runs Innings Runs
One could take the percentages from the first chart (percentage of occurrence of start) and then use simple subtraction to gain the value of each percentage per start.

Below are the final percentages for each start based on the calculations from the first chart. (eg: to get the innings chart percentage for 6 innings, 3 runs, you would subtract the percentage of every start better than it from 100%).

Earned Wins Percentage Values
Innings Runs Innings % Runs % AVG %


The following percentages are of the seasons of pitchers in 2001. The percentages are means and medians. The first debate will be of Johnson v. Schilling, 2001.

  • Randy Johnson (ARZ) EWP-AVG: 72.38%, EWP-MED: 81.50%
  • Curt Schilling (ARZ) EWP-AVG: 66.56%, EWP-MED: 77.00%
I personally prefer the median as the way of analyzing a season because bad start can throw off the entire average, although the average is effective as well. But my calculations show that Randy Johnson was more deserving of the Cy Young than Schilling last year; this was my original support.

Let?s examine 4 pitchers in the New York Mets rotation last year (I leave out the 5th, because it was a combination of Rick Reed and Bruce Chen). I personally am a Mets fan, and they had rather good pitching last year.

  • Al Leiter (NYM) EWP-AVG: 65.08%, EWP-MED: 77.00%
  • Kevin Appier (NYM) EWP-AVG: 59.19%, EWP-MED: 63.63%
  • Steve Trachsel (NYM) EWP-AVG: 55.20%, EWP-MED: 63.63%
  • Glendon Rusch (NYM) EWP-AVG: 47.42%, EWP-MED: 42.50%
These numbers show how little respect Al Leiter got. His consistency was almost up to par with that of Curt Schilling, and Leiter only had a record of 11-11.

A third one that came to mind immediately was the AL Cy Young Award of 2001. Although I had no one stat or point, I felt that Freddy Garcia, Mike Mussina, and Tim Hudson all were more deserving than Clemens. Let’s take a look.

  • Roger Clemens (NYY) EWP-AVG: 59.18%, EWP-MED: 63.63%
  • Freddy Garcia (SEA) EWP-AVG: 64.63%, EWP-MED: 77.00%
  • Mike Mussina (NYY) EWP-AVG: 64.39%, EWP-MED: 79.25%
  • Tim Hudson (OAK) EWP-AVG: 61.98%, EWP-MED: 77.00%

What these numbers mean to me is what I have felt all along. All three are better candidates than Clemens. However, this is only 1 stat; it would not be the lone determiner for me. My top 3 for Cy Young would have been Mussina, Garcia, and Hudson, respectively.

These are just three of the many comparisons that can be done with the new stat, earned wins percentage (EWP). It is quite easy to use, just go to CBS Sportsline and use their game logs. Record every one of a pitcher?s starts, and match the corresponding percentage with a start. Hopefully, this information will be able to help someone compare pitchers.

Questions, comments? Email Me.

Dan Scotto Posted: July 04, 2002 at 06:00 AM | 3 comment(s) Login to Bookmark
  Related News:

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. tangotiger Posted: July 04, 2002 at 12:33 AM (#605411)
You're off to a good start!

You may be interested in my win expectancy chart which can be used for many different things, including establishing win% rates based on runs and innings. I'm sure Wolverton uses something similar for some of his work. Good luck!
   2. JoeD has the Imperial March Stuck in His Head Posted: July 05, 2002 at 12:33 AM (#605430)
Good job Dan!

One thing that I think you'll have to do is adjust each start for the ballpark at a minimum. You probably also want to adjust for the quality of opponent's offense somehow. Leiter keeping up with Schilling is good, but he did it with 1/2 his starts at pitcher friendly Shea, while Schilling was doing the same thing in what's probably the 3rd best hitter's park in the NL.

You are definitely on to something here. ERA and WHIP are unduly influenced by one really bad start. Look at John Smoltz this year, that early 9-run 1-inning outing. He'll never recover from that. David Wells had a start in 1992 where he gave up 13 runs in 4 1/3 IP, and his ERA for the year was almost a run higher, despite the fact that it only cost the team 1 game.
   3. Dan Scotto Posted: July 24, 2002 at 12:37 AM (#605633)
Mr Dimino:

Thanks for the park adjustment suggestion. I looked at the SNWL park adjustment numbers, and I'm experimenting with them now. I'm hoping to calculate every start in 2003, hopefully with park adjustments. Thanks!

The List I've Used

OAKLAND A'S: -2.5%

You must be Registered and Logged In to post comments.



<< Back to main

BBTF Partner

Dynasty League Baseball

Support BBTF


Thanks to
Forsch 10 From Navarone (Dayn)
for his generous support.


You must be logged in to view your Bookmarks.


Page rendered in 0.3855 seconds
58 querie(s) executed