Sunday, July 28, 2013

#killthewin

MLB Network's Brian Kenny has been pushing a new movement through his access to both MLB Now and MLB Tonight to kill the win statistic for pitchers.  The idea is not new, though a mainstream media source has often ignored it, showing win-loss records because it's easy.  How much the network really values what Kenny says is debatable.  What's not is that more and more people are getting it.  Even Chipper Jones recently tweeted to Kenny how he was converting to the idea that there was little value in a pitcher's win-loss record.

It's clear that if you are attempting to gauge how a pitcher has performed, an individual win-loss record won't tell you much.  However, is there is a way to look at an individual game and categorically say he "won" or "lost" based more on how he pitched over run support.  There is and we can thank Bill James once again.  Game Score is a crude, but efficient way to try to find out how well the pitcher performed.  It is not without its issues and James downplayed Game Score's usefulness because of them.  In its purest form, Game Score doesn't take into account park or league effects, nor the defense behind the pitcher, or even if a run was charged to the pitcher after he left the game.  It's not that Game Score is a perfect number and some have attempted to increase its usefulness, including attempting to negate the issues previously discussed or, in this case, present a GSWL record (pronounced "Gaz-Wall").  I am going to get back to GSWL later.

How is Game Score calculated?  Using an example from the page I linked to a few lines ago, I am going to use a completely average start to go over how this number is calculated.

June 22nd at Milwaukee
Tim Hudson: 6 ING, 7 H, 2 R, 2 ER, 4 BB, 4 SO

Every game starts with 50 points.  From there we add +1 point for every out recorded (or easily 3 pts for each completed inning).  Six innings x 3 = +18 points.  We also add +2 points for each full inning commpleted after the fourth.  Hudson completed 2 innings after the fourth in this example so 2 x 2 = +4 points.  Next, we subtract -2 points for each hit surrendered.  7 hits x -2 = -14 points.  Also, let's subtract -4 points for each earned run and -2 points for any unearned run.  2 ER x -4 = -8 points.  Finally, take the sum of strikeouts minus walks.  Hudson walked as many as he struck out in this game, ending with 0.  To sum up:

+50...baseline starting point
+18...for innings
+4...for each inning after the fourth
-14..for hits allowed
-8...for earned runs allowed
0 for K - BB
50 Game Score

There are two ways to take a 50 GS. On one hand, since your score is the same when you started, you can call it average.  On the other hand, the average is actually slightly higher this season at 52.  But for all intents and purposes, let's refer to 50 as an average start.  Hudson took a 2-0 loss that game.

For this season, the worst individual start came from Paul Maholm, who received an abysmal 9 GS in a 10-0 loss against the Tigers on April 26th, also known as the Anibal Sanchez game.  The best game score was Julio Teheran's near no-hitter against the Pirates on June 5th when he finished with a 89 GS.

Game Score was been expanded to include "tough loses" and "cheap wins."  A tough loss occurrs when the pitcher had a 50 GS or better, but suffers a loss.  In this case, Hudson's June 22nd start was a tough loss.  On the other hand, a cheap win is a GS under 50 but a victory is credited to the starter.  Maholm's May 1st start against the Nationals and Hudson's June 6th start against the Dodgers both ended with a GS of 70, but the starter took the loss.  The cheapest win came from Maholm on June 22nd.  He received a 37 GS, but took the win in a 7-4 game.

Jeff Angus in the SABR column I previously linked to also suggested GSWL.  The idea was to take the simplistic Game Score and see how that related to a pitcher.  When a starter had a game score of 55 or better, the team won 73% of the time in 2007.  When the starter posted a game score of 43 or lower, the winning percentage for the team was an abysmal .244.  That left what Angus called the Game Score Tweeners.  Overall, the team won 53% of the time when the game score was between 54 and 44.  He split the remaining games right down the middle, giving half to the win column and half to the loss column.  In the event of an odd number of GS tweeners, the extra game was counted as a win to reflect the win percentage of .528 in the tweener range.  Throwing out the GS tweeners would seem like an idea, but there are problems with that.  Angus goes over why better than I can.

So, where does that leave the Braves, the focus of this blog?  In the following table, I will give the GSWL, number of GS tweeners, average GS, tough loses, cheap wins, and actual starting win-loss record for reference.


Player GSWL # Tweeners Avg GS Tough L's Cheap W's Act W-L
Mike Minor 17-4 5 59.3 3 0 10-5
Julio Teheran 13-7 6 56.1 3 0 7-5
Tim Hudson 13-8 5 53.1 3 1 8-7
Kris Medlen 11-9 3 50.6 2 1 6-10
Paul Maholm 11-9 3 48.5 3 2 9-9
Alex Wood 1-1 1 43.5 1 0 0-1

How impressed are you with Minor now?  Ace enough for you?  Here are some well-known pitchers and their GSWL.  Adam Wainwright = 17-5 with 5 tweeners.  Jordan Zimmerman = 16-5 with 3 tweeners.  Bartolo Colon = 16-5 with 7 tweeners.  Max Scherzer = 19-2 with 3 tweeners.  Cliff Lee = 15-5 with 4 tweeners.  One of the most interesting cases is Lee's teammate, Cole Hamels, who is 15-7 with 3 tweeners and an amazing 7 tough loses.  

Game Score, and by extension GSWL, will not replace other informative numbers, nor become a be-all that we should use as the most important number to look at.  There are just too many variables that the pitcher doesn't have much control of (defense and park to name a few).  However, instead of #killthewin, maybe we can keep a win-loss record that better reflects how the pitcher performed.  Not the worst compromise.  

No comments:

Post a Comment