Another reason to rethink the Save

I don’t know if you could call it a “campaign,” but I’ve lobbied for some time now against the use of the save as a useful relief pitching performance measure.  Win Probability does a far better job on two levels: 1) it does a far better job showing how much credit a reliever deserves for his performance based on the situation, and 2) it gives us a way to measure bullpen deployment, or the way the managers use relief pitchers, based on Leverage.

Last night, in perhaps the wackiest Major League Baseball game of my lifetime, the Texas Rangers scored 30 runs in the first game of a doubleheader against Leo Mazzone’s Orioles.  Here are a few quick facts about the game before I continue:

• 30 runs is an American League record for a single game
• The last time a team scored 30 runs was in 1897, so this is a record for the modern era
• The Rangers had scored just 29 runs in their previous 9 games
• They scored in only four innings, so they actually didn’t score in more innings than they did score
• They scored more runs in the final two innings of the game than in their previous six games combined
• The Rangers’ #8 and #9 hitters (Jarrod Saltalamacchia and Ramon Vazquez) drove in 7 runs each, while arguably their best player (Michael Young) drove in none
• The Rangers won the second game of the doubleheader 9-7, making it also the most runs scored in a doubleheader in the modern era
• Reliever Wes Littleton entered the game for the Rangers in the 7th (when Texas had an 11-run lead) and pitched effectively for 3 innings, picking up the save in a 27-run win

I like what ESPN’s Tim Kurkjian often says, that when you go to a baseball game, it’s possible that you’ll see something that’s never happened before.  Not only did this game show an unprecedented offensive performance, but also it included the most curious application imaginable for the save rule.

By definition, Littleton did meet the requirements for a save, and he picked up a whopping +.002 in WPA.  Still, it’s unthinkable that this situation would be remembered in history the same way as closing out a 1-run game with runners on in the 9th.

Let me suggest a new method:

Since it’s not always ideal for managers to save their best pitcher for the final inning, scrap the idea that a “save” requires the reliever to finish the game.  You could call these Effective Relief Appearances, but that might cause some acronym trouble.  Let’s just go with “EO” for Effective Outing.  The new requirements should be:

1. The reliever enters the game with a Leverage Index of 2.0 or higher, meaning that the situation was at least twice as crucial as an average situation.  This is arbitrary, but it works for me.  If someone wants to derive some kind of threshold at which the game becomes truly significant, that’s fine by me, and we can change it.
2. The pitcher’s WPA for the game is positive.

That’s all.  I’ll track it from now on and see how it works for the Braves.  In my stats, I’ll label EOs, EOCs (Effective Outing Chances, or all relief outings where the beginning LI was > 2), and EO% (EO/EOC).