Kershaw Above Replacement

Clayton Kershaw pitched another outstanding game last night.  It got me thinking about the statheads' favorite statistic:  WAR.  Wins Above Replacement.  I hate this stat as it's entirely theoretical.  How many more theoretical wins does a player contribute to his team than a theoretical replacement level player would?  I listen to a few baseball podcasts that I like, but sometimes they'll talk about players and all they'll talk about is WAR.  For example, instead of talking about Ted Williams hitting 38 home runs, driving in 123 runs, and hitting .342 in 1946, they'll say he was a 10.9 win player in in 1946 (his best season according to WAR, but not his best season according to statistics that are based on things that actually happen in baseball games).

The good thing about starting pitchers is that they actually are replaced (although not by a "replacement level player") each day.  So let's compare the Dodgers' record during the last 10 starts for each of their starting pitchers:

Clayton Kershaw:  10-0

Zack Greinke:  4-6

Hyun-Jin Ryu:  6-4

Josh Beckett:  5-5

Dan Haren:  4-6

If they gave a replacement level pitcher ten starts, what should you expect?  Read this if you want to read more about replacement level.  Here's the important line:  "replacement level is now equal to a .294 winning percentage."  Now, the rest of the team would not be replacement level.  The rest of the Dodgers are pretty good, so let's say we'll add about .100 points to the expected winning percentage for the Dodgers with a replacement level starting pitcher.  So let's we can expect the Dodgers to win four out of ten games with a replacement level starter.  So in Kershaw's last ten starts, the Dodgers have won six more games than you would expect with a replacement level starter.  They've won four more games than they have with Hyun-Jin Ryu, five more games than they have with Josh Beckett, and six more games than they have with Zack Greinke and and Dan Haren.  Now, before you tell me that there's more to that than just Clayton Kershaw, I agree, but look at Kershaw's game log.  If you want to throw out his game against the Cardinals (7 innings, 3 earned runs, no decision), I'm fine with that.  Kershaw had something to do with that win, but he definitely wasn't the only reason for it.  But look at his other nine starts.  He's given up five runs in 70 innings in those nine games (0.64 ERA).  We can give Kershaw credit for those nine wins.  Nine wins compared to 6 with Ryu, 5 with Beckett, and 4 with Greinke, Haren, or a theoretical replacement level pitcher.

So in that ten game stretch, let's say Kershaw has contributed five more wins to the Dodgers than a theoretical replacement level pitcher would.  He's made six other starts.  The Dodgers are 3-3 in those starts.  In the three wins, Kershaw was excellent (0.46 ERA, but he didn't go more than 7 in those games).  Let's give him credit for two wins there (the Dodgers only scored more than 3 runs in one of those games and he pitched a total of 19 and two thirds, a little more than two full games).  So that gets him to seven wins above a replacement level pitcher.  The losses are a little more complicated.  In one game, he was terrible (seven runs in one and two thirds).  Let's subtract that from his total.  His other two starts were both 7 innings and 3 earned runs.  The Dodgers scored two and four in those two games.  Those aren't bad performances, but they certainly weren't great either.  The National League average ERA is 3.69.  Kershaw's ERA in those two games was 3.86.  If you have a 3.86 ERA when league average is 3.69, you're probably a little bit below .500.  I'm not going to pin either loss on Kershaw (just like I didn't give him credit for the game the Dodgers won when he pitched 7 and gave up three runs).

Where does that leave us?  Kershaw contributed seven wins above a theoretical replacement level pitcher and he contributed one loss that a theoretical replacement level pitcher wouldn't have (the Dodgers scored seven runs in that game, you should win when you score seven).  So his WAR should be 6.  What do the stats say?  Baseball Reference and ESPN have him at 4.9.  (The idea of decimal numbers here is pretty stupid.  What's 0.9 of a win?  I know, Clayton Kershaw can be partially responsible for a win and Yasiel Puig, Dee Gordon, and Adrian Gonzalez can all contribute to that same win, but for starting pitchers, it's pretty stupid.  Either you were a big reason they won, you were a big reason they lost, or you were just decent and really didn't have much to do with who won or lost.)  4.9 is not that far off from where I have him.  Let's round up and say five, but I definitely think it's too low.  Fangraphs has him at 4.1 (if you're reading this after Kershaw has pitched more games, the numbers are likely to be different).  What?  The Dodgers have only won four more games with Kershaw than they would have with a replacement level player?  No way.  In his last ten starts, he's definitely been five games better than replacement and he definitely wasn't -1 in his first six games.

That illustrates another problem I have with WAR.  You go to different sites and you get different numbers because it's all theoretical.  I can go to any site I want and find out that he has a 1.76 ERA, 0.80 WHIP (ridiculous), 11.3 strikeouts per 9, a .186 batting average against, .217 on base percentage against, and .285 slugging percentage against.  Why are those the same no matter where I look?  Because they're all based on things that actually happened in baseball games and can be calculated exactly.

I'm sure WAR is generally accurate.  Mike Trout and Troy Tulowitzki are really good, but I don't need WAR to tell me that.

KAR (Kershaw Above Replacement) > WAR.