Brian Kenny is a smart man. His recent book is an outstanding treatise on advanced statistics. The audiobook is especially great, because the author himself narrates it, which I think should be a requirement for all baseball audiobooks. I listened to the book on a road trip last month, and I found myself nodding along with most of the book, actively agreeing with a lot of what he said.
There was one point where he lost me, though. Maybe “lost” isn’t the right word, but he made an argument that caused me to think, “Hmmmm, I don’t think I agree with that.” That was when he made the case that Joe DiMaggio actually deserved the MVP in 1941 over Ted Williams. I wasn’t confident enough to say he was wrong, but while most of the rest of the book made great logical sense to me, his argument in this case — that DiMaggio’s higher Win Probability Added justified his selection over Williams despite Williams’ 10.6-to-9.1 advantage in WAR — seemed like statistical cherry picking to me.
Now, Kenny is making the WPA argument again, this time for Baltimore Orioles closer Zach Britton as American League Cy Young Award winner. And again, I find myself thinking, “I usually agree with you, Brian, but this seems like an argument that only considers one stat to the exclusion of many others that are at least as applicable.”
The AL Cy Young race is weird. Plenty of starting pitchers have had good seasons, but no one stands out as having been great, which opens up the competition to a reliever like Britton, who has been historically good this season. If I had a vote, I’m not sure who I would choose, but I’m nearly positive it would be one of the starters. Chris Sale, Rick Porcello, Corey Kluber, and Justin Verlander all have decent cases, and I’d wrestle with the stats and flip-flop back and forth and eventually decide on some order. But Britton would not be at the top.
There are two main reasons I don’t believe Britton deserves the Cy Young this year. To be clear, I think Britton has been outstanding this year. He has been the best closer in baseball, and I don’t blame Orioles fans for disagreeing with me on this analysis. There are probably at least 20 teams in baseball that would love to have Britton as their closer.
The first reason I wouldn’t vote for Britton: He is a relief pitcher. When you’re discussing whether a pitcher should be considered for the Most Valuable Player Award, you can make the case — the number of plate appearances a position player has in a given season is roughly the same as the number of batters faced by a starting pitcher, so even though the pitcher’s value is clustered into 33 games instead of 150, there is still an equivalency. There is no such equivalency between starting pitchers and relief pitchers. They do the same job, except that a great starter does it for 220 innings while a great closer does it for 65. A relief pitcher would have to be almost impossibly better than the starter to make up for only pitching one-third as much. When we’re talking about good pitchers, more is always better.
Let me illustrate. Right now, Britton is in the conversation partly because he has not blown any saves and because he has the lowest ERA in history for a reliever. Both of those things are really tenuous. If he gives up a three-run home run today in the last game of the season, neither of those things will be true anymore. (I mentioned this on Twitter yesterday, and some Orioles fans mocked the absurdity of the grounder-inducing Britton giving up a homer, but his 0.14 HR/9 rate this season is significantly below his career rate of 0.58, so maybe he is due?) Kenny’s point in reply to me was that Britton hasn’t given up that three-run homer so he shouldn’t be punished for it, and I agree with that. But the fact that his case could be so easily derailed demonstrates just what a tiny sample we are dealing with here.
The second reason is much bigger: WPA is not a good stat for determining awards!
As a quick summary, WPA is an easy calculation. Every game situation has an associated “win expectancy” — that is, historically speaking, what are a team’s chances of winning the game at that moment in time. For every play, that win expectancy goes up or down. The difference is ascribed to the pitcher and the hitter. Here’s an example from the FanGraphs primer on the subject:
If a batter flies out on the first pitch of the game, the home team’s WE goes up from 50% to about 52%. This means that the pitcher who induced the out gets a WPA of +0.02 and the batter gets a WPA of -0.02.
A couple problems with WPA jump out immediately. If a batter hits a ball over the fence that the center fielder jumps up and robs, the batter gets a negative WPA, the pitcher gets a positive WPA, and the center fielder gets jack squat. In fact, it is impossible for a fielder to ever accrue any WPA — all changes are credited or debited to the pitcher and the hitter. With two outs and a runner on first, a strikeout looks exactly the same as a double where the right fielder makes an amazing throw to nail the runner trying to score.
Another problem: context. To quote FanGraphs again:
WPA is the ultimate context dependent statistic. You get credit based on how much your action contributes to the odds of winning, meaning a home run in a 1-1 game in the 9th is dramatically more valuable than one in a 10-1 game in the 9th. For this reason, WPA is terrific at telling the story of the game and the players who delivered in big situations. When did the winning team pull away? Who had the decisive hit? These are questions WPA can answer.
It doesn’t tell you how well a player performed, it tells you how important their performance was.
In this way, WPA has some striking similarities to using a hitter’s RBI total to justify an MVP vote. It also gives an advantage to a player whose team plays a higher percentage of close games, because a higher percentage of that player’s plate appearances will come in situations where he can actually have a big impact on the outcome of the game.
That effect is magnified when you look at WPA for pitchers. Here is the top 10 in the American League in WPA going into the final day of the season:[table "” not found /]
WPA- is the sum of all the negative WPA events in a player’s season, and WPA+ is a sum of all the positive events. Combine the two, and you have the total WPA.
Now, one thing that might jump out at you here: Half of the top 10 in WPA are relief pitchers who spent at least a good part of the season as closers. If we expanded out to the top 30, there are 18 relievers. It’s almost as if a context-dependent stat based on winning might be biased towards a player who plays exclusively late in games his team wins!
If you’re not willing to submit a ballot of Britton, Miller, Dyson, Verlander, and Sanchez — meaning you believe the top three pitchers in the league were relievers — you can’t put Britton at the top based on WPA.
Teams winning by three runs or less in the ninth inning win 92 percent of the time. It’s closer to 97 percent for a three-run lead and 85 percent for a one-run lead, so if you get handed a lot of one-run leads and you don’t blow saves, it’s really easy to rack up the WPA, just like it’s easy to rack up RBIs if you consistently come up with runners in scoring position. Stats that depend so heavily on what your teammates do shouldn’t be big factors in voting for individual awards.
Britton’s FIP this season is 2.00. His xFIP is 2.15. (Both of those stats attempt to identify what a pitcher’s ERA “should” have been based on how well he did the things he can control — strikeouts, walks, and home runs. The difference between the two is because they treat home runs differently.) If Britton’s ERA were actually 2.00, it’s safe to assume he probably would have blown a few saves. Maybe two, maybe six (Kenley Jansen has six blown saves this year to go with his 1.83 ERA). If Britton had blown three saves, his WPA would be much lower — probably right around Miller’s number.
I know we’re dealing in hypotheticals here. I am not saying Britton should be treated as if he gave up home runs at his career rate instead of at his actual rate. I’m not saying we should pretend that he blew three games. My point is that there are numerous little things that could have gone differently — in entirely believable ways — that would have made Britton a non-factor in the Cy Young race. And that much instability is only possible when you’re dealing with a sample that is too small to be meaningful. In fact, that instability in those small samples is why they aren’t meaningful.
Zach Britton has had an excellent season. He has been nearly perfect all year. But he has thrown 65.1 innings, and the value he has packed into those innings can’t compare to the value a great starting pitcher brings.