[What follows emerged out of a debate in the comments yesterday and Monday. Rather than offer an 800-word comment of my own that would explode the bottom of the page, I thought it better to offer an article.]
Since he emerged as a big story, one thing has terrified me about the Masahiro Tanaka posting. No, it isn’t the chance that he isn’t posted in the end, or even the chance that the Cubs come up second for his services.
It’s the 24-0.
When Tanaka completed the 2013 NPB season with a “perfect record,” I knew that, once his story got going, we were going to hear regularly about just how good he was in Japan, using the 24-0 thing as a justification for that good-ness. To be clear, Tanaka was very good this year in Japan – and in recent years, too, where heaven-forbid he actually lost a few games – but his win/loss record tells us almost nothing about just how good.
Indeed, win/loss record is perhaps the most pernicious statistical dud to be thrust upon the well-meaning baseball public since the invention of the sport.*
*(CERA – catcher ERA – and RBIs are up there, too.)
It’s the worst statistic not because it doesn’t do what it says it does, or because it’s meaningless. Instead, win/loss the worst statistic because what it does is very little, and it is held up to do a whole lot. It is incredibly misleading, and it has historically been offered up as the standard bearer of how well a pitcher pitched in a given season.
Consider that, over the course of their respective careers, Felix Hernandez has amassed a record of 110 and 86 during his time in Seattle (.561 winning percentage), while his predecessor with the Mariners, Freddy Garcia, who the left the team the year before Hernandez arrived, had a record of 76 and 50 (.603). Was Garcia simply a better pitcher with the Mariners than Hernandez? Of course not – Garcia’s average season WAR in that time was in the 3.5 range, dwarfed by Hernandez’s average mark around 5.0. Given the clear failure of W/L to tell us which pitcher was better – generally in the same ballpark, pitching for the same team – over a long stretch of time, we can conclude that there is obviously a fundamental flaw in the stat.
That flaw, as expressed in this example? Le duh: the Mariners team for which Garcia pitched was one of the best teams in baseball. Hernandez’s Mariners have been one of the worst.
I find that the last remaining, modestly credible argument anyone can make for wins and losses anymore is that, “well, it’s not a perfect stat, but it still gives you a rough idea of how well a pitcher has performed. It still tells you something about the pitcher’s performance.”
Sure, the win/loss statistic tells you something about the pitcher’s performance. But what it tells you is about as useful as a stat that tells you which starting pitcher in a given game gave up fewer homers. You can’t argue that giving up fewer homers than your opponent in a game isn’t desirable, and thus it tells you something about the pitcher’s performance. But you also can’t argue that the stat would be subject to a whole lot of noise that detracts from its utility – what if the starter lasted just an inning because he got blasted by doubles or walks? What if the pitcher’s team just went on a homer binge? The flaws in the stat are immediately apparent, and the value is minimal. (I’d argue this hypothetical homer stat is probably more useful than wins/losses in terms of correlation to future performance, but I have done no research to back that up, so it remains me just puffing my chest in a parenthetical.)
In the end, win/loss record for a pitcher answers a question that wasn’t worth asking: did the pitcher happen to leave the game when his team had the lead?
It does not tell us, in any meaningful way, how well the pitcher actually pitched. Do wins and losses tell us something? Sure. Do they tell us far less than ERA, FIP, BB/9, K/9, HR/9, BABIP, LOB%, HR/FB%, and on and on? Yes.
So why continue to use it? Why fight for its utility simply because it offers the teeniest, tiniest scrap of (frequently misleading) information about a pitcher’s performance? I just don’t get it.
Actually, I do get it: the reason win/loss record has gotten so much play over the years is because of the unfortunate pairing to the words “win” and “loss,” which, at the team level, are the most important numbers in any given season. Unfortunately, a team’s “win” is not the same thing as a pitcher’s statistical “win,” and the conflation of the two has led to decades of frustration for all of the reasons discussed herein.