By Jason Wojciechowski on August 26, 2011 at 4:50 PM
Jeff Passan published a piece on Yahoo this week with a lot of statistical tidbits. A whole bunch of those tidbits involve pitch-type linear weights values as found at Fangraphs. For instance, he told us that R.A. Dickey's fastball had saved significantly more runs than his knuckler, strongly implying without coming right out and saying that Dickey's knuckleball was not his best pitch. ("Long live the knuckleball? The prospects don't look so good.")
A conversation involving Colin Wyers, Dustin Parkes, and Bill from The Platoon Advantage broke out regarding Passan's blameworthiness for suggesting such silly things. (Other people may have been involved, but if they were, I didn't see the posts.)
The core tweet representing Colin's views (I hope) said this:
If pitch type values without chaining are at best misleading, why get upset at Passan for being mislead?
To expand on that (it is Twitter, after all), the esteemed Mr. Wyers is referring to the fact that pitch-type linear weights values, which are calculated by finding the outcome of each pitch (strike, ball, walk, homer, double, out, whatever) and totaling up the linear weights values of those outcomes for each pitch type, are problematic from the get-go. I'm sure there are other places the issues have been pointed out, but here's a comment on Beyond the Box Score by Dan Turkenkopf, noting the selection, sequencing, and game theory concerns that these raw pitch-type values raise. Further down in that thread, Justin (jinazreds on Twitter -- I don't think he uses his last name on the internet) says, "I honestly don't know if those run values mean anything." And a quote from Fangraphs's own Dave Cameron is what started that whole thread on Beyond the Box Score, as he noted in a chat that the numbers shouldn't be used "as gospel," pointing out that they are not defense-independent.
So! You've got these numbers at Fangraphs that, due to all of these unaccounted-for factors, are, paraphrasing Colin, misleading at best. The question, then, comes down to how we apportion the blame for spreading misinformation about R.A. Dickey's fastball (among other uses of the data in Passan's column) between Passan and Fangraphs. Colin's made his view clear above.
Bill, on the other hand, wrote:
I tend to think more information is always good. We don't blame ESPN for publishing Fpct when a writer cites it.
(Fpct being fielding percentage, if it's not clear.) And Dustin:
What's bothersome to me is Passan's reach, the annual reception to the piece and that you don't have to be statistically inclined to figure out the flaws with what he's doing.
I'm going to split the difference on this one, which is why I'm writing a blog post rather than just a tweet or two. I think Fangraphs has to think long and hard about what this statistic says and what it can be used for. Is there actual value in knowing the linear-weights values of the outcomes of the pitches that major-league pitchers have thrown? Given all of the information that is not included in these values (everything mentioned a few paragraphs up), are there positive uses for this data? If those uses can be articulated, then that data should be left up so people can engage in that analysis. I am, however, unaware of such uses.
That said, I'm not leaving Passan blameless here. (I told you: splitting the difference.) Journalists are often people without any particular expertise in the area they are covering. Most science writers are not scientists. Many politics writers have not held elected office or earned PhDs in political science. Even when someone with experience does become a journalist (say, a former communications director becomes a reporter), the beat is typically broader than the person's knowledge base -- how many campaign staffers can reliably explain sampling error in polling?
What do journalists typically do about this? They learn how to talk to people who know what they're doing. When Grigori Perelman won (and declined) a Fields Medal in 2006 for his work in differential geometry, journalists called up a bunch of geometers and asked them what was up. In baseball, when a minor-league player is traded, scouts and other sources get asked to describe this player's talent and future prospects.
The rhetorical question, then, is why baseball statistics should be any different. If Jeff Passan is not up to the task of thinking about the stats he is quoting and the possible limitations and problems inherent in those stats, then perhaps, trained and experienced journalist that he is, he should call someone. Drop Tom Tango a line. Text David Appelman. Do anything but go around parroting numbers that may or may not mean what you think they mean. Even if all you end up doing is a classic politics-journalism "here's one side and here's the other, you decide," I think that's better than uncritically reporting that Edward Mujica has the most valuable splitter in the game.1
The "not up to the task" part of this paragraph is similar to the Dustin Parkes quote above, with the difference being that I'm cynical enough to think that mainstream journalists, as a class, are simply not rewarded enough for critical thinking for it to be worth their while. I don't expect Passan to look up how these values are calculated and realize on his own that, "Hey, there's a whole host of stuff these stats don't take account of that might really call their value into question. And might explain why R.A. Dickey isn't a fastball pitcher!" I do expect him to do the journalist thing and have someone else tell him what those problems are. ↩