Jonah Keri decided to take on the Angels at Grantland, hoping to suggest a solution to the question of why the Orange County squad has beat its Pythagorean record every year for eight years running (assuming they hold to their three-game gap this year).

Keri proposes two theories. First, perhaps the team's excellent defense has created a cascade effect, allowing starters to pitch longer, thus keeping the bullpen fresher, thus (presumably? This isn't really spelled out) permitting the Angels to win more close games than you'd expect from the mere coin-flip that we normally expect close games to result in.

Cool idea, right? So first Keri establishes that Angels starters have, in fact, pitched deeper in games than other teams. Then he uses some basic leverage-based analysis combined with defense independent stats to show how the Angels have pretty consistently got their best relievers into the game in the tightest spots. All of this doesn't prove the point, of course, but it does at least strongly sugg-- wait, what? Keri doesn't do any of this work? He spends 1300 words setting up the problem and describing (to no real point) the Angels as a team, but doesn't actually show any support for his theory? That's pretty impressive.

Alright, maybe the second theory? It's the old "Mike Scioscia is a genius" thing. Evidence cited for exactly what Scioscia is doing? None. Evidence cited for the general theory? None.

In fact, as I understand it, there is no evidence that managers can beat Pythagoras on a consistent basis. Chris Jaffe, who's done more work on managers than anyone (as far as I'm aware), wrote here (albeit back in 2006) that, "Personally, the only two components I firmly believe has a managerial impact are effet [sic] on individual hitters and pitchers."

A manager's effect on individual players, of course, would show up in runs scored and allowed, not in a team's beating Pythagoras. This is why the final sentence of Keri's second theory is silly: "But the best managers might make a bigger impact in other areas, such as motivating his players to succeed, or simply putting the best nine guys on the field at any given time." Is there any reason to believe that these things would not show up in runs scored and allowed but would show up in wins and losses?

Anyway, that last part is a quibble,¹ and I don't want my main point to be lost: this sort of empty theorizing without even a hint of evidence is, however many times Keri cites xFIP or UZR (ahem), however many SABR presentations he cites in footnotes, however good it is that he's trying to popularize Pythagorean records, not sabermetric analysis.

Obviously, not every question has to be fully answered. A person might, in particular, not have access to the data or the skills necessary to perform the analysis. "Better batted ball data could test this theory," one might say, or "Can someone give me a pointer on how to extract what I need from Retrosheet so I can test this?" (The latter, perhaps, in email or on Twitter rather than in a published piece, especially on Grantland.) But a simple theory with no acknowledgment that the actual work of supporting the theory still needs to be done? That doesn't sit well with me, not coming from one of the three most prominent sabermetric voices out there (Schoenfield and Neyer, I'd say, being the other two).