December 27, 2005

Shannon Sharpe Imitates Philosophy

No sooner do I declare hiatus than I break it to note that Shannon Sharpe:

But playoff teams have the record to get into the playoffs. It's as simple as that

seems to be making an argument of the exact same form that Doug Lavin discusses in an early draft of "Practical Reason and the Possibility of Error" (which analyzed the flaws in the argument that the best team would always win the division, because the best team was by definition the one that won the most games).

Posted by Matt Weiner at December 27, 2005 07:51 AM
Comments

Whether there are flaws in the argument that the best team wins the division is begging the question. Because it assumes that the best team should go to the playoffs. That's not the way the game is structured - there isn't some arbitrary multivariate bastard exponentiated version of quarterback ratings (or even footballoutsiders.com) that determines the best team, and then that team gets the trophy. That's the way fantasy football (ugh) works. In the real world of sports - choke on that for a moment - the "best team" doesn't win the division or go to the playoffs, only the team with the best record and/or best thoroughly irrational but rule-defined set of tiebreakers does. Recognizing this is why they pay Shannon Sharpe the big bucks, not philosophers or bloggers.

If you care about this sort of thing, in bicycle racing (which has many very clear applications of game theory strategies, much more so than say running, because of the way drafting works) there are quite a few examples where "the strongest rider didn't win" due to someone else's better-applied or more guileful tactics. Okay, he was the strongest, but he wasn't the _best_.

Posted by: Ben at December 28, 2005 01:24 AM

Of course the team that goes to the playoffs is and should be the one with the best record, not the best one -- but one would hope that when Shannon Sharpe explains why San Diego isn't a playoff team, he would come up with something a bit more enlightening than "Their record isn't good enough." We knew that. The race is not always to the swift, blah blah blah.

In re this: best thoroughly irrational but rule-defined set of tiebreakers does

"Thoroughly irrational" can be rigorously defined here: The tiebreakers violate one of the Savage axioms. If Kansas City and Pittsburgh tie for the #6 seed in the AFC, Kansas City will make the playoffs, but if Kansas City, Pittsburgh, and San Diego tie, Pittsburgh will make the playoffs. This violates the principle that says that if A is preferred given the choice between A and B, B should not be preferred given the choice between A, B, and C. (I do not necessarily think that the Savage axioms define rationality.)

Posted by: Matt Weiner at December 28, 2005 07:14 AM

he would come up with something a bit more enlightening than "Their record isn't good enough."

I don't know. I think this is at least as good an explanation as G.E. Moore's "Here is one hand" argument for realism (from "Proof of an External World," but you knew that). It's basically the same argument. You can make up a lot of byzantine arguments for skepticism and brains-in-a-vat, or for how the Chargers got one or two or four unlucky breaks and lost games they could have won. But the bottom line is that disbelieving your common sense in the matter of an external world is tendentious, and so would be arguing that teams that didn't win games when they needed to win games are qualified to be in the playoffs, should anyone other than a homer actually make such an argument.

Also, I hate to say this because it spoils the simplicity of my argument (that Sharpe is a Moorean anti-skeptic), but by bringing up several examples of games-the-Chargers-woulda-shoulda-coulda won but instead lost, he's implicitly pointing out that playoff quality teams put these kind of games away and don't count on avoiding unlucky bounces.

Speaking of inscrutable tiebreaks, with two weeks left on the schedule, Jacksonville was 10-4, SD and Pittsburgh both 9-5. If they all won out, Jax and Pittsburgh went to the playoffs (Pittsburgh over SD on head-to-head). But if Jax split, they would all tie at 11-4, and Jax and SD go based on better conference record. (Even though Jax beat Pitt and Pitt beat SD.) This is possibly even more perverse than your example because Pitt/SD's fate depends on an unrelated Jacksonville game, and because Pittsburgh owned the head-to-head over SD.

However, in both these examples, it's not clear that they violate the Savage axioms. Because if A and B are tied and C has a worse record, that is a different state of affairs than if A, B, and C are all tied. That is, A, B, and C are not equivalent between the two cases. A Kansas City team that winds up tied with Pittsburgh but better than San Diego is not the same as a Kansas City team that is only good enough to tie Pittsburgh and San Diego. (As Dr. Shannon Sharpe has proved in his recent article in the Journal of Quantifiable Outcome Assessment, San Diego is just not good enough.) So it may be quite rational to disfavor Kansas City in that situation. It's like arguing about counterfactuals - well, would you pick Kansas City, if it were tied with San Diego, which it isn't? How would you like Kansas City if it weren't in Kansas? (Oh, wait...)

By the way, how likely is it that the savage axioms define rationality? In the usual Western ethnocentric view, savages are rarely held to be rational. Although they don't play football, which is a point in the savages' favor.

Posted by: Ben at December 28, 2005 10:45 PM

>by bringing up several examples of games-the-Chargers-woulda-shoulda-coulda won but instead lost, he's implicitly pointing out that playoff quality teams put these kind of games away and don't count on avoiding unlucky bounces.

I think that is right.

At midseason it makes sense to say something like "team x is looking like a playoff team, they are 3-3 and 2 of those losses are due to fluke plays and bad calls" on the expectation that they will do better in the rest of the season without the fluke plays and bad calls. At the end of the season that doesn't matter; the close games are just additional losses.

Posted by: joe o at December 29, 2005 03:05 PM

G.E. Moore's "Here is one hand" argument for realism

I was going to respond by saying "But it's only philosophers who need to be hit with arguments that repeat the obvious in that fashion" (I apologize to my colleagues), but actually football homers do too.

I was expecting something a bit more of the shoulda-woulda-coulda analysis, "Good teams don't play kajillions of close games." I suppose it's not fair of me to expect Sharpe to have written a column that I could have written before clicking the link.

Yet I still think it can make sense at the end of the season to say, "Non-playoff team X played better than playoff team Y, but they have a worse record because of bad luck in this way and that." Not that San Diego is in that category.

Posted by: Matt Weiner at December 31, 2005 07:23 AM

Incidentally, Doug reminds me that he put the original example in an early draft entirely to make fun of Ed Bou/chette, and that it isn't in the published version. I've updated the post to reflect that.

Posted by: Matt Weiner at December 31, 2005 07:24 AM