November 07, 2004

Rooting For Evidence

While watching a bunch of NFL games at once, I had this thought:

Suppose that your only interest in the NFL is that team X be successful as possible. Teams Y and Z are playing, midway through the season. The outcome of this game will not have any effect on team X's fortunes (we can assume that Teams Y and Z are both unlikely to compete for playoff spots*); team X beat Y earlier in the season, and team X will not play team Z.

Does it make sense to root for team Y?

Here's the argument for so rooting: You want X to be as successful as possible. The better X is, the more successful they will be (most likely). The better Y is, the more evidence X's previous victory provides that X is good. So if Y beats Z, you have more evidence that your goal will be achieved.

The argument against rooting for Y is basically this: What happens between Y and Z has no effect on X's fortunes. All you care about is X's fortunes. So why should you care what happens between Y and Z?

This seems to have some resemblance to Newcomb's paradox, but it's not quite the same. Even I, the one-boxer**, think that you have no reason to cause team Y to win should that be in your power. But when you're rooting, you're not making any decisions--even when you root for team X itself, it doesn't make any difference to team X's fortunes. (Certainly if you do so from a sports bar in the basement of the UWM student union.) And if rooting ever makes sense, it makes sense to root for team X.

Nevertheless, rooting isn't just wishing that something were the case. If that were so, you could root for the past to be different--and that seems to me clearly absurd.

Perhaps rooting is much like hoping--but I think the same case arises for hoping. It makes sense to hope for something that is a downstream effect of something you want to have happened--if the lottery has been drawn but not announced, I should hope that my ticket is announced. Does that transfer over to rooting for team Y? I'm not sure.

You might say that your problem is essentially epistemic, and the best way for that epistemic problem to be solved would be for team X, for the second week in a row, to pound a theretofore undefeated team like British currency.*** In this you would probably be right.

*We could assume that teams Y and Z are in the other conference from team X, but as NFL schedules are currently constituted this is inconsistent with the rest of the setup--if Y and Z are in the other conference from X, then X plays Y during the regular season iff X plays Z during the regular season.

**I'm not a one-boxer who subscribes to evidential decision theory; my position is that you should (if you think it at all likely that you'll find yourself in a Newcomb situation) form the intention to one-box, and that it is rational to follow through on such intentions. Nothing like that comes up in the rooting case. I suppose that an evidential decision theorist would find rooting for Y unproblematic; she would think that you shouldn't rig the game for Y (if that be in your power) because then Y's victory would provide no evidence of X's future success.

***No punts! No third-down conversions allowed! 41:49 time of possession! Yoi!

Posted by Matt Weiner at November 7, 2004 03:10 PM

Of course, given the BCS, in college football we always root for victories for those we have beaten. We want our teams to look good, not just be good.

I'm not sure I have a clear enough idea of the practical rationality of rooting to know whether we should root for evidence. Even without the BCS, though, I want my team to look good, and I'm encouraged when those we've beaten win. Go Michigan! (And Tennessee.)

Posted by: Chris at November 8, 2004 02:07 PM

Yes, in college football, the rules are such that the success of my former opponents really does impact my success.

Here's a thought: what if *respect* given a team is partially constitutive of that team's success? Then it's rational to want them to look better, which means it's rational to want their opponents to look good.

Also, I don't think it matters whether X beat Y. I think the argument goes exactly the same way if Y beat X.

Posted by: Jonathan at November 8, 2004 03:47 PM

I'm not sure you can't root for the past. Suppose we think of the Supreme Court cases as like sporting events, and think of cases that involve a lot of historical questions. So we have competitive historical research. It seems that we can root for one side to win the argument, though the historical materials are in the past.

Or think of watching a game we haven't seen before on Classic ESPN. Can't we root in those cases? It seems that ignorance is enough for rooting.

Posted by: Chris at November 9, 2004 07:07 AM

Jonathan is right that the argument is the same if Y beat X--but the only team that has beat X is their closest division rival (albeit two games back), and wasn't playing at the time anyway. (OK, I'll stop with the trash talk.)

The respect stuff is interesting--but I meant to be stipulating that all you care about is X's actual success, measured in victories and playoff appearances. I think the problem I mentioned comes up there. Also to stipulate that strength of schedule doesn't matter (in the NFL it's a distant tiebreaker, but it almost never comes up).

Chris is right about rooting for the past--that was sloppily put. What I should've said was that you can wish p was false even when you know p is true, but you can't root for p to be false when you know p to be true.

One of the issues here is that it's not clear that rooting is rational at all--though a theory that rules out rooting does seem to me somewhat impoverished and puritanical.

Posted by: Matt Weiner at November 9, 2004 07:47 AM

Ex hypothesi, rooting for a team is something over and above wanting that team to do well or being glad when that team wins. That missing element is behavioral.

In its purest form, rooting is a behavior intended to increase your team's chances of winning. Pure rooting is what you do when you're actually in the stadium cheering on your team. Derivative rooting is isomorphic with pure rooting but may proceed even without a well-founded belief that your cheering will help your team win. Derivative rooting is rational if you enjoy rooting behavior (and/or the results of rooting, like pissing off the Yankee fans in your vicinity.)

Pure rooting for Y would be irrational for an X fan, even if she were a 1-boxer. Why? Because the fan would thereby be helping to make Y better. If the rationale for indirect rooting is the desire for evidence that X is very good indeed, it would be self-defeating for a fan to root purely for Y because she would thereby be confounding her variables and clouding the evidential picture.

All derivative rooting is rational if the fan enjoys the process. So, derivative rooting for Y is always rational if the fan enjoys the rooting and is able to sustain it. (I am agnostic as to doxastic voluntarism.)

Posted by: Lindsay Beyerstein at November 11, 2004 05:37 PM