Taltos’s Wager

If 538’s model is accurate, this guy has a 1.1% chance of looking like a genius tomorrow morning. I think the odds of a McCain win are higher than 1.1%, but calling the election for McCain at this point is ludicrous. It appears that the core of his argument is that McCain and Obama are campaigning in Iowa and Maine, states that are polling solid blue, therefore both campaigns must believe that the polls are badly wrong.

I don’t agree with that interpretation. I think what we’re seeing is both campaigns applying Taltos’s Wager, the principle of strategy outlined in this quote from one of Steven Brust’s Vlad Taltos novels:

If your only chance of living through something is if your enemy isn’t a sorceror, or doesn’t have a spare dagger, or can’t jump an eleven-foot crevasse, then you assume your enemy isn’t a sorceror, or doesn’t have a spare dagger, or can’t jump an eleven-foot crevasse

Suppose there’s a 60% chance that Georgia, Montana, and North Carolina are closely divided swing states, that Colorado, Nevada, Florida, Virginia, and Ohio are leaning Obama’s way, and that Pennsylvania, Maine, and Iowa are locked up for Obama. Neither campaign cares about that possibility, because if that’s true, then the election’s as good as won for Obama no matter what either campaign does. When planning their last few days of campaigning, they’re only concerned about the scenarios where their last few days of campaigning will affect the outcome of the election.

McCain is screwed unless there’s a systematic problem with the polling, so by Taltos’s Wager he can safely assume that the polling is systematically skewed towards Obama. If that’s the case, then states like Iowa and Maine are the swing states that could decide the outcome of the election, so that’s where McCain is campaigning. The reverse logic applies to Obama, who is forced to assume the same systematic polling skew because that’s the only scenario in which he has a significant chance of losing the election.

Update: Continuing to read the linked article, I see he has some very harsh words for Nate Silver of fivethirtyeight.com. I’ve got some quibbles with 538’s methodology that I’ve discussed in previous posts, but despite Silver’s acknowledged sympathies for Obama I think he’s been doing a careful job and has been trying to fairly predict the election. Note that Scott Elliot’s election projection site (based on a different, independently-developed poll aggreggation model) has a very similar map, and Elliot is a conservative Republican who strongly favors McCain.