Monday, February 28, 2011

Voting Method and ROY

I realize I'm a little late on this issue, but there was a minor storm about one writer's ballot for NL Rookie of the Year. One voter from Pittsburgh voted for Buster Posey first and then for two Pittsburgh Pirates second and third. There have been suggestions that this was defensible (see here and here). He has been criticized for homerism.
The voter himself defended himself by saying that he picked Posey (the eventual winner) first. (Since he defended himself on Twitter, it is hard to find the appropriate tweets). He just used his second and third place votes to recognize flying under the radar Pirates, but picked the best player first. However, I think the fact that he picked Posey first makes his other picks even more irresponsible.

The key decision for the award was Posey vs. Heyward. Since 1st place votes are worth 5 points and 2nd place votes are worth 3 points, a voter who had Posey first and Heyward second would increase his top choice by 2 points. However, Kovacevic increased Posey by 5 points over Heyward. If his motivation was for Posey to win, this was pretty close to the perfect ballot. I actually would have preferred if he had just chosen three Pirates or three random players. In that case his ballot would have had no influence on the final outcome. Or he could have just picked the two Pirates first and second and then picked Posey third.

Part of the problem is that, according to Arrow's Impossibility Theorem, there is not a perfect way to design a ranking method among more than two alternatives. The choice of a Borda count method pretty much opens up the possibility of strategic voting. I do not think his intent was strategic, but his ballot looks exactly like that of a strategic voter trying to help Posey.

No comments:

Post a Comment