MIKE OSSIPOFF nkklrp-at-hotmail.com |EMlist| wrote:
<cut>
Russ continued:
The "strategy" boils down to a single, short inequality, but Mike got wrapped around the axle trying to explain it. Mike also failed to recognize that there is really only one "strategy," with several corollaries for special cases.
I reply:
It would seem that, according to Russ, we've all been mistaken then :-) We here, and Fishburn & Tideman, and a various other authors on Approval strategy. Surely Russ will write to those authors to let them know their error.
Many strategies can be related to Weber's strategy of voting for candidates with positive strategic value, according to Weber's strategic value formula. But that doesn't make them the same strategy, as we've been using the term here. Only Russ claims that to vote for whichever of the 2 expected frontrunners one prefers to the other, and for everyone whom one likes better is the same strategy as voting for all the candidates who seem better than the election's expected value--one's perceived expectation in the election.
If only two candidates have any chance of winning, then the "best-frontrunner" strategy is a corollary of Weber's formula. If other candidates have a chance of winning, then it is not necessarily the optimal way to vote. Sure, it's a "strategy," but it's not an optimal one. I could just as well say that my football "strategy" is to always go for it on fourth down. Yes, it's a "strategy," but it's not a very good one. It's not optimal. Any strategy that is not consistent with Weber's formula for some reasonable set of inputs is not optimal. Are you taking notes, Mike?
Russ continued:
Then he has the nerve to chide me for pointing out the *real* potential problem with Approval, which is that voters will depend largely on polling results, which could be inaccurate or deliberately manipulated.
I reply:
But it's common knowledge here that voters in Approval won't depend on polling results. That's been explained to you a number of times, but you'll never understand it, and that's ok. I'm not going to repeat why voters won't depend on polling results, because I've repeated it for you enough times already.
Yes, it is true that voters won't "depend on polling results," but you are the one who claimed that voters will have some idea of how popular the candidates are. How will they know that, Mike? By word of mouth? Yes, the previous election results will provide some information, but ... I'm not going to waste my time repeating myself again, Mike. When you don't want to hear something, you are *very* good at not hearing it.
Russ continued:
, it has little bearing on reality. Approval may "converge on the sincere Smith set" if voters base their votes on previous elections only (ignoring pre-election polls)
I reply:
Which they'll do, if they know what they're doing.
Russ continued:
and if EVERYTHING IS STATIC. In other words, it is only true if all the candidates are the same and the issues and concerns of voters haven't changed one bit from previous elections. Is that reality? Of course not.
I reply:
Sure, the electorate might be completely and radically unlike it was before :-) But that isn't likely. And if that isn't the case, then Approval converges as some of us have been describing. Approval strategy is typicallly based on best estimates, Russ.
Oh, listen to Mr. Precise. Either the electorate is "completely and radically unlike it was before :-)" or it is so nearly identical that mathematical convergence can be asserted. Mike, the mathematical convergence property that you parrot only applies in the idealized case in which everything is static: candidates, issues, voter attitudes, etc. Yes, it may still apply approximately if the changes are small, but I doubt you or anyone else has analyzed the sensitivity of the convergence to the changes.
By the way, I would like to know what the significance is of those little ":-)" symbols you sprinkly throughout your messages. Do you think its funny when you make an ass of yourself?
Russ continued:
But in Mike's little world, reality doesn't matter much. Clue for Mike: some voters actually check out candidates rather than blindly voting for party only.
I reply:
I wasn't aware that I or anyone else had said that the convergence depends on party voting. If the electorate can be expected to be anything like it was in the previous election, then the vote totals of
Oh, I see. Is "anything like" a mathematical term, or is it something you pulled out of your rear end?
Clue for Mike: if a voter's approval strategy depends on the previous election results, yet all the parties (except perhaps for the incumbent party) have new candidates, then the voter is voting by party rather than by candidate. That would be obvious to anyone who doesn't have his head stuck in his rear end.
---- Election-methods mailing list - see http://electorama.com/em for list info
