[EM] Expected Utility Differences strategy

MIKE OSSIPOFF nkklrp at hotmail.com
Thu Jan 20 16:01:32 PST 2005


  I've described an enhancement of the Best Frontrunner strategy. It could 
be further enhanced, by considering the possibility that if there's a tie or 
near-tie maybe neither F1 nor F2 is in it. That leads to a way of getting 
the Pij for Weber's strategic value formula. Any easily estimated way of 
getting those Pij will do, and that's one, if you identify 2 likely 
frontrunners.

But the ordinary Best Frontrunner strategy, and the enhances one that I 
described previously should be good enough if you identify 2 likely 
frontrunners.

If you don't, then I suggested the Better Than Expectation strategy.

But there's another possibility, one that simplifies Weber's strategic value 
formula:

  For a particular candidate, X:

  Db is the imporovemnt that you expect if someone better than X wins 
instead of X. How much better would it be if someone better won?

  Dw is the worsening that you expect if someone worse than X wins instead 
of X. How much worse would it be if someone worse won?


  You have a feel for those things, for how much better or worse you expect 
it to be if someone better or worse than i wins.

   Pb is the probability that someone better than X will win.

  Pw is the probability that someone worse than X will win.

You have a feel for that too. How likely is it to be worse? Or better?

  So vote for X if:

  DbPb < DwPw

  or, as may or may not be more conveniently estimated:

  Vote for X if:

  Db/Dw < Pw/Pb

Maybe the ratios Db/Dw  and Pw/Pb would be, for some, easier to estimate 
directly than would Db, Dw, Pb, & Pw.

  I call that the Expected Utility Differences strategy, and it's another 
one to consider when you don't have strong opinion of who the frontrunners 
will be.

...and if you need one of these expectation-maximizing strategies because 
you don't agree that there are completely unacceptable candidates who could 
win, or that there are 2 sets of candidates such that the merit differences 
within the 2 sets are neglligible compared to the merit difference between 
the 2 sets.

Mike Ossipoff

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/




More information about the Election-Methods mailing list