[EM] Condorcet vs RV

[Dave Ketchum]

THANK YOU!  A difficult task but I like your words.

On Thu, 08 Feb 2007 21:17:01 +0000 Michael Ossipoff wrote:
> 
> I’d said:
> 
> Range Voting isn't a bad method, and it's very winnable, but let's not 
> get carried away and say that RV is as good as Condorcet.
> 
> Warren replies:
> 
> --my earlier post which stimulated this, was merely claiming that, as a 
> matter of experimental fact, Range Voting produces Condorcet winners 
> more often than Condorcet methods do. At first this statement seems 
> impossible by definition, but the reason it is true is I am speaking of 
> a 50-50 honest+strategic voter mix
> 
> I reply:
> 
> So you’re making unjustified assumptions about how people will vote. You 
> seem to be assuming that people will vote as sincerely in RV as they 
> would in wv Condorcet. Not a safe assumption.
> 
> Warren continues:
> 
> , and asking what it the probability the winner is the same as the 
> (honest voter) Condorcet winner. My computer sims showed that range 
> voting does so more often than Condorcet methods based on rank-order 
> ballots. Paradoxical seeming. But true. Since this is an experimental 
> fact, it is indisputable.
> 
> I reply:
> 
> It’s indisputable in the world of your simulation. To claim that it’s 
> indisputable outside that world is to assume a lot. You’re assuming that 
> the premises on which your simulation is based are all correct, 
> including your assumption that people will vote as honestly in RV as in 
> wv Condorcet. Always be careful when drawing conclusions about the real 
> world from simulation results for single-winner methods. With 
> apportionment, simulations are more reliable, if we know the form of the 
> state-size frequency density distribution.
> 
> 
> Warren said:
> 
> Computer simulations. The page discussing this is 
> http://www.rangevoting.org/StratHonMix.html (Also, same results happen 
> for other mixes than 50-50...) CAVEAT: I should note that IEVS presently 
> is only doing rank order ballots, I.e. equalities in rankings were 
> forbidden in these sims.
> 
> I reply:
> 
> That invalidates the simulations. Equal ranking is allowed in every 
> Condorcet proposal that I’ve heard of, including all the proposals of wv 
> Condorcet, such as SSD and BeatpathWinner in particular.
> 
> Warren said:
> 
> Quite possibly the situation would change in future IEVS versions that 
> will permit rank-equalities and wv-condorcet methods. I urge EMers to 
> help me program more such stuff into IEVS. The IEVS source code (which 
> continues to evolve) is available http://www.rangevoting.org/IEVS/IEVS.c 
> To riposte to the caveat, I note that there are theoretical reasons why 
> range voting is highly likely to generate honest-voter-Condorcet winers, 
> when the range voters are strategic. The theorem is presented at 
> http://www.rangevoting.org/AppCW.html .
> 
> I reply:
> 
> Riker showed that, with any method, a sufficiently well-informed 
> strategic-voting electorate will choose the CW every time. It was 
> referenced in _Scientific American_, in ‘76 or ‘77. Probably July or 
> thereabouts.
> 
> Warren said:
> 
> I am not sure that Ossipoff was aware of this theorem
> 
> I reply:
> 
> As I said, it’s true for all methods.
> 
> or that he was aware of my computer-sim results. I want to make it clear 
> I am not just ranting the counterintuitive claim that range voting is 
> more likely to generate Condorcet winners than Condorcet methods. I am 
> proving it via both theorem and computer simulation. (The computer sims 
> are independent evidence from this theorem since they were not using the 
> kind of strategic voters in the theorem, at all.) In the model of 
> strategic behavior in this theorem, range voting with strategic voters 
> generates Condorcet winners 100% of the time when they exist. Condorcet 
> methods - even wv-condorcet methods with rank-equalities permitted - 
> presumably do so less than 100% of the time?
> 
> I reply:
> 
> Riker’s statement applies to wv Condorcet as well as to RV, and to every 
> method.
> 
> What you’re missing is that Condorcet wv allows sincere voting much more 
> than RV does. That’s why my single-winner criteria have always been 
> about how much a method penalizes sincere voting. That’s the subject of 
> all of my defensive strategy criteria.
> 
> 
> 
> If yes, then range is definitely better than Condorcet methods for the 
> purpose of generating Condorcet winners; if no, they are the same.
> 
> I reply:
> 
> You haven’t proven that RV is equal to or the same as wv Condorcet for 
> choosing C.W.s.
> 
> It’s all in the assumptions on which your work is based.
> 
> Mike Ossipoff
-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.