[EM] WDS repsonse to Tarr re Condorcet v Range & strategy

[Warren Smith]

>Adam Tarr:   I'll now respond to Warren's earlier message.
>I don't debate that the "more-favored front runner first, less-favored
>front runner last" strategy is useful (often optimal) in Borda, but I
>can't easily imagine a scenario where it is useful in Condorcet.
>WDS: I did not say it was the best strategy.  I merely claim it is
> an obvious strategy, which *sometimes* is best,
>Tarr: In all seriousness, when?  Make some simulations that demonstrate
> this, or at least show some examples.

WDS RESPONSE:
Sorry, I was under the wrong impression you had already
seen the example given on the CRV web site.  Go to
  http://math.temple.edu/~wds/crv/RangeVoting.html
click "Condorcet" on the left, click
 "incentive to strategically exaggerate"
and you reach an election example where frontrunner max/min exaggeration is optimal
in a very large number of voting systems all simultaneously.  Then you can click
the  "click here"  thing to get even more discussion.


> This exaggeration is a strategy that many members of the public will adopt.

>tarr: This seems like an unsupportable assertion, particularly given your
assertion that voters will tend toward honesty unless given a
compelling strategic reason to do otherwise.  (An assertion that I
agree with wholeheartedly, by the way.)


--Well first of all, this kind of exaggeration IS the optimal strategy,
essentially ALWAYS, in Borda and in every nondegenerate weighted positional system.
Ok?  So under the Tarr-sometimes-used-implicit-assumption voters will always be
exactly rational, they will vote in exacly this manner.   Now given they
do that, it seems plausible many will also do it under Condorcet or IRV,
because as I said it is plausible Joe Public is not going to look deeply
at the Schulze Beatpaths Condorcet system.


-- second, the "given your assertion" is not actually what I said.
But I agree it is hard to support this kind of argument for obvious reasons...
that does not seem to stop you of course right below here from
making your own unsupported assertions about human behavior
and "urges" which are equally hard to support:


> tarr:
I don't debate that, in the rare cases where voters can be trusted to
be both honest and expressive, that range voting maximizes total
satisfaction.  A good example of that would be the voting of a small
committee whose votes are subject to public scrutiny.  But in large
public elections with secret ballots, the urge to push all your
rankings to 99/0 and vote strategically is huge.

--WDS in fact, I did an actual study of real people
doing range voting and urged to act as they would if this were the real election,
and <24% succumbed to that urge.
   http://math.temple.edu/~wds/homepage/works.html  #82.

>This contrasts with the best Condorcet methods, where strategic voting
is less important than in basically any other method known.

--another unsupported and meaningless assertion by Tarr.
Contradicted by my explicit example in an earlier EM post about "DH3."
Also contradicted by the fact that approval and range voting are specifically
designed to be insensitive to strategy, whereas Condorcet's design
decisions nowhere even considered strategy and nobody so far has even
worked out what good voting strategy in Condorcet methods even *is*.

>tarr:
Again, your "honest range" voting is, in my opinion, very unrealistic.
 Especially after an election or two, voters will get the clue.

--WDS: you may be right.  In that case it will reduce to a mixture
of stratgeic-range, i.e. approval-style, voters, and honest ones (from
among the voters not yet with a clue).  That will perform quite well,
thank you, in fact my simulations suggest better than any Condorcet method.


>The situation is nearly completely reversed when it comes to
Condorcet.  Your "honest Condorcet" is fairly realistic, since there
isn't really a strategy that has been shown to be consistently better
than an honest vote in winning votes Condorcet.*

--WDS: what nonsense.
I claim it is usually trivial, in any given well publicized IRV or
condorcet election, to think of a strategy highly likely to be better than honest.
I do not have any algorithm in mind. I simply claim that I personally, as a voter,
can do it.
Easily.  Most of the time.


>tarr: And your "strategic
Condorcet" strategy is a strategic nightmare - basically never the
optimal approach.

--WDS: I repeat my offer I made privately to Tarr, now publically.
If anybody is willing to code up as a C program their own Condocret voting strategy, either
optimal strategy if you can figure out what that is, or what you think is a good but perhaps
suboptimal strategy, then I will be happy to include that in the 2nd edition
of my giant sim comparative study whose first edition was
   http://math.temple.edu/~wds/homepage/works.html  #56.

>tarr:
* OK, in the interest of fairness, here is one winning-votes Condorcet
strategy that is arguably superior to sincerity.  This is from Blake
Cretney.  It's pretty simple: if you have a sincere tied ranking, it's
better to rank those candidates in some random order than to rank them
equally.  So instead of ranking three candidates tied for fourth,
rank them 4, 5, 6, (in some order) and kick any candidates below
fourth down two slots.  There are situations where this strategy can
hurt you, but on average (aggregating over a large number of voters
with similar preferences) it will not.

--WDS well, I think I agree with (a corrected version of) what you just said.
wds