[EM] Re: SCRRIRVE; my folly

[Chris Benham]

James Green-Armytage , in his first post on CC SCRRIRVE, wrote 
(Mon.Jan.19, 2004):

"....these "hybrid" Condorcet methods haven't sparked my
interest yet. I prefer a "parsimonious" voting method, because it seems
that the more complexity there is in the tally rule, the more convoluted
the strategic incentives will become."

Designing a voting method should IMO be thought of as an engineering task, and so "prototypes" should
be judged purely on performance. Would you turn up your nose at a Toyota Prius (a hybrid petrol-electric
car), because you believe it would be better if it were one or the other?
If one of our aims is to make strategising as difficult as possible, then isn't it in a way a good thing 
if the strategy incentives are "convoluted"?
On the other hand I can see that it would very interesting and more tidy if we can show that SCRRIRVE
is Schwartz efficient, so that the method can have a briefer description.

Marcus Schulze commented (Tues.Jan.20,2004):

"Another problem is: When we agree e.g. that Reverse IRV finds the
worst candidate, then it is difficult to argue why IRV doesn't find
the best candidate."

Since we are only interested in finding one winner, for SCRRIRVE to work Reverse IRV doesn't have to
find the worst candidate;it only has to reliably avoid finding the best candidate.

Kevin Venzke (Tues.Jan.20,2004):

"Raynaud can give weird results:

11 C>A
7 A
9 B
10 D>B

B>C 19 (margin 8), B>A 19 (margin 1), A>D 18, C>A 11 (m 4), C>D 11 (m 1), 
D>B 10.

Eliminate C and then A.  D beats B and wins."

SCRRIRVE eliminates D, and then B is the majority favourite and so wins.

Here is an example that Mike Ossipoff posted (Tues.Jan.20)
Presumed sincere preferences:
40:A>B>C
25:B>A>C
35:C>B>A
100 ballots. B is the Condorcet Winner (and the IRVE winner).

A voters Bury (offensively order-reverse against)B, while all other voters vote sincerely:
40:A>C>B
25:B>A>C
35:C>B>A
All candidates in the Schwartz set. B>A>C>B.
RP, Beatpath, Simpson, Raynaud all elect A.
SCRRIRVE eliminates A and elects C, so the A voters' strategy back-fires.

A voters Bury B, and B voters respond by truncating:
40:A>C>B
25:B
35:C>B>A
As in the last example, B>A>C>B.

Discussing Raynaud, Mike writes:
"B has the biggest pairwise defeat, 75. B is eliminated, and with just A & C 
remaining, A is now the unbeaten candidate and wins.

With wv, the A voters' offensive order-reversal backfires, due to the 
defensive truncation, and is therefore well-deterred."

Wv elects C, as does SCRRIRVE.

Chris Benham