[EM] Hey guys...

[Kristofer Munsterhjelm]

I think I just did it!

Let's have some strategic manipulation stats with a similar setting as 
before: impartial culture, 97 voters, 5 candidates, 5k elections, 32k 
coalitional tries per election. Here's Smith,IRV:

Burial, no compromise:  398     0.0796
Compromise, no burial:  1237    0.2474
Burial and compromise:  4       0.0008
Two-sided:              0       0
Other coalition strats: 791     0.1582
==========================================
Manipulable elections:  2430    0.486

As I mentioned, I would expect pushover/nonmonotonicity to appear as 
"other coalition strats", because nonmonotonicity is either of the form 
"A doesn't win, lowering A makes him win" or "A wins, raising him makes 
A lose". Let the election where A loses be called "before" and the one 
where A wins "after". Then if voters who prefer A to the winner can make 
A win by going from before to after, then that's a valid other strategy 
and would show up under Other.

Now, here's Smith,Method X:

Burial, no compromise:  449     0.091614
Compromise, no burial:  1092    0.222812
Burial and compromise:  40      0.0081616
Two-sided:              878     0.179147
Other coalition strats: 0       0
==========================================
Manipulable elections:  2459    0.501734

So roughly the same strategy susceptibility as IRV, *but no Other 
strats*. So it's very likely there's no pushover, and thus the method is 
monotone.

What is this method X? More to come :-)

-km