[EM] Hey guys, look at this...

[Colin Champion]

Another correction I'm afraid...

I mentioned evaluation results for Condorcet+random (which are quite 
good). They can't possibly be right. Anyone who attempts tactical voting 
needs to make decisions from imperfect knowledge, since he can't know 
for sure how people will vote. Studies of tactical voting tend to factor 
this difficulty out, assuming onmiscience on the part of the manipulator 
and therefore making tactical voting seem more effective than it really 
is. My evaluation does this... but the omniscience must stretch to 
knowing in advance the result of the random coin toss.

The conclusion would seem to be that tactical voting is harder than I 
allow for in randomised voting methods, and randomisation better, not 
worse, than my figures suggest. I think I need a sad face emoji here.

CJC