[EM] non-deterministic methods

[Paul Kislanko]

On Behalf Of Paul Kislanko

Yes, I noticed I didn't add right and got the pairwise percentages wrong...

Blame it on trying to do too many things at once but wanting to get a point
in.

Probabilistically, A should not have twice the chance of winning as B and C
in the example. The actual probabilities that would be "fair" would all be
nearer 67:66:66.