[EM] Hierarchial MAM

[Michael Ossipoff]

The motivation for the deterministic Ranked-Pairs (DRP) methods is to
avoid mid-count randomization. But maybe some such measures could
_reduce_, even when not eliminating, the need for mid-count
randomization.

An important advantage of MAM is that it avoids some unnecessary
discarding of defeats, by randomly determining their order of
considration. But using non-random considerations when appropriate
needn't discard unnecessarily many defeats.

So I suggest Hierarchial MAM (HMAM).  Or, if it were named purely
descriptively, it would be called Hierarchial Random Ranked-Pairs
(HRRP).

It would have the 3 paragraphs of HDRP, with their hierarchial relation.

Paragraph 3 would have the contingent clause (I forget what my other
name for it was, but it, too, started with "c", probably "con-"), and
the weaker-preferring clause..

The difference from HDRP would be that, in paragaphs 2 and 3, instead
of equal defeats being simultaneously discarded, they'd be discarded
according to their randomly-determined dominance-order, as in MAM.

Michael Ossipoff