[EM] Call for Ideas on Automatic Approval Cutoff Finding

[Forest Simmons]

Here's an idea that combines Kevin's above mean idea with the generalised
Bucklin idea:

The voters mark CR ballots.

The levels are collapsed one at a time, from top to bottom until either
the next to bottom level is reached or a level is reached at which one or
more candidates have at least fifty percent approval.

When this level L is reached, the set S of candidates with fifty percent
or more approval is formed. [This set might be empty if L is l above the
bottom 0.]

For each ballot B, let r1, r2, etc. be the ratings of the members of set S
on ballot B.  [If two or more members of S have the same rating, some of
these numbers will be repeated.]

Set the approval cutoff for ballot B at the value

(r1 + r2 + ... + (L-.5))/k

where k is one more than the cardinality of S.

Forest