[EM] Taking advantage of cycle proof conditions

[fsimmons at pcc.edu]

Here's a useful cycle proof condition that we used in step three of my previous
message under the topic "tanking advantage ...":)

If there are 2n levels, and only those comparisons that involve a difference of
n or more levels are used, then there can be no pairwise beat cycle.

Proof:  The above condition is more stringent than allowing all comparisons
between levels from the upper half with levels in the lower half, and this
weaker condition is equivalent to placing an approval cutoff between the middle
two levels, which in turn is strong enough to prevent cycles, because pairwise
approval comparisons do not cycle.

Forest