View source for Pairwise Sorted Methods

Pairwise Sorted Methods all share these properties:
* The list of candidates is first sorted top to bottom according to some metric.  This can be considered to be a seed for the next sorting stage.
* Starting at the top of the list, each pair of candidates is processed in some way in accordance with their pairwise (one-to-one) contest.
* The winner is the candidate that ends up at the top of the list.

Almost all methods of this type satisfy the [[Condorcet Criterion]] because a candidate that pairwise defeats every other candidate will eventually be bubble-sorted to the top of the list.

In other fields, the pairwise sorting stage is known as Local Kemenization or Rank Aggregation.  See, for example, [http://www10.org/cdrom/papers/577/ Rank Aggregation Methods for the Web].

One nice by-product of rank aggregation is that the initial seed order is acyclic --- any pairwise cycles that might exist are precluded from the final result.

Among pairwise sorted methods that have been suggested on the election-methods mailing list are
* [[Pairwise Sorted Approval]]:  Use total Approval score in descending order for initial seed.
* [[Pairwise Sorted Approval]]:  Use total Borda score in descending order for initial seed.
* [[Approval Sorted Margins]].  Use total Approval score in descending order for initial seed.  For subsequent pairwise sort, repeat until all pairs are ordered pairwise:
** Search through list for pairs that are out of order pairwise.  
** Reverse the pair with the least difference in approval.