Calculating the Smith set?

[Steve Eppley]

Mike Ossipoff wrote:
>Yes, that algorithm is good, but you could reduce the number of
>pairwise comparisons needed by adding a rule to admit an alternative
>to the Smith set if it isn't beaten by more alternatives than some
>member of the Smith set. 
[snip]

That looks to me like another kind of pairwise comparison.  How would 
it save time?

--Steve