JB>now... race  a b e h i and take the two fastest,
JB>and then race f g c d and the third place getter from
JB>the last race and take the two fastest

JB>that'll do it

JB>12 races.

JB>is there  a better solution?

Using (excellent observations wrt) relative spacing, more horses could 
be eliminated in the early races.  The (correct) answer given was 8 but 
I think that depended too heavily on there being 1 or 2 best case 
scenarios (a relatively even spread of the fastest horses among the 
various groups).

þ CMPQwk 1.42 999

--- Maximus/2 3.01