initially, it is an intellectual pursuit. can you look at ten horses, each with their own performance characteristics, and determine their optimal path?

you have to start on the most superficial level and ignore a great deal of detail. however, once you have that piece done, you move in a layer. sort of an onion approach. for example, let's say you start with the assumption that you need to minimize the distance travelled. (it is probably wrong, but we have to start somewhere.) you model this. the only decision the "horse" has to make is how soon can i get to the one path. now you go in a level. what if i'm in the two path and there is no horse inside me, but there is a horse in the one path in front of me. should i move to one path or stay in the two path?

you have to do these a little at a time or you will drown in the "what-ifs." the saving grace is that it is fun. the computer generates a overhead animation of the race so i watch little races all day.