I've been using data mining techniques to create a playable system that's pretty straightforward but does require some data entry. I'll eventually incorporate ABBY into my research, but for now, I've exclusively used AI Handicap data. My results are based on 1,650 races. I excluded races with fewer than 6 entries, more than one first-time starter, and those held at a "minor" track. Below are the results from a held-out data set (not used to create the betting schema):

48% winners
ROI: 12.4 % ($2.25)
Races played: 129/380 = 1/2.9

The results are acceptable to start using it live, except that only 1/3 of races yield a bet. I'm going to do some more research to increase the % of playable races, and anyone who wants to assist in this is welcome—up to three people.

Thanks!

Andrew.