Surely the accuracy of the ratings service's assessment of a horses chances are paramount. In your example you have a bit of room to move and if they assess a horse at 25% and they are 5% out you are still betting on a horse that is 5% better than the public's 15%. I would think it's more likely you will be looking at differences that are less than this and a (say) 5% swing either way will tip a correct overlay into an actual underlay or vice versa. If you allow a hefty margin for error on the part of your ratings service you will be cutting your number of bets down a lot.

KV