I guess of all of us you would be the most qualified to make a set of rules like this. I wonder if you could quantify them at all. I realise the number of bets must depend on the size of the divvy etc but I would like to see some figures. How many bets for what average price using how many rules would you say would guarantee a system is going to keep winning.
Aside from the obvious benefits of knowing your answer to this question I've always seen backfitting as a fun project to try to undertake and a few figures would give a target to aim at. Like is it possible to devise a system that would win over a thousand bets just using the letters of the horses names as a silly example.
Come to think of it a good example of retrofitting like that would serve as a good lesson to would be system creators to show how easy it is to fool yourself into thinking you have a winning system.

KV