>>/22428/
I am doing the same, just juggling in my mind some eyeballed expected values and sums across of sets of interchangeable runners.
Still tinking about putting it in some machine learning algo but I don't even know how to manipulate the loss function to make it work across a linear programming problem further divided up into separate runner categories.
It is better this way, to not research it, just play the gem like a human being.