James,

depends on what you want. You could do them individually which would give you the most accurate result. So do it for Neural Rank =1 then neural Rank =2, etc. then do it for the others.

You could also get an approx by doing Neural Rank = 1 to get the number of races (which would be the number of selections as very few would be dual top selections). Then just do the Neural Rank between 1 and 4 and it will give you the total number of winners. Divide one by the other and you will have a good approx of the strike rate.

I hope that makes sense.