I still don't know if it's possible to deduce the equation being used to generate the primary score for this. I figured it out by establishing a range that the target rate was in and then iteratively narrowing the range by submitting solves at the 25% and 75% points and throwing out the half with the larger score until I happened to get a score of 0. Overall I thought figuring out the metric was made more interesting by it not changing in clear increments. The solve executes a period 6 loop that makes 3 of each output which is repeated 200 times, followed by a 171 cycle pause.