So I decided to finish what I started and abandoned a few months ago. A few months ago, I strove to make acc pp more accurately reflect how statistically accurate you are. After a discussion with Full Tablet https://osu.ppy.sh/community/forums/topics/727540?start=6741333, I made some major changes to the algorithm. I used Full Tablet's idea of using the quantile formula of the normal distribution (or is it quantile of the normal distribution formula, idk how to grammar) to relate accuracy, OD, and number of objects. I then made regressions of its partial derivatives to get an idea of how they were related and then made a regression function with inputs of all three variables (it sounds complicated when it's really not). It turned out like this:

key:

OD = overall difficulty

n = number of circles in map (does not include sliders and spinners)

acc = accuracy on map (with respect to circles not total objects)

s = 2.06199*n^(-0.0468658)*(1+0.196922e^(-0.0103434*n)*(1.01937-acc)^(0.339792)*(79.5-6*OD)

r^2 = 99.251%

Keeping to the original, we then get

acc pp = 1.6988647*1.52163^(13.25-s/6)

Edit 1: It's still a work in progress, the OD and accuracy parts are fine, the number of circles part is the problem.

key:

OD = overall difficulty

n = number of circles in map (does not include sliders and spinners)

acc = accuracy on map (with respect to circles not total objects)

s = 2.06199*n^(-0.0468658)*(1+0.196922e^(-0.0103434*n)*(1.01937-acc)^(0.339792)*(79.5-6*OD)

r^2 = 99.251%

Keeping to the original, we then get

acc pp = 1.6988647*1.52163^(13.25-s/6)

Edit 1: It's still a work in progress, the OD and accuracy parts are fine, the number of circles part is the problem.