The stat thing: this is how I do it - I use variables rather than numbers which for my pass and fail thresholds, but you can do exactly the same thing with numbers.
After coding each chapter (and after writing a chapter if I’m feeling extra studious, which I don’t always), I add a file with sets of code and commented out lines to my scene list. The below is an example, using implicit control flow, but you don’t have to use that.
*if hb_ath < hb_ath_lo
*comment LO ATHLETICS FAIL
*elseif hb_ath < hb_ath_mid
*comment LO-MID ATHLETICS FAIL
*elseif hb_ath < hb_ath_hi
*comment MID-HI ATHLETICS SUCCESS
*else
*comment HI ATHLETICS SUCCESS
I run 10000 RandomTests, generating the randomtest-output file and including line coverage numbers afterwards.
Then I check each *comment line and calculate the percentage of how many times the randomiser lands on the lowest Athletics, middle Athletics, high Athletics, and so on.
I do this for each stat including opposed stats. If the randomiser is mostly not passing a test, or it’s much easier for one stat to be high than another, this method will show it.
If that sounds arduous, it is - it’s very boring. But I find it less boring and time-consuming than wading through a chapter to check how many times something is increased or decreased, and it helped me balance much better and more consciously with a sense of the big picture (for example making Royal Affairs easier in general than Creme de la Creme).