>>532280886
The website brute force calculates the chances of where a card could show up for training, how many stats it would give, how this would affect which training option you would pick and how that would contribute to your overall stats
Then it does that for every turn in a run while adjusting a bunch of factors, it includes events on cards but is completely blind to skills
>Explain it like I'm a dumbass.
It roughly tries to guess how many raw stat points you would gain from using that card
>I see adding a card sometimes adjusts the tiering for other cards, even in different stats
Kitasan at max dupe has a strong chance of always being on speed training, this means other speed cards with speciality priority are going to be better because more of the time they'll both show up on speed and that's what you pick
but a power card like Vodka with high speciality training too is less likely to show up on speed and instead will be on power more, this is actually worse since ideally you want the most supports on a training even if they're not matching types
Imagine every support had a 100% chance to only go on their own training and you used one of each, you wouldn't gain stats from overlaps ever which would lower the rankings of each one if you had others selected using the tool
There's other minor things like trying to hit the race reward target (35% for URA) that factor in too
Hope this helps