I was looking into this topic lately and came up with an idea. Actually the formula for cancels used to be:
start-up frames + active frames + recovery frames. I wanted to see, if one can calculate the exact numbers that way (first white column).
Then I was wondering: Why does noone take blockadvantage into account? The new formula would be:
start-up frames + active frames + recovery frames + block advantage (usually a negative number)
This leads to new numbers (second white column) and I think, these might make sense indeed.
My theory now is that you have to subtract -12 (duration of a fast, but doable PCC) to get somewhat decent numbers.
What do my fellow Predators think?