Well, I stand corrected. Went out and got a bunch of data, and compiled it into a nice neat little spreadsheet.
https://docs.google.com/spreadsheets/d/1J-fvqfF6yqaHQL-07YSMTzxoVigvMJf-Xu3AGsSpEVw/edit?usp=sharing
It is not in fact RPPM. It also does not appear to be PPM either, as far as I can tell, though we’d require data using a 2-hander to tell for sure. The data suggests that it’s just straight up a 50% chance per AA crit to proc. If that were PPM, it would be 11.5 PPM, which seems a very odd value, especially given how close to the rounded 50% rate it is. If using a 2-hander gives a proc rate of closer to 70% (3.6/2.6 * 0.5 = 0.6923), then we can say it’s PPM, but I suspect it’s simply a flat percentage.
Either way, there’s absolutely zero chance 2h can ever compete with that proc rate with it still triggering off of crits. So basically, they need to change how it works (like restoring the old RPPM on all AA hits, perhaps scaled by crit chance as well), or find some other way to prop up 2h, because even if 2h had a 100% chance on AA crit, they’d still be getting ~28% fewer KM procs per minute than DW at any given crit level.
Also went back to look at the time data to see why it disagrees with this, and I think I see the issue:
My bold. This is incorrect. RPPM operates off of the time since last proc attempt, not time since last proc (except for the bad luck protection component). To properly compute the time-sensitivity, you’d need to look at each AA crit, how long it had been since the last AA crit (not KM proc), and then compute the probability of an AA crit generating a KM proc relative to the time since the last AA crit. I suspect that data will be substantially less clear.