So.. yeah.. overhealing. Who always gets the blame? The guy with the
slightly longer cast.. and the guy with 0.1s more latency.
Right now it's hard to tell the difference between a flawed strategy (too
many healers per target) or someone that needs to L2P.
I'm thinking it would be feasible to average out overhealing numbers in an
offline parser like stasiscl; something along these lines:
Keep track of the needed healing for a target over the past ~0.5s (i.e. on
the heal that healed him to full, scan heals backwards for 0.5s as long as
you see non-overhealing). Say you get 10k healing needed after a big nasty
blow from a boss. Now for the overheal occuring for that heal, and all
heals for another 0.4s (total 0.9s window i.e. < gcd for uberhasted
people), compute the factor of actual healing needed vs actual healing
received and blame all healers equally.
As for the actual algorithm, I'll non-helpfully dump that on you
squarely :) Though I imagine it'd be something along the lines of
keeping a list of the healers+healing done tuples from the start of the
0.5s+0.4s window and assigning healing/overhealing at the end of it. This
window would obviously terminate early if a heal is seen to actually
_heal_ again, and not just overheal.
Yes, it's probably somewhat of a bitch to code. But wouldn't it rock? :)
Original issue reported on code.google.com by mikeclue...@gmail.com on 29 Jan 2009 at 3:56
Original issue reported on code.google.com by
mikeclue...@gmail.com
on 29 Jan 2009 at 3:56