search

simulationcraft / simc

Simulationcraft engine/GUI
GNU General Public License v3.0
1.4k stars 692 forks source link

Heavy discrepancy in Wdps scaling sim #2991

Closed ghost closed 7 years ago

ghost commented 8 years ago

http://imgur.com/a/GLkxu

I simmed twice with the same gear and setup, and I had greatly varying results for Wdps scaling. It just about always comes out around 0.95 (normalized) like the second picture, but when I saw 0.64 I knew something was up. Not a big deal to me personally as Wdps isn't something I can really customize for my character, but thought you guys might wanna know this happened. Seemed to be outside the margin of error for such a big discrepancy to occur.

10003 iterations for both.

Collisionc commented 7 years ago

Weapon damage has fairly high variance, and probably needs 25-50k iterations to get a solid value. I'll take a look into it though.

Collisionc commented 7 years ago

Ok, should be much better after that change. Not sure why we had the delta set so low for it .

scamille commented 7 years ago

It should still roughly be inside the dps error boundaries, no matter how much it varies. So eg. if you run the scale factors 100 times, 95 samples should have a maximum range of about 2*dps_error. (If I understood this stuff correctly, I'm no statistician.)

As far as I remember, there are a couple of hacks throughout the modules because setting the scaling delta for weapon dps happens later than some modules initialize stuff. Maybe some problem there.