BiB consistently recommends AZ trait choices that are not consistent with Raidbots/Bloodmallet and what I see on WarcraftLogs.
These AMR BiB setups sim lower than other setups when run on Simcraft.
It seems like AMR overvalues Laser Matrix, Pandemic Invocation and possibly Sudden Onset and undervalues Inevitable Demise, Dagger in the Back, and Wracking Brilliance.
I know every gear set is unique and AMR sometimes recommends counter-intuitive suggestions but recommending Laser Matrix (with no Reorigination) over Wracking Brilliance or Dagger in the Back seems really wrong.
If you have an addon export for a specific character (the data you copy from addon to the website) that you were using to test, we can take a look and give you more feedback on what the optimizer is choosing.
Note that we use a slightly different model of the game than simc, and we use a much more rigorous approach to analyzing the data to recommend gear – so minor differences are not unusual. When we look at your specific case we can give you more detail.
Thanks for your response. I have been a long time subscriber to AMR because I believe in the approach.
We can ignore simc results because of differences in the models, even though the differences are substantial. But how can we ignore the fact that data from Warcraftlogs shows AMR is recommending a trait not used by any of the top 400 affliction warlocks in favor of one that is used by almost all of them? I believe in analytics, but theoretical models have to be evaluated when they don’t match empirical data. Could all the top Warlocks in the world be wrong? Maybe but I find it hard to believe.
Thanks – I’ll check it out when I get a chance (whenever a new raid comes out, we get bogged down doing new analysis of the data, we’ll probably have some minor tweaks for some specs coming next week).
I highly doubt that all of the top warlocks are wrong – but I am also pretty sure that there is a self-reinforcing trend at the top of the charts that makes the set of possible top-performing setups seem narrower than it is. Put another way: if all the top affliction warlocks use setup X, all it proves is that setup X is good. It does not prove that setup Y is bad. (This is one of the main reasons we added the “customize” feature to Best in Bags – to show people that there is significant flexibility in stats and azerite traits in many cases.)
But either way, we like looking at specific cases when people bring them to us – keeps us on our toes. We’ll get back to you soon-ish.
Thanks for the post. It looks like when I updated Inevitable Demise after they changed it to stack to only 50, I forgot to update the rotation accordingly, so that trait is definitely being under-valued. I fixed the rotation and we will re-rank it.
Laser Matrix still does more damage than Inevitable Demise - I’ve checked our numbers on that trait many times, so I’m pretty confident in our value for it. They are really close though - would be hard to tell the difference and it is more engaging to use Inevitable Demise so I understand why people would favor it.
I checked the other traits you mentioned and everything looks good. I know I’ve specifically looked at the differences between simc and amr on dagger in the back. I tested dagger in the back a lot in-game to make sure we have it right. I don’t know if SimC assumes casters are behind or in front of an enemy, we just assume casters are in front and only get one stack per proc for a conservative estimate. If you think you can be behind enemies, it could be slightly better than we have it estimated.
Thank you for looking at it. You mention that you sim Dagger in the Back as if the player is always in front of the target. Do you sim Pandemic Invocation as if the player hits the 5 second window 100% of the time?
The “truth” for both Dagger and PI is probably somewhere in the middle.
If you have pandemic invocation, I have the rotation waiting until there is less than 5 seconds left, since it is worthwhile to do so. That is something that you can control yourself, so I don’t see a reason not to do it.