AMR gearing putting Devourer DH at 1% Haste - intended or bug?

I’m seeing something that doesn’t add up with Ask Mr. Robot’s gearing strategy and wanted to get clarification.

Right now, the optimizer is putting me at ~1% Haste, which seems completely off for Devourer Demon Hunters, where Haste is generally considered one of the stronger stats for rotation flow and overall performance.

Instead, AMR is heavily prioritizing other stats while almost entirely ignoring Haste.

So the main question is:

Is this behavior intended based on current stat weights, or is this a bug/miscalculation in how AMR is handling Devourer DH stat priorities?

For me to look at what you are looking at, I need you to press the “help” button next to BiB and post the generated 32 digit snapshot.

When I look at a generic devourer and do BiS, I’m seeing it going for a lot of haste. Maybe there is something specific to your setup resulting in this low haste optimization. It is a new spec, so the models will need tweaking as we learn more.

One thing I plan to do after this week is use some of the combat logs out there to try to pin down a more precise fury drain rate during void metamorphosis. As far as I am aware, no one knows the exact rate of decay and how it ramps up. I am using an estimate right now based on beta logs. I imagine that rate of decay will decide a lot about how good haste ultimately is for devourer.

42c3f7559078487ab8c24d7db2f0a5bb

After playing around with the optimizer, it still does not like haste %:

fe2d215af05347eaa3ff1e8babe90408

Only way to “force” Haste is adjusting stat values from “customize” options by decreasing Vers/Crit in favor of Haste/Mastery.

It appears that the simulation is seeing jumps in DPS when it crosses particular thresholds of haste. So, haste is not scaling linearly. The first threshold can be reached with just the haste from Emptiness and Swift Erasure. If you can’t get enough haste on your gear to reach the next threshold, the optimizer won’t try to get more haste.

My guess is that this would be the haste thresholds at which it is able to cast another collapsing star during void metamorphosis. It makes sense that haste would scale in a step function for devourers.

I will set up some tests and look at log outputs from the simulator to see if I can verify this theory.

1 Like

I looked into this more this morning. I did some analysis of combat logs, and found that I need to revise my fury drain formula.

We know that the fury drain accelerates the longer you have been in void metamorphosis. I cross-referenced some logs and fitted a formula to it. I will post an update with the new, empirically correct formula.

This will not change your optimization, for the reasons that I mentioned before. Haste scaling is not linear for devourers. Our optimizer is running real simulations to do the optimization, so it can fit your gear very well to actual simulation results. I did notice that Best in Slot will actually try to get less haste now with this updated implementation - presumably because the exponential drain increase results in a serious diminishing return on getting more haste.

Here are some other items to consider (and i will be writing this up a bit and creating a link right from the optimizations, so it isn’t hard to find)

Add deaths: when adds die and spawn soul fragments, that can significantly affect how long you keep void metamorphosis going. Our simulation is not assuming any frequency of add deaths. I think we should make this a user-configurable setting, as it can have such a large effect on your void meta duration, and thus stats you would want.

Most guides/discords, etc. base their stat recommendations on simulations from simc/raidbots. As of the time of this post, I can tell you for sure that the fury drain function in simc is not matching what we see in logs. They are doing a linear drain that starts at the wrong value, and diverges from the real game very quickly as you spend more time in void metamorphosis. I assume they will update it soon, but if you are someone who interacts with that community, it might be worth reaching out. I don’t know any of those folks, and they usually don’t like me jumping in to their conversations.

Here is the fury drain/sec formula I was able to reverse engineer from combat logs:

// t = seconds elapsed since Void Metamorphosis activated
// Returns fury drained per second at time t
double VoidMetaDrainRate(double t)
{
    return 337.72 * Math.Pow(55.0 - t, -0.7619);
}

And I implement this in a function to get the interval between fury drain ticks while you are in void meta. Note that this maxes out at 55 seconds - there is no log data right now for a void meta that lasts longer than that, but this fits the shape of the drain increase that I have observed in logs.

2 Likes

Alright, that’s some good information :abacus: there. As long as you are on top of it, I am going to go with AMR :robot: BIB recommendations.

:rofl: :rofl: :rofl: :rofl: I know the history and know exactly what you are talking about here. Not going to waste my time there.

This is highly recommended :wink: