Balance Druid/NF "Convoke the Spirits" during Celestial Alignment

The current AMR Balance Druid rotation has a comment about not using Convoke during CA because “It is best not to use this during Celestial Alignment or Incarnation”. This goes against usual advice, so I went and tested it out and I couldn’t reproduce the results that would lead to that conclusion.

Note: all results here are using my patched version fixing Starlord/Eclipse bugs with Convoke, see Balance Druid + Night Fae "Convoke the Spirits" bugs - #3 by delroth

Comparison here: Not Available – you can see that using Convoke during CA is a good 2-3 %pt improvement, well outside of stddev. This is using (HasBuff(CelestialAlignment) or CooldownSecRemaining(CelestialAlignment) > 15) as the condition for Convoke – basically, trying to keep alignment as much as possible (at T+0, T+360), but not delaying for long enough to miss casts in practice.

PS: As I continue finding such issues, I really miss being able to simulate my own gearing strategies instead of having to wait for you busy folks to update the default ones. Now that the gearing is less absurd than it was in BFA (azerite, corruption, etc.) is there any hope to see custom gearing strategies coming back?

I can update Convoke the Spirits today. That spell is a rough one… who knows if our logic for picking what spell to use is anything even close to what the game uses. We just had a few relatively vague blue posts and in-game testing to go off of. It was technically difficult to implement - it’s triggering those other effects it shouldn’t because I forgot that when you put a trigger at the spell level, it fires the cast complete event - so I need to add some checks like you found to stop it from proc’ing a few things it shouldn’t.

We don’t have any plans to bring back custom gearing strategies right now. They optimizer has evolved over the past couple years to a point where they won’t really work anymore. It’s not as easy as generating a few thousand data points and then plugging it in - we have a layer of mathematical modeling that sits over the top of the simulation results to make the optimizer work well.

1 Like