Arcane gearing strategy undervalued T20 set bonus for Arcane



I’ve just created a new gearing strategy for my 924 equipped Arcane mage. Askmrrobot seems to be keen on keeping me away from using any of the new t20 set pieces.

Is this a bug or is it just that bad?

char setup here
personal gearing strategy here


I’ll take a look at it - I was finding that set bonus tricky to use. I still want to play around with the rotation some to see if I can make better use of it. I don’t think it’s going to be an amazing set bonus, but I think that I could get more value out of it.

As an aside, if you have gone through the trouble to make a custom gearing strategy, you should keep the “Machine Learning” option checked when you optimize your gear: you will get much better results. “Better” means that the results will match the simulations better.


The T20 2pc tricky to use?

It gives you 10% bonus damage for 6 secs and 4 Arcane Charges when you activate PoM.

I don’t think it’s rocket science to figure out that you just use PoM on the pull together with all other cooldowns, right after MoA and RoP. Then use it every subsequent AP.

When coupled with 4 set you use PoM lined up with your MoA mini burns as well. Wheter it is ready for your second AP depends then on how many AMs you get out after your 2nd MoA.

If you dont have it ready then it’s up to maths if it’s worth delaying AP by a few seconds.

Edit: I was using machine learning, but I found that because of this issue it would recommend different stats than SimCraft would.


What I meant by “tricky” is that with 4T20, you can get an extra PoM in between burn phases sometimes.

Also, I have found that optimal use of T20 includes using an Arcane Barrage at the start of your burn phase. It would be like: MoA, RoP, AP, ABarr, PoM, AB.

Machine Learning will always recommend different stats than SimC. SimC doesn’t have anything built in that will calculate a function that gets even close to what an AMR Gearing Strategy does. The single simulations from SimC are accurate, but there is nothing in SimC to turn those data points into a function to use for optimizing gear. The very simplistic stat weight scaling that it has built in is nearly useless for gear optimization.

Using stat weights instead of machine learning on your custom gearing strategy will result in significantly less accurate results, especially for Arcane, which benefits more than most specs from the machine learning function.

I made some improvements to the Arcane rotation this morning and posted them up. I don’t think they’ll change the results too much, but, the DPS numbers will be higher. T20 is looking a lot weaker than T19. Some initial testing is showing me that you’d need your T20 gear to be 10-15 item levels higher than your T19 gear to be worth using.


Oh yeah, I get what you’re on about. Getting that extra PoM is up to AM generation rng. Figured you were talking solely 2p, which in itself is about as complex as a 2 piece jigsaw puzzle. :smiley:

Anyway, it does seem you fixed it now. The gear optimizer is now recommending T20 pieces as the best option. (where as before it would recommend even a 885 non tier headpiece from Nighthold over my 915 socketed Tempest head)

Is that Arcane Barrage before the burn phase really worth it’s global? I guess it’s only with 4 charges and even Resonance. You lose out on a global of AP there.

I’m a user of Machine Learning, as I can see the theory behind it. Even though alot of people seem to disagree on it being a legit thing (such as my guildies).
I don’t really care to be honest, the interface is alot friendlier and it sims things on actual boss scripts, rather than just the ever dreadful patchwerk sims with single point stat weights. Especially as Arcane it’s a huge pain to have to re-sim every time you even switch a single gem.

While SimCraft, not going to lie here, might a bit more on point with it’s APL’s and results sometimes, AMR is so close that it’s less than a few dps off, if any.

Anyway, thanks for the response. Have a good day,


Edit: By the way, do you know what the deal is with the Tarnished Sentinel Medallion? According to my [Trinket spreadsheet] ( it’s supposed to be crazy as hell, yet it only sims a minimal dps increase for me, on AMR and Simcraft as well.

Is the fact that it can trigger more attacks on every single AM hit calculated in correctly? It’s tricky to program rng on top of rng probably though.


They nerfed that trinket fairly recently - maybe you are using the old values? It’s an RPPM that procs on hit, so it is working correctly in the simulator as far as I can tell.

Have your guild members read Yellowfive’s article that explains the whole idea behind machine learning? It explains why using a more complex predictive function just makes sense now.