Snapshot ID: 7e4da12215fc478493121d952358945b (Assuming that’s the thing you guys need!)
Hey folks, as the title says I’ve just tried out AMR for TBC Classic and I’m getting some odd suggestions.
Firstly, I’ve disabled raid gear and BoEs and it’s still suggesting them in the BiS list. It’s also suggesting BoP crafted Leatherworking gear when my professions are mining/BS. PvP filter is set to None, and it also suggest PvP gear.
Another weird one is that it is suggesting odd stat combinations, e.g.
+10 Haste on gloves, instead of +15 Strength.
+8 Crit (Smooth Dawnstone) gems, instead of the likes of +8 Strength (Bold Living Ruby)
+24 AP on Wrist, instead of +12 Strength
This suggests the stat weights are a bit whacky to me; Strength should be at the very least equal to AP (more with Kings), so having it slightly worse is definitely odd. Crit and Haste again are definitely behind Strength, too.
I assume you guys have some form of stat weights for profiles to use in the background, will they be available for players to edit any time soon? Especially so we can find upgrades with different things, such as higher stamina/Resil builds whilst still aiming for decent stats.
I don’t see any of those items chosen - they’ll still show up in the per-slot lists of items by design. Same comment for the PvP gear.
The Fel Leather set can be worn by non-leatherworkers. It is only the set bonus that requires leatherworking.
Because of the filters on available items, Dragonmaw is a much better weapon than the other choices, but there are no 2.7 speed off hands to go with it. Having two weapons of the same speed with sync’d auto attacks greatly increases the uptime of Flurry. In this case, the optimizer is finding that there isn’t a great weapon combo for you with two weapons of the same speed, so now you need more crit and haste to keep Flurry up more often.
Unleashed Rage makes attack power just as good as strength. If you want to optimize for kings, you should add that to your settings. I’ll take a look at it some, I thought we had code that would pick strength when they were equal.
We do not use stat weights. We created a mathematical model of the spec and calculate a predicted DPS output for each gear setup. This is how we are able to take into account all the additional settings like fight length, buffs, debuffs, consumables, etc. The optimization changes appropriately for each of those settings.
I understand the Fel Leather set is BoE, but what it is suggesting for me was the Primalstrike LW crafted BoPs, and the Vengeance Wrap (BoE) cloak. I’m guessing it’s not showing due to me manually disabling them per slot and re-optimising prior to the snapshot? I’ve undone it to be “fresh” if that’s the case - Here’s the new snapshot: 880432e86a6b422fb2b4f2f8f8a3a003
For the stats, I manually selected Gladiator’s Cleaver (2.6 speed) in the MH slot, and it automatically chose The Harvester of Souls (also 2.6 speed) OH - but it is still using pure Crit gems (as well as Hit gems now) in yellow sockets, and Stre/Crit in Reds. Should that still be the case with same-speed weapons without Kings? With Kings, it looks right though, i.e. Red = stre, yellow = stre/crit
Snapshot specifically for this if you’re curious: 5d43f8d3d4b14d7aa27b47bb326c1603
ARM has been suggesting Primalstrike for my Enh shaman (7c00251cdbff4d3faf5854976ecb3759), as well, which i do find quite odd - Almost all the best performing Enhancement Shamans in the world have the Ebon Netherscale set, or sometimes epic off pieces, but never the Primalstrike one (Warcraft Logs - Combat Analysis for Warcraft). ARM suggests Primalstrike even with the stock-enhancement char (e8d250b8486547ba8fad2aceb3a5eae4)
Something definitely seems off. Not sure what is causing it, though.
I see, we must be missing a crafting requirement for those items, I’ll update that.
I feel good about the calculations we are doing right now for enhance. If having kings vs not having kings is showing some sort of inflection point in what gems are chosen, I feel comfortable following what the optimizer is suggesting. I think it will be impossible to empirically test a difference when the margins are that small.
This makes little sense to me. Ebon Netherscale has stats that enhance doesn’t need (intellect, mp5).
I’m trying to very diplomatically say: I do not trust the original EJ EP values in this case. I like the model that I made more for making this decision. I do not believe that Ebon Netherscale is a good set for enhance shamans. It doesn’t make logical sense.
We can’t get the 100% absolute best scoring setup with the optimizer unless we are willing to take a very long time to optimize. Set bonuses are one of the trickiest things to work around. I’ll take a look at this case and see if we can improve it somewhat. We try to keep our margin of error in the 0.5% - 1.0% range and this is falling outside of that.
I am very confused by the bot as it disagrees with all other sources for gearing, recommending spellpower gemming over int gemming and hit miles above what is needed for an arcane mage . What is this simming against ?
If you post a snapshot id (use the help link above the gear table to generate one) I can answer your question more specifically.
The optimizer figures out how to get the most damage over the course of the entire fight you have the optimizer set up for. If you are very mana-limited, you’ll have to use a lot of filler spells, which are usually frost bolts, depending on your spec, and thus more hit rating can often be optimal. Other sources are assuming that all you care about is your arcane damage - we assume what you care about is doing maximum damage for the fight length specified.
Dont forget one point. Most of the raid has a form of multi target. So if you see al Arcane Mage go Venthyr (for exemple) is because they push overall rank in DPS using the multitargets.