Here is the snapshot ID for my frost mage: 143bd3b24702472da92af6a9bf798144
I have noticed lately that when selecting best in bags it wants to really bring my crit rating down to pump in more mastery. Since mastery tends to be the worst stat I was starting to wonder why it wanted to drop my best stat for my worst stat. I ran a sim using what BiB is suggesting along with what was recommended based on frost discord, raidbots, and general tutorials on what to look for to beef up my mage. The sim is showing that the BiB is roughly 3k short of the top setup which is in my eyes a big difference. I do apologize if this is somewhat confusing since im still relatively new to using AMR to its fullest. Hopefully the screenshots will help explain.
Here is the best in bags
There is something going on there causing the optimizer to pick a sub-optimal setup. I’ll look into it and figure out how that happened. That’s a really big difference… must be some sort of bad/missing data point or something. Thanks for pointing it out.
We’re looking into this – I think the issue in the simulator is a crazy interaction between the necrolord covenant legendary, deathborne, and kevin’s oozeling soulbind trait. I’ll have to double check how it is actually working in-game.
Look at the uptime of icy veins, rune of power and deathborne. This is using the same BiB gear from AMR in both sims.
Maybe the icy propulsion conduit scaling is off? It also looks like shattered ice lances (from flurry) don’t seem to extend icy veins as they should (due to thermal void).
The scaling is correct on icy propulsion as far as I can tell - we pull it directly from the game data. Shattered ice lances from flurry are definitely extending icy veins with the thermal void talent - you can see it in the logs.
The main difference between those reports is that there are higher crit chances for everything in simc, especially the non-ice lance spells. I think I had comet storm eating up winter’s chill procs, which does have some small effect. I can change that. But that is a minor issue.
The difference is coming from So’leah’s Secret Technique giving crit in simc where in our simulation it is giving haste. We pick the player’s highest secondary stat to give with that trinket in the simulator - this player has higher haste than crit. If I change things so that the stats end up the same, the simulations give equal results for buff uptimes and dps.
The dps difference is ~3.2k, uptime on icy veins is 57% vs 84%, deathborne 32% vs 48%. AMR casts IV at 52, 143, 275 and 379 seconds; Raidbots casts it at 45, 107, 153, 214, 293 and 359 seconds. As a result of re-casting IV, RoP is up for almost half the fight, whereas in AMR it’s up for 38%.
At that item level, it is 212 crit rating, which definitely does change the dps by that much, because crits are what feeds the increased uptimes on all of those buffs. When I change the AMR sim to have the same stats as the simc sim, I get all the same buff uptimes. (Including that fix to comet storm)
It’s actually really hard to set up and exactly apples to apples comparison between the two simulators, unfortunately.
So you claim there is nothing wrong apart from the comet storm “minor issue”, yet the dps difference without So’leah’s is still 3.2k?
Both sims were run with the same settings except for the varied fight duration, but I tried again with 0.2 on both and it’s the same result regardless. Could you link your sim please?
Another difference I can see is that simc is refreshing volatile solvent to keep a near-100% uptime on it, that’s another 120 mastery that AMR only has up for 28%. It’s enough to start casting it and instantly interrupting it to get the benefit.
I updated the site with the fix to comet storm. Don’t forget you have to use the training dummy script as well, since our single target script has some random movement. You have to subtract out the damage from the noxious bolt to compare - I added the crit you’d get from so’leah’s. Our simulator will always get less total damage than simc because of how we chose to model the game, so don’t get too caught up in comparing exact absolute numbers. The important thing is that the skill breakdowns and buff uptimes all look correct. I could have the simulator keep that mastery buff up with plague advisor… but it’s pretty small. I can circle back to it sometime.
Fine if you say the DPS value can be overall lower as long as it’s proportionate - I get that. But the uptimes on IV (and thus RoP and slick ice) are still well below simc. I don’t particularly care about the DPS value being absolutely correct, I’m more worried the simulator is gonna push me into the wrong stats because something is simming incorrectly. I’m not sure what’s wrong though, the rotation looks sane.
Something I just noticed which might account for the lower overall DPS: If you check the rotation in that log it never uses a potion.
i’ve encountered a similar problem, stat distibrution on BiS / BiB are different from a simulation, it suggest haste/versa and crit somewhat equal or leaning towards haste, versus only crit/haste on a simulation.
You don’t have any gear that would significantly change your stat distribution and yield a dps gain. You would have to reduce your item level to change the stats much, which isn’t worth it.
This still happens. Comparing simc results to AMR shows that all the crit calues are close enough to be insignificant, with the exception of comet storm. On AMR it’s crit rate is around 45%, but on simc it’s 54%. I’m not sure if that’s enough to push you over from one IV to the next, but it’s something at least. Maybe this is why? simc/T28_Mage_Frost.simc at shadowlands · simulationcraft/simc · GitHub
AMR is also still not using any potions, whereas simc uses two of them.
I don’t really know what to say about the comet storm crit rate. I’ve checked it multiple times… it is behaving how it should. This must be some sort of timing difference between the two simulators. I’d have to go through simc logs and check the timings on everything to figure out where the difference is. If I thought it would affect which gear is being suggested, I’d do it, but in this case I think the main difference is in comparing simulations, not gear selection, so I don’t think we have to worry about it.
