Balance Druid SimC v AMR


#1

I’m trying to sort out the difference between AMR and SimC when it comes to doing the sims. I greatly prefer the AMR web interface and gear change tools so I’d like to figure out what’s going on in the sims differently.

Here are two results:
AMR:
http://www.askmrrobot.com/wow/simulator/report/8ee6db138d29406f9d4edb170b1466f7
SimC:
https://www.raidbots.com/reports/S123JVmGZ/index.html

I noticed that since the last time I’d been playing around with AMR you had updated the Emerald Dreamcatcher rotation and given it a special APL. Everything looks a whole lot like the SimC APL now. But to me it looks like we’re getting pretty different results. SimC gives back 965k dps, AMR gives back 878 dps.

I keep seeing, however, that it doesn’t seem like the AMR sim is using 2x SW between Starsurges when the ED buff is up. I’ve even tried just lowering all the lag and human reaction times to zero in case that was the issue. But I’m getting the same results.

I did a log output check and saw that at high enough haste, the simulator is casting SW and LS between SS. Otherwise, however, it seems to be alternating between LS between Starsurges. In other words, it just casts, SS -> LS -> SS until it runs out of AP (sometimes it weaves in another spell) and goes back to building. But that seems suboptimal. It would be better to build high and then:
SS -> LS -> SS -> SW x 2 repeat until it runs out of AP and goes bak to building.

I was tipped off to the difference by the difference in the damage breakdown. Both are getting about 37% of total damage from Starsurge, but AMR is getting 21% of damage from LS and 15% of damage from SW. SimC is getting 20% of damage from SW and 17% from LS.

Since SW is higher baseline DPS and AP/sec, it’s not surprising to see the SimC doing more damage.

I also noticed a pretty big discrepancy in resource generation.

Both fights averaged about 300 seconds but:
AMR generates: 2686.1 Astral Power
SimC generates: 4571.91 Astral Power

I thought maybe something was going wrong with the fight lenghts, but SimC only generated 40 more AP with the moon spells (ie one extra cast of Fullmon).

The big differences were in the Lunar Strike and Solar Wrath totals. That’s reflected in the total casts.
AMR: 72.3 SW, 74.5 LS, 90.1 SS
SimC: 98.9 SW, 63.7 LS, 99.7 SS

I’d love to crack this nut and figure out what’s going on.
Is SimC making some unrealistic assumptions?
Why isn’t AMR casting 2x SW even though the character has enough haste to do so?
Where does the AMR APL differ?

Any help/discussion would be much appreciated. I really appreciate everything being done by AMR and would love to figure out why there’s such a large discrepancy with SimC.


#2

Also, I know that this specific issue of 2x SW is largely irrelevant because of the changes coming to ED in 7.2.5 but I’m worried that the same kind of issue will crop up then since it’s functionally the same idea being implemented.

And, just in general, I’m trying to sort out why SimC is getting almost 10% more dps out of moonkin than AMR.


#3

I should note that the same discrepancy shows up without Emerald Dreamcatcher in terms of AMR v SimC output.

Here’s an OI/IFE setup:
http://www.askmrrobot.com/wow/simulator/report/7e13f3082d5d47a4815d2713797574cb

Here’s a SimC comparison between OI/IFE and ED/IFE:
https://www.raidbots.com/reports/BkGDDBmMb/index.html

For OI/IFE,
SimC DPS: 945,517
AMR DPS: 865,074

The damage breakdowns are somewhat similar:
SimC: 29.2% SS, 20.4% LS, 17.7% SW
AMR: 29.0% SS, 18.8% LS, 17.4% SW

But the casting ratios are pretty different:
SimC: SS 76.1, SW 90.9, LS 74.4
AMR: SS 71.1, 88.1 SW, 67.6 LS

One thing that stands out there is that AMR is wasting a Lunar Empowerments, though that can’t account for the size of the gap we’re seeing. The raw casting activity seems just much higher for SimC. It doesn’t seem like it can be explained by a difference in fight length variation either since SimC is averaging 21.1 moon casts while AMR is casting 22.3.

Still, though, the output difference is about 9%.


#4

One thing I’m noticing is that the spells are just hitting harder on SimC compared to AMR when I just look at the OI/IFE numbers:

AMR:
Moonfire tick: 81,853 average non-crit
Sunfire Tick: 82,161 average non-crit
SS average non-crit: 757,721
LS average non-crit: 490107
SW average non-crit: 401,194

SimC:
Moonfire tick: 93,784 average non-crit
Sunfire Tick: 85,271 average non-crit
SS average non-crit: 770,808
LS average non-crit: 508,850
SW average non-crit: 414,795

The SimC moonfire number looks extra suspicious to me. The nukes and sunfire are doing about 3% more damage in SimC. SS is doing about 1% more

Interestingly, the moon spells do less damage in AMR
AMR / SimC
New: 357475 / 334486
Half: 718006 / 671536
Full: 1374826 / 1331046

edited to remove misreported crit%


#5

Thanks for the detailed comparison – we’ll take a look and see if we can figure out the source of the difference.

We’re a bit buried with 7.2.5 stuff (and hope to have 7.2.5 version of the simulator available in the next couple days), but can hopefully spare a few minutes to look.

One small difference could be the concordance buff… I think our live version still has it at 2000 primary stat instead of 4000. We’ve updated it in our 7.2.5 version but need to copy it to the pre-7.2.5 version as well. I can try to update that today sometime.

@Swol would probably be able to figure this out more quickly than I can though.


#6

The major difference here is that SimC is allowing the global cooldown to get below 0.75 seconds when casting Solar Wrath with Solar Empowerment and the Starlord talent.

This is a major bug in SimC which heavily inflates the value of Starlord. Nothing reduces the global cooldown below 750ms in WoW.

I tried prioritizing double SW casts in the dreamcatcher rotation, but it just wasn’t resulting in a DPS increase for me. There might be a way to make the rotation a bit better in that case - I’m going to hold off optimization on that for 7.2.5, like you mentioned, due to the changes to ED.

Sorry for the brief answer - I’m totally buried getting the simulator ready for 7.2.5 - usually I’d deep dive into this and make some example rotations, etc.


#7

Thanks a ton, that’s actually super helpful and makes some sense about the increased amount of casting.

It also yields a slightly higher uptime on Inc. Compounded with the Concordance difference that might be enough to explain the damage differences of about 3%.

SimC moonfire looks bugged in some way like I mentioned above and that’s probably the last piece of the puzzle that can be added together to explain the damage difference between the sims.

Re: [quote=“Swol, post:6, topic:965”]
I tried prioritizing double SW casts in the dreamcatcher rotation, but it just wasn’t resulting in a DPS increase for me.
[/quote]

I’ve actually seen the same thing even in simcraft and had a post about it a few days back on Slippykin’s Theorycrafting thread on MMO-C. From the responses there it seems like that happens for a lot of people and has been trending that way ever since the 7.1.5 changes to ED.

What was more bothering to me wasn’t so much that it wasn’t coming out as a dps increase but that the AMR sim wasn’t even actually ever casting 2x SW.

RE: [quote=“Swol, post:6, topic:965”]
This is a major bug in SimC which heavily inflates the value of Starlord.
[/quote]

That’s really helpful to know. I had been getting Starlord valued at 9% over WoE in SimC without ED, 14% with. Only comes out to a bit under 3% better than WoE without ED with AMR.


#8

I did notice that it looks like AMR has the 2 piece set bonus as 10% increased crit instead of 15% - so we’ll update that.

I need to get the spell id into the wiki so that our automated tool picks it up. I thought I had caught them all, but must have missed that one.


#9

One more quick note on the balance APL:

I noticed it’s not multidotting with Stellar Flare. I suppose this is something worth testing and obviously depends on time to live for the adds but in a sustained situation you probably want to keep it running on at least 3 targets.

With the new legendary ring coming in 7.2.5 Stellar flare is going to generate a lot more interest from moonkin


#10

I haven’t done exhaustive tests, but, in all the tests I have done, multi-dot’ing with Stellar Flare is and extremely minor DPS gain.

Here is a test I did:
http://www.askmrrobot.com/wow/simulator/report/4f93a9f884014fefa52937e8114a5cd5

As you can see, in this case, multi-dot’ing with Stellar Flare is DPS-neutral or even a slight loss.

I’ll revisit this for the 7.2.5 rotations and the new legendaries, of course.


#11

That makes sense with two targets in starfall range of each other.

When they’re not, though, (for example odyn and dragons in T19), I’d expect it to be a more straightforward gain.

I’d also expect with the new legendary and 40 AP starfall together with Stellar Flare, the
AP spent on Stellar Flare is more of a net gain than 60 AP starfall.

Like you said, though, no need to redo anything until after 7.2.5 and especially whatever number tweaks are coming after the first week of heroic raid data in ToS.


#12

Scratch that, I think you’re right that even if the targets are spread Stellar Flare Multi-dotting isn’t worth very much.

Here’s a test with the Default rotation: http://www.askmrrobot.com/wow/simulator/report/dadb56776ddb475383ba68e8dad4efdc

933k, error of 4660

Here’s a test that just adds multi-dotting for Stellar Flare (up to 3) and Sunfire (up to 5):
http://www.askmrrobot.com/wow/simulator/report/db36905096d64adcb83e5f863e5a2731

941k, error of 4703.3

So, barely any gain, even when the targets are too far apart to Starfall. And I actually think that’s overestimating the value of Stellar Flare multidotting because even though I put in that Sunfire should Multi-dot, the Sim is not multidotting. You can see that the uptime stays only at just under 100% instead of 200% like Moonfire and Stellar Flare.

That raises another issue, though:

Why isn’t the Sim multidotting with Sunfire when the targets are too far apart for the Sunifre spread mechanic to work?


#13

Hmm…when I go to two targets in Starfall range and multidotting I get more of an increase from multi-dotting with Stellar Flare:

Multi-dot Stellar Flare:
http://www.askmrrobot.com/wow/simulator/report/8d08122e8e6849ef8ebc92b430debf2c
1.214m DPS, 6060error

Default:
http://www.askmrrobot.com/wow/simulator/report/792213495c4d4d04adec2433e3047350
1.131m DPS, 5649 error.

That would be a 7% increase even with the 60AP cost Starfall.

Edit to add: Nevermind, I’m seeing that you used M+ script. I was thinking Stellar Flare would only really be useful for council kinds of situations. Really long lived multi-target.

It’s interesting, though, that the result I get is that Stellar Flare multi-dot is more valuable for Starfall-able multi-target as opposed to spread multitarget.

The Sunfire question I asked above still stands, though.


#14

It makes sense, since Stellar Flare benefits from starfall’s stellar empowerment.

I’d have to look at your script more a bit later to figure out the multi-dot’ing issue.


#15

I just added a multi dot count to sunfire and Stellar flare. 3 for flare and5 for sunfire. The entries for sunfire and moon fire are identical.


#16

I’m guess it has something to do with the spacing of the enemies in the script. It’s not straight forward to make the positioning of the enemies work how you want in the scripts - it’s on our list to improve when we have time.


#17

Interesting, I had built two scripts. One in which the two boss targets are 5 yards apart. With the 5yd boss apart script, sunfire gets just about 200% uptime because of the spread mechanic.

With the other script I put them 40yds apart. In that one, Moonfire and Stellar Flare get 200% uptime but Sunfire only gets 100% uptime even though the entries for Moonfire and Sunfire are identical and the Stellar Flare entry is nearly identical but has multi-dot count set to 3 instead of 5.


#18

Looks like the Sim is charging 15 AP for Stellar Flare. In game and according to tooltips on wowhead, wowdb, etc., the cost is only 10AP.

Here’s a log output: http://www.askmrrobot.com/wow/simulator/report/7983d8de31f648ad87469771374aff37


#19

Hmmm… not sure how that change got missed. It’ll show up in the next update.


#20

Found another weird thing going on.

Here’s a log without the new Legendary Cloak for moonkin that allows you to Cast an extra Full Moon before Full Moon goes back to New Moon:
http://www.askmrrobot.com/wow/simulator/report/29abca06dd2d44499318eec38bed99ac

The cloak is not equipped but the buff—Radiant Moonlight—has almost a 100% uptime.

If you look at the log and filter for casts and damage of the moon spells, it looks like the sim casts New Moon, Half Moon, and then Full moon. For the rest of the encounter, it’s only casting Full Moon.

Here’s what it looks like:
0.644 333346 New Moon
4.186 739519 Half Moon
6.118 2986797 Full Moon
49.252 1413386 Full Moon
70.186 1413386 Full Moon
87.672 1522118 Full Moon

Etc.

The overall spell cast data confirms that: there’s 1 new moon cast, 1 half moon cast and, on average, 13.8 full moon casts.

So, two problems:

  1. I have the radiant moonlinght buff even without equipping the legendary
  2. The legendary isn’t working correctly, allowing me to only cast Full Moons.