An issue with Sasszine script and Elemental Shamans


#1

Hello,

I have been playing around with the Sasszine script, as an elemental shaman, and I think that something is wrong.

Indeed, I think the add positionning / lifetime / … is way to favorable to Earthquake damage.

Whether simming myself (http://www.askmrrobot.com/wow/simulator/report/177639967a9f42db943605bfdcff1071) or the generic elemental shaman (http://www.askmrrobot.com/wow/simulator/report/a89d871695a845fd944dac5f5c227f7f), Earthquake damage makes up for more than 30% of total damage, and it is the 1st source of damage.

I have looked at dozens of logs, whether mine (and my two eleshamn guildies) (https://www.warcraftlogs.com/reports/z2HmT1GZBVryAwtj#fight=14&type=damage-done), or other logs selected using a very advanced sampling technique: randomly picked from WarcraftLogs.

  • I have not found a single one where Earthquake is the first damage source, and seldom where it is second.
  • In any case, it does way less than 30+% damage: I have had a hard time finding one with 20%.

I have found a mistake in the spell script, when Elemental Focus is up:

= 0.775 * TotalSpellPower * TotalDamageMultiplier * (1 + 0.05 * ArtifactTraitRank(TheGroundTrembles))
if not HasSetBonus(19,4)
= Damage * 1.1
if HasSetBonus(19,4)
= Damage * 1.2

should be

= 0.775 * TotalSpellPower * TotalDamageMultiplier * (1 + 0.05 * ArtifactTraitRank(TheGroundTrembles))
if not HasSetBonus(19,4)
= Damage * 1.1
if HasSetBonus(19,4)
= Damage * 1.17

to account for T19 nerf in 7.2.5.

This may make things better, but the results are so different from actual logs that I don’t think this is the whole thing…

Cheers,


#2

Yeah, I forgot I had special handling of elemental focus on earthquake and liquid magma. I’ll update those.

I would assume earthquake damage is lower on logs because the enemies get moved around, which makes you not hit with all the earthquake ticks. We’d have to decide if we want to model that or not. The amount of movement of the packs would vary a lot from encounter to encounter and raid team to raid team.


#3

I don’t know what is the best way to handle it internally, but I really believe something needs to be done.
I had a bit more time, I’ll expand on the issue, and why it is problematic.

First of all, I have built a little something with the WLogs API in order to get more significant data on EQ usage. I have randomly fetched a log in every 100 logs between rank 1 and 5000 (60th percentile).

The source code is available on GitHub, and the results are in the repo, the average damage of EQ + Seismic Lightning is 15% of overall damage, roughly half of what AMR predicts.

This has some important consequences when it comes to optimization, in particular the choice of legendaries. Indeed, the damage that AMR attributes to EQ is actually done by Chain Lightning.

WLogs makes it very clear that two legendaries are way better on this fight, Sephuz Secret + Al’Akir Acrimony.

The lack of damage attributed to CL leads to a situation where using 640 Uncertain Reminded (http://www.askmrrobot.com/wow/simulator/report/526bc0b7327940378e29cfeea2ac9940) sims higher than 970 Al’Akir Acrimony (http://www.askmrrobot.com/wow/simulator/report/2d754dd9d306490097726f13f88e9f3c).

While it is easy to overcome this effect by locking Al’Akir, I think there is important effects on the desired stat distribution (for example, Mastery has no effect on EQ, but Chain Lightning damage scales with Mastery)

Thanks for reading. I remain available for anything you may need to improve this damage distribution


#4

So, I wonder why people aren’t getting more earthquake DPS in logs? Are people moving the boss/adds a lot?

Maybe players are using one earthquake when the adds spawn, and then just moving to chain lightning spam when they get lower? Also, top parses could have the adds dying faster than the default script as well.

I’ll take a look at some of the parses and see if we edit the duration of the adds to match the logs, if the results look closer. I know we saw this with some assassination logs as well - some teams were killing the adds significantly faster than the script assumes.


#5

It may be an APL issue.
According to this post (https://us.battle.net/forums/en/wow/topic/20756795646), spamming CL and overcapping Maëlstrom seems to lead to higher DPS than using EQ, when there is less than 7-8 targets.

It seems like most people have read this, or intuitively figured it out.


#6

Well, that is an easy test to do.
Default rotation (uses earthquake with 4+ targets):
1848.7k DPS

Only earthquake with 7+ targets (so never happens in this boss script):
1698.9k DPS

Only earthquake with 5+ targets:
1737.5k DPS

So, it looks to me like the math in that post is incorrect. Maybe people should be using Earthquake, but aren’t?


#7

There was pure bullshit here about bad CL damage calculation. I tested it, it works fine.


#8

I’m not sure I follow. What was bullshit? And, what works fine?


#9

I thought that AMR’s calculation for Chain Lightning damage was wrong (too low). I wrote a 1k character piece to explain in detail what it was bad, posted it, and then, had some tests on a dummy to make some nice screenshots, highlighting the bug.

It turns out everything works fine, so I edited the long message, that was indeed BS, to what you see now :slight_smile:

I am still experimenting though, because I made the test on M Harjatan last night, and it definitively feels like I am doing more damage spamming CL than ‘wasting’ GCDs on EQ. But it is definitively not CL damage.


#10

Hmmm… I’ll try to look at some more logs too, see if anything jumps out at me as possibly off.


#11

I may be onto something.

I noticed that Chain Lightning Overload damage is quite low in AMR, compared to real logs (you would expect something between 80 and 120% of ‘normal’ CL damage, and AMR gives 50%)

I have looked at the calculation (http://www.askmrrobot.com/wow/theory/mechanic/spell/chainlightningoverload?spec=ShamanElemental&version=live) and you may be missing a
* (1 + BuffStat(DamageMultiplier,MasterOfTheElements)) to account for the artifact trait.


#12

Yeah, taking a look at that, there is a problem there. It will actually pick up the Master of the Elements buff… but it is not picking up a bunch of the other buffs for the overload. I fixed that and will post it shortly - I got the chain lightning damage to match some logs I was looking at.

The other thing I noticed is that maybe the script is missing some flags for “aoe checkpoints” which is causing stormkeeper to not be saved for AoE.

Once I fixed those things, using earthquake still does significantly more damage than not. I’ll post some new sims once I put them on the live site.

Just mathematically… well, you can look at the “damage per execute time (DPET)” on the reports. The DPET of earthquake is higher than that of chain lightning. Just sitting on that maelstrom isn’t going to be a DPS gain. If the adds were to die or move before all the earthquake ticks hit, then it might not be worth casting, but if you can, it’s worth it.

All my calculations show 4 enemies to be the cutoff. If you can hit 4 enemies with a full earthquake, it is worth casting.


#13

On this topic, there has been some blue confirmation of the community calculation (https://us.battle.net/forums/en/wow/topic/20755546115?page=28#post-554), with a fix targeted for 7.3

I look forward to the overload update, so I can check if any differences remain.

Cheers


#14

It is updated already in the live simulator.


#15

OK, great.

I have tested this on the AoE training dummies in Warspear, using this boss pattern (ST dummy * 3) (http://www.askmrrobot.com/wow/theory/boss/291e1a08548546f98e8577fffe956395?spec=ShamanElemental&version=live&share=true)

First test: spamming Chain Lightning for 5 minutes (http://www.askmrrobot.com/wow/simulator/report/7911db7c96784c259caa11cc4d8c0b20): Live matches the logs within 5% .

Second test: spamming Chain Lightning, using EQ when available (http://www.askmrrobot.com/wow/simulator/report/0f685073314244548218fbbb6e6226f6): Live matches the logs within 5%

Conclusion, I think the damage of both spell is good :slight_smile:

There is one last thing I wanted to test, but I could not find how and if it is handled in the ‘standard’ rotation, and therefore could not include it in my simplified ones:

In an AoE situation, to maximize Lightning Rod damage, one would typically avoid casting Chain Lightning on a target that already has the Lightning Rod debuff, in order to get as many targets affected as possible (Only the first target of a CL can be debuffed).
In most real cases, one simply ‘tabs’ between each CL cast, to try and maximize the number of debuffs.

This has a real effect on CL damage (and is also really RNG, so it is gonna be a bit trickier to test!).

Thank you for your help,


#16

When you look at the default rotations, expand the “target priority” section and you will see there is logic to switch to targets without lightning rod.


#17

So I was looking at this a bit more… and you get a significant DPS increase if you multi-dot with flame shock. But, no one is doing that in all the logs that I look at.

Multi-dot’ing is annoying to do… but lots of specs do it with no problem. Why wouldn’t elemental shamans be doing this?


#18

I have done a first approximation of this. I am talking tooltip damage here and 0 haste, so depending on set bonuses, buffs and so on, the exact numbers may vary, but I think the order of magnitude is OK. For FS to do more damage than CL, the adds need to live more than 10s when there is 4 targets, and that’s without any procs, such as Mastery, Elemental Overload, or Lightning Rod

I think most adds don’t live that long, and therefore, it’s not worth it.