[Elemental] Rawr Module Development - Page 6

Lord Helmchen · 09/23/09, 3:54 PM

Release 2.2.19
* 3.2.2 Spell and glyph changes
* Shamanism changes to Lightning Bolt and Chain Lightning
* Flame Shock Glyph changes
* Flame Shock Duration + 6s

♦ masanbol · 09/23/09, 4:02 PM

Rawr now shows at T8->T9 gear levels that glyph of ToW is better than both FS and Lava, and FS being the glyph to drop for Lava at around 3400-3600 spellpower + 4T9. Does this follow with the preliminary numbers we've been seeing post-3.2.2?

Lord Helmchen · 09/23/09, 4:56 PM

It seems I am still valuing Lightning Bolts higher than Binkenstein does. We will compare these numbers, when I have time.

Lord Helmchen · 10/07/09, 10:37 AM

Release 2.2.20
* Corrected base damage values for Earth Shock, Frost Shock and Lightning Bolt.
* Changed the DPS display to show both: DPS with and without mana limitation, when necessary.
* CastTime for all spells could go below 1s
* Corrected the mostly unused LB and CL DirectDPS property
* A few changes to Rotation so that I can interface a csv export tool with the resulting dll.

Lord Helmchen · 10/18/09, 2:40 PM

Release 2.2.23
* The totem of Hex is overvalued due to a bug. It is fixed in the source already.

Shegokigo · 10/18/09, 5:33 PM

Quick question, is there a reason that the Worldbreaker Tunic is valued higher over the Hauberk? I can't seem to figure out what's causing it.

Actually I'm noticing for ALL the Tier pieces with mp5 on them. Anything I can do to remove that value from my opitimizations?

tufy · 10/19/09, 3:24 AM

Because you run out of mana before the fight is over, rawr boosts mp5's value. Just make sure you have replenisment ticked and it should display correctly.

Lord Helmchen · 10/19/09, 7:07 AM

The only reason that I can imagine is, that you did not tick all the raidbuffs. Especially the ones which provide you with mana are probably missing.
Rawr calculates that you run out of mana and suggest countermeasures.
In order to completely disregard the mana calculations go to the options tab and set the ratio to 100% Burst.
You can do the same for optimizations.

Zamir · 10/22/09, 11:06 AM

In this post last month it seems to me you've calculated the rotation based around waiting for Lava Burst's cooldown if the figure is less than LB's cast time. Is this the case or am I reading it wrong? I ask because Rawr's rotation output is giving me exactly the same cast sequence as Bink's SEIC model, which in my understanding always fills wait times, no matter how short, with LBs.

I'm also unclear as to whether the lack of wait time is meant to represent a real-world best DPS practice, or was just to facilitate the modelling process. My assumption until now has been that casting an extra LB/CL even in a very short cooldown gap is higher DPS than not casting anything and waiting for the cooldown. [e] I modified SEIC's rotation to use wait times and it did result in a lower DPS output.

Lord Helmchen · 10/22/09, 11:44 AM

Currently I am calculating all possible rotations with and without wait and in the end I return the best one. A W(time) would be visible in the displayed rotation if a wait would return more dps.

Binkenstein · 10/22/09, 7:48 PM

Originally Posted by Zamir

In this post last month it seems to me you've calculated the rotation based around waiting for Lava Burst's cooldown if the figure is less than LB's cast time. Is this the case or am I reading it wrong? I ask because Rawr's rotation output is giving me exactly the same cast sequence as Bink's SEIC model, which in my understanding always fills wait times, no matter how short, with LBs.

I'm also unclear as to whether the lack of wait time is meant to represent a real-world best DPS practice, or was just to facilitate the modelling process. My assumption until now has been that casting an extra LB/CL even in a very short cooldown gap is higher DPS than not casting anything and waiting for the cooldown. [e] I modified SEIC's rotation to use wait times and it did result in a lower DPS output.

I do a bit of extra math.

Firstly, work out the "wait time", ie: the time between the last full LB cast and the next LvB cast being available.
Then take LB damage divided by normal cast times less that wait time.
Compare the DPS of that extra LB to the dps of the portion of the rotation that we're looking at (ie: compare the extra LB dps to 4Lb + LvB)

Lord Helmchen · 10/26/09, 4:55 AM

Release 2.2.24
* Totem of Hex bug fix

Ferala · 11/11/09, 8:03 AM

Since i have been using the last (.24) version of Rawr i often have that when i optimize my char i get a suggestion that the new gear rawr came up with scores a little better than what i had. I accept the new outcome and run optimze immediately again and it comes up with a slightly different setup that is just a bit better. This can be repeated for a few times until it comes up with a new setup that is slightly worse combination than the last. Mind you, i havent changed any gear setting except constantly accepting the lastest improved setup it came up with.

Is this something in the Elem plugin or something in the core of Rawr? Does it mean it doesnt check all combinations but follows paths that seem to be good (and thus not always coming up with the same conclusions)?

In my guild lots of players (lot of different classes) have this same issue.

Any comments on this behaviour?

Lord Helmchen · 11/11/09, 10:11 AM

The reason is, that checking all possible combinations would require an unbearable amount of time. Rawr uses a genetic algorithm in order to follow the more useful paths. The behavior of this algorithm can be modified a bit by changing the Thoroughness slider in the optimization screen.

Lord Helmchen · 11/11/09, 10:13 AM

Release 2.2.25
* Added default buffs for 25 man raiding.
* Added latency to the spell cast time calculations.
* Added number of targets to the options for later use.

Ferala · 11/11/09, 10:49 AM

Originally Posted by Lord Helmchen

The reason is, that checking all possible combinations would require an unbearable amount of time. Rawr uses a genetic algorithm in order to follow the more useful paths. The behavior of this algorithm can be modified a bit by changing the Thoroughness slider in the optimization screen.

That sounds reasonable!, but ..

Rawr should at least check if the score of the setup you are currently wearing isn't better than the best suggestion he comes up with. If that happens he should just suggest the setup you have right now. Of course you can always select 'Keep current setup', but it looks so silly that he comes up with a worse scoring setup after 10 mins of calculating (and no one in their right mind will ever choose the worse setup). :-)

p.s. Can i enter then random starting seed for the genetic algorithm somewhere? In that case it should make the outcome stable.

Lord Helmchen · 11/11/09, 11:00 AM

I will suggest this in the issue tracker since you have a point there

.

Levva · 11/12/09, 1:46 PM

For information, as the Elemental relic slot totems are often great for Enhancement. I've added the Elemental tier 10 totem to Rawr. It's implemented with a trigger of Trigger.ShamanFlameShockDoTTick so you will have to implement that as a relevant trigger.

Lord Helmchen · 11/13/09, 9:12 AM

Thanks. I added it.

V_lad · 11/13/09, 1:15 PM

Does current implementation takes default lvl 83 "resistance" into account?

I've tried to compare values for simple LB spam without any proc effects, hero or Elemental Mastery in-game with Rawr and Seic spreadsheet, and I get a perfect match when I add damage that was resisted. I've adjusted for crit, latency (34ms average over 3 minutes according to combat log) and specced out of overload too.

Will it be possible to add this option if it's not yet there?

Lord Helmchen · 11/13/09, 8:52 PM

It did not take the 15 points of boss resistance into account. It's default now and already in the SVN repository.

V_lad · 11/16/09, 5:05 PM

hm.. seems like people are not yet decided on whether it's 5 or 8 points per level. See this thread:

Spell Penetration

Is it possible to have a variable setting?

Lord Helmchen · 11/16/09, 6:44 PM

I actually found a function in the StatConversion class, that is used by other models. It calculates 2% per level.
I'll switch to use this one.
I guess it is written by Kavan.

V_lad · 11/24/09, 11:16 AM

any chance of adding an optional usage of demo lock buff and magma totem? that screwsup rotation with 2pt9, but it can be renewed instead of 1 LB cast. Also will help to prepare for ICC and fire nova.

Lord Helmchen · 11/25/09, 8:53 AM

I'm still unsure what would be one pass of a rotation then. Since several cooldowns and/or durations compete with one another.
Currently the Flame Shock recast marks one pass. It will be much more difficult when Chain Lightning, Magma Totem and Fire Nova will be added to the game.

Currently I would tend to output a priority list depending on the number of targets and do a small simulation until the rotation repeats itself in the background in order to calculate the actual values. However I don't like this very much, because it is not elegant and probably too slow.

Suggestions are welcome here.