Jump to content


Photo

Cataclysm Mage Simulators and Formulators


  • Please log in to reply
443 replies to this topic

#21 Jaerel

Jaerel

    Von Kaiser

  • Members
  • 86 posts

Posted 04 December 2010 - 03:34 AM

I'm currently running the sims with the option "aura_delay=0.150" which causes the sim to model both the tick-munching bug and the tick-rolling bug, so it is currently not behaving as if it's been fixed.

There was a fix to Hot Streak yesterday that was causing it to have a chance to trigger on a single Scorch crit. Not currently aware of anything else that's going on, but I'll keep looking.

ETA: Are you looking at Hot Streaks from my output, or are you running the sim yourself? If running it yourself, are you building from the latest source in the repository, or from the release? If you're using last week's release, there are many fixes that have made since that you're not seeing in your results.



I guess I'm either confusing the issue by calling both behaviors munching or if the .150 is meant to handle the refresh lockout ignite loss then it's performing poorly at high crit frequency. There should be a condition in the sim that an ignite with a duration >4 seconds cannot be refreshed, and any crits that occur while ignites duration is >4 seconds generate 0 ignite damage. Its definitely alive and well on live servers and its a pretty dramatic effect even at this first tier stat level (you'd lose ~15-25+% of the ignite damage that you'd expect, worse in the cases where ffb & scorch see heavy usage, whereas last I checked I think the sim is reporting <5% loss).


As far as hot streak goes, I had been looking at your couple days old outputs but the most recent fix seems to have brought it mostly in line with what I'd been seeing/estimating.

#22 Roywyn

Roywyn

    Bald Bull

  • Members
  • 1,398 posts

Posted 05 December 2010 - 02:41 PM

I've been able to improve the Arcane rotation a bit (~800 dps) by [...]

Something that I haven't figured out yet is the DPS Timeline. It is obviously an average, but is it taken over a time frame, a weighted time frame, over the iterations (does not look like that to me)?
http://chart.apis.go...,20&chco=2459FF
http://chart.apis.go...&chts=000000,20
What I cannot see from the graphs alone is whether Arcane Blast spam at 25% is better or worse than the Conservation Cycle at 95%. And if it is better, then by how much?

More important and more general is the following question: When ignoring Mana Adept, how does the damage D_AB := 1s of AB spam + Xs of Evocation (the mana needed for 1 spammed AB) compare to D_CS := (1+X)s of Conservation Cycle? (Here, more max. mana improves both parts - it shortens Evocation time and allows for a more powerful Conservation Cycle.) Once we have this damage ratio, we can calculate how much Mana Adept A% and current mana M% we need such that (1+A%*M%)*D_AB is more damage than (1+A%*95%)*D_CS.

This allows us to find out (given a certain amount of Mastery) whether we should drain our mana with AB spam down to 25% or whether we should stop at 40% and save one tick of Evocation. AB damage, AB mana, AB cast time, Mana Adept, etc. are just too many non-linear mechanics to determine this in general.


[Edit]: Thanks for the info. I'd like to point out another potential issue than is not fixed by smaller time slices. Image one iteration with a burn phase from 0:10 to 0:30 and a second iteration with a burn phase from 0:20 to 0:40, both followed by 5s Evocation. Averaging those two, you get a semi-burn from 0:10 to 0:20, a full burn from 0:20 to 0:30. From 0:30 to 0:35 your average mana goes up to 60% while being at half burn DPS, then drops as a semi-burn from 0:35 to 0:40 and goes up to 95% from 0:40 to 0:45 with half-conserving DPS.

We're calculating the average performance per timeslice, but the more interisting data is actually the average time at which the mana bar drops to a certain value. I don't think it's a real issue for right now since we have a lot of iterations and the mana timelines look really good. Just something to keep in mind when reading those graphs and what they mean.
Chaotic Meta Gems in Cataclysm: http://elitistjerks....p2/#post1794256

DPS spec and class comparison in Naxxramas gear: http://code.google.c...ki/SampleOutput
The Blue Bar and you - the complete Fire Mage 2.4 mana compendium: http://elitistjerks....0-post3191.html

And doesn't proc on AM.
Neither does since 3.1.

#23 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 05 December 2010 - 03:11 PM

The timelines in SimC are a bit mushy.

The course of a fight is sliced into 1sec buckets. Every time damage is done, resource is consumed, or current resource is monitored, the value is added into the appropriate time-slice. This is a running total over all iterations.

When the sim is finished iterating, the value at each time-slice is divided by the number of iterations.

The DPS timeline is populated by looking at the damage done over the period Time-10sec to Time+10sec (subject to bounds).

EDIT: Roywyn, I like seeing those graphs vertically stacked. I'll change the output so that is the norm.

EDIT2: We had a long-standing Issue in which the tail end of graphs bottomed-out do to varying fight lengths. This has been resolved with r5947.

#24 Shaewyn

Shaewyn

    Piston Honda

  • Members
  • 144 posts

Posted 05 December 2010 - 06:25 PM

This allows us to find out (given a certain amount of Mastery) whether we should drain our mana with AB spam down to 25% or whether we should stop at 40% and save one tick of Evocation. AB damage, AB mana, AB cast time, Mana Adept, etc. are just too many non-linear mechanics to determine this in general.


Does the behaviour of Evocation change after level 80? At level 80, it is only returning 60% total, not 75%. The arcane thread concurs with this - it would appear that the tooltip ingame is incorrect. This would change your stop values to 55% (skipping one tick) and 40% (using full Evocation).


[Edit:] Ahh, apologies. Because the numbers matched up with the incorrect tooltip numbers, I made a quick assumption and forgot about mage armor/replenishment.

#25 Roywyn

Roywyn

    Bald Bull

  • Members
  • 1,398 posts

Posted 05 December 2010 - 07:16 PM

Does the behaviour of Evocation change after level 80? At level 80, it is only returning 60% total, not 75%. The arcane thread concurs with this - it would appear that the tooltip ingame is incorrect. This would change your stop values to 55% (skipping one tick) and 40% (using full Evocation).

The behaviour doesn't change, going from 40% to 100% is fine in a world without mana regeneration. But you have Mage Armour, base MP5, BoM and Replenishment ticking while you cast Evocation and your first two Arcane Blasts, maybe even three if Clearcasting procs. That can easily go up to 10% of your mana, which is why you go down to ~25% before using Evocation as Arcane spec.
Chaotic Meta Gems in Cataclysm: http://elitistjerks....p2/#post1794256

DPS spec and class comparison in Naxxramas gear: http://code.google.c...ki/SampleOutput
The Blue Bar and you - the complete Fire Mage 2.4 mana compendium: http://elitistjerks....0-post3191.html

And doesn't proc on AM.
Neither does since 3.1.

#26 Nathyiel

Nathyiel

    Von Kaiser

  • Members
  • 96 posts

Posted 08 December 2010 - 08:58 AM

I am wondering why in all the frost test your not crit capped? Is it because of SimC config? Because in both 359 and 379, it can be possible to obtain the crit cap whithout Molten Armor.
I have made a proposal for a 372 Frost BiS here.

#27 Power

Power

    Glass Joe

  • Members
  • 17 posts

Posted 08 December 2010 - 10:49 AM

Frost with Frostfirebolt as filler seems to be quite good (~100 DPS less then normal Frost):
- Replace Frostbolt glyph with Icelance glyph.
- Cast Frostbolt when Early Frost is up.

Spoiler

Edit:
This was done with simc-403-7. It seems Version 403-9 broke the Early Frost Function.

Edit2:
My Fault. Version 403-9 no longer has Early Frost in the BIS Frost Sim.

#28 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 08 December 2010 - 01:01 PM

I am wondering why in all the frost test your not crit capped? Is it because of SimC config? Because in both 359 and 379, it can be possible to obtain the crit cap whithout Molten Armor.
I have made a proposal for a 372 Frost BiS here.


The set I'm using in the simulations crit-caps Ice Lance due to the set bonus. Reforging or regemming to shift additional haste or mastery to crit consistently reduces the final numbers. That doesn't necessarily mean the set can't be improved, but shifting crit doesn't seem to improve it.

Power: that action list would almost never hard-cast Frostfire Bolt; it would cast Frostbolt unless you had Early Frost, not if you had Early Frost. Change that line to:

actions+=/frostbolt,if=buff.early_frost.react

A quick run here puts it around 500 DPS lower than using Frostbolt filler.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#29 Power

Power

    Glass Joe

  • Members
  • 17 posts

Posted 08 December 2010 - 01:42 PM

Lhivera:
My action list casts 17 Frostbolts in a 300s fight. Your action list casts 0.

PS: Version 403-9 no longer has Early Frost in the BIS Frost Sim.

#30 Nathyiel

Nathyiel

    Von Kaiser

  • Members
  • 96 posts

Posted 08 December 2010 - 02:18 PM

The set I'm using in the simulations crit-caps Ice Lance due to the set bonus. Reforging or regemming to shift additional haste or mastery to crit consistently reduces the final numbers. That doesn't necessarily mean the set can't be improved, but shifting crit doesn't seem to improve it.


I have forgot that it's also a 200 spell power loss. After a bit of reflexion, if I put Food and Flask in consideration, I can obtain: 7175 ps, 1737 hit, 1204 haste, 2170 crit and 921 mastery (by reforging/regemming). I can also add +160 Intel form Enchant/Jewel. Theoricaly, With Crit > Haste, this stuff could be a dps increase.

edit: Fro the FB/FFB discussion, what about hard casting FFB for refreshing the 3 stacked DOT ?

#31 Power

Power

    Glass Joe

  • Members
  • 17 posts

Posted 08 December 2010 - 02:47 PM

edit: Fro the FB/FFB discussion, what about hard casting FFB for refreshing the 3 stacked DOT ?


I tried this (actions+=/frostfire_bolt,if=dot.frostfire_bolt.remains<3) but it didn't change the results much (-14 DPS).

#32 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 08 December 2010 - 11:02 PM

Lhivera:
My action list casts 17 Frostbolts in a 300s fight. Your action list casts 0.


Something is wrong, then. .down should only be true if the buff is not active. .up should be true if the buff is active. .react should be true if the buff is active and the player has had time to notice that it is active before choosing the next cast.

Dedmonwakeen checked in an adjustment to the EF code today, but so far it isn't making any difference in my test runs. If he doesn't see this before I get home, I'll hop on IRC and explain the weirdness and maybe he can find the cause.

Edit: I apologize -- this is what I get for replying during lunch while dealing with other things. You're saying that the profile itself has dropped EF from the spec. That's probably my fault. I'll fix it and retest.

Edit again: You're correct! I get 24650 with your actions, vs 24891 with the default Frost actions, making FFB-Frost much stronger than I was estimating a few weeks ago.

You're also correct about using buff.early_frost.down -- I had not realized that because of its unique mechanic, the normal conditional syntax is reversed, as it's tracking more like a debuff than a buff.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#33 Brainpower

Brainpower

    Glass Joe

  • Members
  • 8 posts

Posted 09 December 2010 - 04:25 AM

About playing frost with frostfirebolt, is there an amount of crit that would make this spec definitely better than the normal one because ignite works on ffb? This is something you may be able to sim. Also, does this effect the dps weight of mastery and if so is it stronger or weaker (mastery with ffb compared to fb). Could it be something viable for raiding in next tiers or would it be more like some counter-design spec that would be broken because blizzard said they wanted frost to use fb plus brain freeze becomes close to useless with this spec?

#34 Roywyn

Roywyn

    Bald Bull

  • Members
  • 1,398 posts

Posted 11 December 2010 - 09:32 PM

Looking though the Sims and results, I noticed two things on Simulationcraft Results

1) The blue gear profile has 251 haste rating, 1.96% haste unbuffed, 7.06% haste raid buffed. This is for all profiles, including Arcane. Is Netherwind Presence properly implemented? It probably is and works in a layer behind the raid buffed stats, I'm just pointing out something that might look odd.

2) The blue gear set is below the first fire haste threshold. That's probably also the reason why is shows such a large DPS value for haste rating, since adding a decent amount of haste puts it above the haste threshold. The value to reach is 516 haste rating (12.5% haste total for a 5th tick on Pyro/LB) and can likely be reached easily through reforging, the set has 1690 total haste.
You can verify th results pretty easily, just compare the Living Bomb ticks to the quadruple of the Living Bombs explosions (Pyroblast gets overwritten instead). It's around 4 for the blue set and around 5 for the other sets.

The change to 516 (or slightly above) haste rating should be made for Fire and Frostfire specs.

3) How does Early Frost finally work? It reduces base cast time by 0.6s. But how is the GCD affected? Is it reduced to a fixed value? To a lower value that is then divided by (1+haste)? It was changed quite a few times and without an update on how it seems to be working right now, we can't really tell how good or bad it is and how it will scale.


[Edit i]: To avoid post clutter: I was referring to http://elitistjerks....p7/#post1781043 which shows that Early Frost reduced the GCD to 0.8s, seemingly not further affected by other haste effects. It worked that way on live, but not on beta, so that's why I'm asking.

[Edit ii]: Early Frost grants you 0.6s/(1+haste) time every 15s, which is 0.6s/2s = 0.3 Frostbolts regardless of haste every 15s, or 0.02 Frostbolts per second. In raid gear, that's around 230 DPS or 1%, so the Simcraft results seem to be alright. There are also slightly more procs, so it's just a rough estimate.
Chaotic Meta Gems in Cataclysm: http://elitistjerks....p2/#post1794256

DPS spec and class comparison in Naxxramas gear: http://code.google.c...ki/SampleOutput
The Blue Bar and you - the complete Fire Mage 2.4 mana compendium: http://elitistjerks....0-post3191.html

And doesn't proc on AM.
Neither does since 3.1.

#35 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 12 December 2010 - 01:51 AM

Roywyn:

1 & 2) I have not yet had any time at all to try and optimize the preraid or non-heroic T11 profiles. If anyone wants to put some time into that, it'd be appreciated!

3) I'm not aware of any changes to the EF GCD behavior since it started setting the GCD to a fixed value of 1.0 secs on the cast. It appears to be consumed at start of cast rather than end (probably to prevent "ghosting" the buff), and the cooldown appears to start then as well.


Update: So, this is kind of funny.

It turns out Arcane Power wasn't working.

Arcane now looks almost exactly equal to Frost. Which is still not quite ideal, because (a) Fire is still much higher than both, and (B) Arcane doesn't degrade as well as Frost as we deviate from Patchwerk parameters, but it's still considerably better than it used to be.

New results running now, should populate the directory over the next hour or so.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#36 Power

Power

    Glass Joe

  • Members
  • 17 posts

Posted 14 December 2010 - 11:32 AM

Lhivera:
I think I found a small improvement for your Frost sims:

Frostspec with Magearmor is mananegativ.
If you switch to Magearmor at 15% mana, the sim will run OOM (<5%) and use Evocation.
If you switch at 30% (300 sec Fight) the Frostsim will not use Evocation and gain ca. 200 DPS.

#37 Kavan

Kavan

    Bald Bull

  • Members
  • 1,018 posts

Posted 15 December 2010 - 02:49 AM

Has anyone investigated the impact of Power Torrent on arcane? I don't model all the details currently in Rawr, specifically I might be overpenalizing the negative impact on mana adept since when you're hovering close to max mana you could potentially go above normal max and negate some of it. But without that I'm getting that the enchant actually lowers your dps. Can someone do a comparison between Power Torrent and the LK straight spell power enchants?

EDIT: Ignore above, I had an error in calculations.

#38 Cuhjo

Cuhjo

    Glass Joe

  • Members
  • 7 posts

Posted 15 December 2010 - 10:44 AM

In Lhivera's sims frostfire uses the same logic as fire, namely cast scorch if mana <95% and evo CD > 60s. How would it effect frostfire dps if the sim would try to keep up 3 stacks of FFB DoT?

#39 Venthos

Venthos

    Glass Joe

  • Members
  • 24 posts

Posted 15 December 2010 - 11:08 AM

Something I noticed in arcane's priority list is that presence of mind is simply being activated for any arcane blast when it comes off cooldown/ gets to priority, rather than being saved for the highest stack, and/or right before popping AP and Gem. Is this intentional, and gave a dps increase not to save it, or was it simply an awkward thing to code?

#40 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 15 December 2010 - 12:17 PM

Kavan: I checked it anyway (or rather, Dedmonwakeen checked it for me), and it looks like SimC is not currently adjusting max mana with Int procs, so it's not currently measuring that.

Cuhjo: Yes, that's correct. It's definitely worth some experimentation.

Venthos: The original priority list saved it for the highest stack; removing that condition increased DPS a bit, rather than decreasing it -- I expect because during mana maintenance phases, POM never got used, and increased usage outweighed the increased crit. I have not tried using it in conjunction with Mana Gem; that's worth trying out (I did try delaying AP to line up with Mana Gem, that was a loss, not a gain).

Power: Yup, looks like a gain at both 5mins and 7mins; I've made that change. Thanks.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users