Jump to content


Photo

SimulationCraft for Shamans


  • Please log in to reply
135 replies to this topic

#21 Zimeron

Zimeron

    Cookies!

  • Members
  • 423 posts

Posted 19 November 2009 - 01:20 AM

You can add in the option for CL to only be cast when you have CC up, but it's not enabled by default. Over the course of 1000 iterations, assuming BIS gear, you do more dps spamming CL and possibly going oom at very end, then only using CL when CC is up.

Also, the sim knows your mana levels and knows the boss health percentage, so adding an option to only cast CL when your mana percentage is higher then the bosses is something that can be done.

#22 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 19 November 2009 - 01:29 AM

I'm in the process of re-vamping our action conditionals such that all sorts of information will be dynamically available:

...
actions+=/chain_lightning,if=(mana_pct>target.health_pct)|buff.clear_casting
...

#23 Zimeron

Zimeron

    Cookies!

  • Members
  • 423 posts

Posted 19 November 2009 - 02:56 AM

I'm in the process of re-vamping our action conditionals such that all sorts of information will be dynamically available:

...
actions+=/chain_lightning,if=(mana_pct>target.health_pct)|buff.clear_casting
...


Until then, I added an option to chain lightning called "conserve", which will only cast when your mana percentage is greater than the bosse's health percentage. It resulted in a net gain for simming my own character and no change, as expected, in the BIS profiles. It's not perfect, but it's the closest we have to player's intuition at the moment.

If you build SimC yourself, it's in 3859, otherwise you'll have to wait till dedmonwakeen releases the next build.

#24 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 19 November 2009 - 08:37 PM

SimC v9 (r3883) is available for download. This includes a conserve=1 option by default on chain_lightning actions.

#25 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 23 November 2009 - 03:15 AM

The wiki page generation has been overhauled to provide information more clearly ala html output from the gui.

#26 pfg

pfg

    Glass Joe

  • Members
  • 18 posts

Posted 31 December 2009 - 02:35 AM

dedmonwakeen if you could help me a little it would be much appreciated. I have been given a job of showing the raid dps difference with flame tongue, totem of wrath and demonic pact / meta lock spell power buffs. If you could point me in the direction of the files I would need to modify that would be awesome. What I'm thinking is have three versions of the magic damage classes (mage, warlock, elemental, boomkin etc) and then put together a spreadsheet that has our raids number of the above. If there is a better way to do this then I would be interested to hear as well.

#27 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 01 January 2010 - 02:07 PM

I believe you also need to measure the opportunity cost of using Flametongue instead of Searing, etc.

We ship a ton of example profiles. You can combine them in one run.

You could type the following in the Simulate tab:
Shaman_T9_57_14_00_258.simc
Mage_T9_57_03_11.simc
Warlock_T9_55_00_16.simc

And then do another run in which you replace the Shaman with Shaman_T9_57_14_00_258_ToW.simc

Similar idea for DP, etc.

EDIT: You will have to make sure that you remove the appropriate buffs in the Options tab.

#28 pfg

pfg

    Glass Joe

  • Members
  • 18 posts

Posted 02 January 2010 - 12:27 AM

Well some testing done but... I read more on the examples and found the guild import so was using that with the following profile.


# Downloading guild characters of certain rank(s)
guild=Zephyr,region=us,server=Proudmoore,ranks=0/3/4/5/6/7

# Saving simulation output to a text file
# output=Zephyr_output.txt

# Saving simulation output in both text and html format
output=Zephyr_output.txt
html=Zephyr_output.html

# Calculate scale factors (default is off)
calculate_scale_factors=0

# Specifying enough iterations for scale factor generation (defaults to 1000)
iterations=1000

# Changing the length of the fight in seconds (defaults to 300)
max_time=300

# Automatically varying the the length of the fight +/- 20%
vary_combat_length=0

# Turning on specific raid buffs after they have all be turned off
optimal_raid=0
totem_of_wrath=0
flametongue_totem=1
abominations_might=1
arcane_brilliance=1
arcane_empowerment=1
battle_shout=1
bleeding=1
blessing_of_kings=1
blessing_of_might=1
blessing_of_wisdom=1
blood_frenzy=1
bloodlust=1
bloodlust_early=1
celerity=1
crypt_fever=1
curse_of_elements=1
divine_spirit=1
earth_and_moon=1
elemental_oath=1
expose_armor=1
faerie_fire=1
ferocious_inspiration=1
focus_magic=1
fortitude=1
heart_of_the_crusader=1
heroic_presence=1
hunters_mark=1
improved_faerie_fire=1
improved_moonkin_aura=1
improved_scorch=1
improved_shadow_bolt=1
infected_wounds=1
insect_swarm=1
judgement_of_wisdom=1
judgements_of_the_just=1
leader_of_the_pack=1
mana_spring_totem=1
mangle=1
mark_of_the_wild=1
master_poisoner=1
misery=1
moonkin_aura=1
poisoned=1
rampage=1
replenishment=1
sanctified_retribution=1
savage_combat=1
scorpid_sting=1
strength_of_earth=1
sunder_armor=1
swift_retribution=1
thunder_clap=1
trauma=1
trueshot_aura=1
unleashed_rage=1
windfury_totem=1
winters_chill=1
wrath_of_air=1

# Simulating something other than a tank-and-spank requires use of the raid_events option
# If cooldown_stddev and/or duration_stddev are not specified, they are assumed to be 10% of cooldown/duration values.

# Every 30sec, the boss is casting something that MUST be interrupted, prompting players to use their interrupt skills
# The duration of the cast is 3sec and the first one occurs 15sec into fight
# raid_events+=casting,cooldown=30,duration=3,first=15

# Every 30sec, all players are required to move for 5sec
# raid_events+=/movement,cooldown=30,duration=5

# Every 40sec, all players within 5yards are required to move for 7sec
# raid_events+=/movement,cooldown=40,duration=7,distance<=5

# Every 60sec, all players are stunned for 2sec
# raid_events+=/stun,cooldown=60,duration=2

# Every 2min, the target becomes invulnerable, interrupting players and shedding all DoTs
# This can be used to model target-switching
# raid_events+=/invulnerable,cooldown=120,duration=3

# Every 1min, 3 additional targets are nearby the boss and remain for 20sec
# raid_events+=/adds,count=3,cooldown=60,duration=20

# Every other 45sec, all players are distracted suffering a temporary 20% skill drop
# raid_events+=/distraction,skill=0.2,cooldown=90,duration=45

# Specify the patch level to match PTR (defaults to 3.2.2)
# patch=3.3.0

# Adjust lag values (in seconds)
# They represent the different delays depending upon the preceding action.
# channel_lag=0.250
# gcd_lag=0.150
# queue_lag=0.075

Then changing the top few lines on a case by case basis. That seems to work except the numbers are off by allot. By no means am I expecting what we get on the logs but when the warriors are on the bottom instead of the top I have to wonder am I doing something totally wrong.

The other item, that you mentioned, is how do I change that action list to use magma etc ? This looks very promising and thanks for the help so far.

#29 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 02 January 2010 - 02:52 PM

Remember that SimC models a simple tank-n-spank by default.

It has options to model movement, nearby adds, etc. However, we make no claims about modeling "real" fights.

The only time I see warriors at the top is if they have some opportunity for "nearby target" abilities.

I would experiment with single actor sims until you get comfortable with the tool. The learning curve is a bit steep I'm afraid.

#30 Zogar

Zogar

    Von Kaiser

  • Members
  • 27 posts

Posted 01 February 2010 - 01:21 PM

I tried simulation craft, and was disappointed. First the mana problem is very weird (I removed CL from the rotation, but i'm still getting oom way too fast). And I get the same nonsense high EP values for crit, not using fire totems and making sure I don't get oom. Has there been an explanation for this ?

#31 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 01 February 2010 - 02:32 PM

I tried simulation craft, and was disappointed. First the mana problem is very weird (I removed CL from the rotation, but i'm still getting oom way too fast). And I get the same nonsense high EP values for crit, not using fire totems and making sure I don't get oom. Has there been an explanation for this ?


When you click "Simulate!" a simc_gui.simc file is created which is passed to the simulator which in turn creates simc_report.html.

Package those up with a detailed description of your problem and I can turn around explanations and/or fixes in less than a day.

#32 Zamir

Zamir

    Piston Honda

  • Members
  • 197 posts

Posted 01 February 2010 - 03:19 PM

Does SimC avoid "stacking" haste effects for elemental or do things like EM and Heroism get used together if they happen to coincide?
Zamir's Blog // Cataclysm Elemental Primer // TotemSpot Community

Every one of us is, in the cosmic perspective, precious. If a human disagrees with you, let him live. In a hundred billion galaxies you will not find another.

-- Carl Sagan


#33 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 01 February 2010 - 04:07 PM

Does SimC avoid "stacking" haste effects for elemental or do things like EM and Heroism get used together if they happen to coincide?


If I can change your question slightly....

"Does the default action list generated for Elemental Shaman avoid "stacking" haste effects with Bloodlust?"

I just checked an there is no conditional logic associated with the "elemental_mastery" action. This would certainly be easy enough to add. I think we have been getting "lucky" to some extent by virtue of the fact that EM is used on CD and the default behavior of BL is to trigger it in the last 60sec or so.

How would you describe the optimal usage of EM? I will convert that into a conditional and update the default actions generated.

#34 Zamir

Zamir

    Piston Honda

  • Members
  • 197 posts

Posted 01 February 2010 - 05:07 PM

If I can change your question slightly....

"Does the default action list generated for Elemental Shaman avoid "stacking" haste effects with Bloodlust?"

I just checked an there is no conditional logic associated with the "elemental_mastery" action. This would certainly be easy enough to add. I think we have been getting "lucky" to some extent by virtue of the fact that EM is used on CD and the default behavior of BL is to trigger it in the last 60sec or so.

How would you describe the optimal usage of EM? I will convert that into a conditional and update the default actions generated.


Unfortunately it's a bit complicated. All things being equal, it makes no difference whether haste effects are used in series or in parallel. Three factors make all things not equal:
  • Maximising effect usage (uptime)
  • Maximising benefit from secondary effects (e.g. on-use trinkets/procs/pots, temporary damage buffs)
  • The 100% haste "hard cap" for elementals

With expected haste levels, (3) only really comes into play when stacking a haste effect with BL. (1) is probably more important than (3), so it'd be better to use EM during BL than not at all, except that if BL alone is enough to cap haste then EM provides zero benefit. The interaction of (2) with (1) and (3) is rather more complex.

What would be great is if SimC could be configured by the user to answer these questions - but I fear that may be a bit difficult?

I guess an option to make sure that EM/berserking are not used during BL would be good, assuming that with the actions+=/bloodlust,time_to_die<=59 setting it's used as close to 59 seconds as possible (to allow for time for EM to be used after BL fades). In pseudo-code:

IF BL active AND [ ( Fight time remaining ) minus ( BL time remaining ) ] > EM duration THEN save EM ELSE use EM at time_to_die<=17
Zamir's Blog // Cataclysm Elemental Primer // TotemSpot Community

Every one of us is, in the cosmic perspective, precious. If a human disagrees with you, let him live. In a hundred billion galaxies you will not find another.

-- Carl Sagan


#35 Zimeron

Zimeron

    Cookies!

  • Members
  • 423 posts

Posted 01 February 2010 - 05:21 PM

IF BL active AND [ ( Fight time remaining ) minus ( BL time remaining ) ] > EM duration THEN save EM ELSE use EM at time_to_die<=17


These actions would model what you want:
actions+=/elemental_mastery,time_to_die<=17
actions+=/elemental_mastery,if=!buff.bloodlust.react

When I last revamped the action lists in prep for BIS T10 profiles, I tested adding in not casting EM during Bloodlust, but it resulted in a few dps lost compared to using it on cooldown. Adding the first conditional should solve the problem of losing dps.

I'll test it and commit it when I get home, if Nate doesn't beat me to it.

#36 Rouncer

Rouncer

    Deeper Shade of Blue

  • Allied Members
  • 4,132 posts

Posted 29 November 2010 - 07:47 PM

I was looking at the outputs someone posted on the D/D forums and I was wondering how you have flametongue set to scale for enhancement in the level 85 model.

On live right now it's "double-dipping" by scaling from spellpower as expected and then by a separate coefficient off of AP. I personally doubt that is the model they are going to stick to since it makes FT/FT with a caster mainhand too tempting but I could see them getting it to scale with the AP and then turning off FT's spellpower scaling for direct damage altogether.

Anyway I was just curious if that output I saw on the D/D was with the AP scaling and proper scaling with enhancement mastery in place or not.

#37 Zimeron

Zimeron

    Cookies!

  • Members
  • 423 posts

Posted 29 November 2010 - 08:15 PM

We're simulating Flametongue with a spell power coefficient of .1539 and for enhancement specializations an attack power coefficient of .04611. Flametongue is benefiting from mastery as well in the sim.

#38 Rouncer

Rouncer

    Deeper Shade of Blue

  • Allied Members
  • 4,132 posts

Posted 29 November 2010 - 09:06 PM

We're simulating Flametongue with a spell power coefficient of .1539 and for enhancement specializations an attack power coefficient of .04611. Flametongue is benefiting from mastery as well in the sim.


If enh specced is it getting both scalings or just the AP one? Dedmon pmed me the code and it looks like both but just want to confirm. Also where did you get the 0.04611 AP coefficient? Is it just from experimentation on the Beta or is there something in the client that shows that value for the scaling? I've been getting a range of coefficients in testing from as low as 2.8% up to 6.5% when I was testing it yesterday. Seems to change values based on how much AP is involved but I'm not sure where to start to try and work out the formula. I think the 4.611% looks about right for testing in the pre-made 85 raider's gear but don't have my notes with me to confirm.

#39 Zimeron

Zimeron

    Cookies!

  • Members
  • 423 posts

Posted 29 November 2010 - 10:06 PM

If enh specced is it getting both scalings or just the AP one? Dedmon pmed me the code and it looks like both but just want to confirm. Also where did you get the 0.04611 AP coefficient? Is it just from experimentation on the Beta or is there something in the client that shows that value for the scaling? I've been getting a range of coefficients in testing from as low as 2.8% up to 6.5% when I was testing it yesterday. Seems to change values based on how much AP is involved but I'm not sure where to start to try and work out the formula. I think the 4.611% looks about right for testing in the pre-made 85 raider's gear but don't have my notes with me to confirm.


Enhancement spec gets both the AP and SP scaling in the simulator. The testing was done on live and the coefficient was found through testing. I asked the dev who did the testing and he didn't notice anything along the lines of the coefficients changing with AP, but he only tested up to 9.7k AP.

#40 Rouncer

Rouncer

    Deeper Shade of Blue

  • Allied Members
  • 4,132 posts

Posted 29 November 2010 - 10:56 PM

Did he test in low AP conditions as well? Very strange that he didn't notice the coefficient shifting, I'll try to do more testing on it later tonight.

EDIT
Just did a bunch of testing and checked my values for the previous testing I had and your values correlate really well. Thank you very much for the help, saves me a lot of mental energy that is better spent on beer and BGs.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users