![]() |
SimulationCraft - for fur and feather wearers
Quote:
|
Do you happen to have a list of what things are currently not modeled (e.g. Starfall, Treants right before Heroism) and their statuses (e.g. Not developed, Not possible)? While I love using SimulationCraft for all my modeling needs, it would be nice to have a comprehensive list of what its known weaknesses are, outside of the theoretical vs. practical argument.
|
When i tested it last it seemed to give coherent values, but i haven't run detailed comparison with spreadsheets.
One thing i'd like to see though is the ability to use dots at different points in the rotation. It is for example widely accepted that you don't want to have 100% dot uptime if that means refreshing them during/at the 2nd half of an eclipse proc. There's also some debate as to wether we should clip dots at precise points (maximizes uptime but loss of GCDs) or just care less about dot uptime. The current behavior is enough to give a rough idea of where we'll land in terms of dps but doesn't allow us to test the result of dot management possibilities. (would be nice if the current action list system would allow for such options. If not, then too bad, but it's still a nice tool) |
Quote:
/starfire,eclipse=benefit at the front of your action list (assuming Lunar Eclipse). I don't believe there is currently any support for clipping your DoTs, or looking at actual timers. I would imagine that syntax for those options would be possible. Something like /moonfire,eclipse_cooldown=3,clip=7 to mean refresh moonfire during the last 3 seconds of the eclipse cooldown, but only if you are clipping no more than 7s of an existing MF. The Warlock 2/13/56 configuration file has a cast which is conditional on Bloodlust being active. I don't know if that syntax works for Druid action lists. |
Quote:
Treants just before bloodlust would need some tweaks in the sim itself i think, so it announces a few seconds earlier that bloodlust will be fired. I made a wikipage for druids once, but i think it is a bit outdated, I'll look into updating it. Edit: Well, actually you can throw treants just before bloodlust, with the "health_percentage<=XX" option, because bloodlust is fired when the mob reaches 35% hp. Quote:
just something simple: skip if eclipse active? or more precise like: skip if eclipse > x seconds? Quote:
bloodlust=1 ONLY while under the effect of bloodlust and with one of the latest commits bloodlust=-1, the opposite, only when not under the effect of bloodlust. |
It depends on if we just want to model current "accepted" rotation or work out what rotations could be best.
Currently most moonkins do : no dot clipping, refresh MF once in per cycle (usually right before Eclipse ICD is finished though there are other options) and IS twice (right before Eclipse ICD is finished and at the end of the proc, as soon as the next SF won't benefit from the effect anymore). If we want to be more flexible, yes it'd have to be a condition based on how far in the Eclipse proc we are (refresh before x, don't refresh after). In a research scenario, well, we'd need more options, concerning MF specifically. IS is short enough to find an easy place in the cycle. |
Quote:
Implemented new options for IS/MF with r1587 and r1588:
skip_on_eclipse=-1: If an eclipse is up, only cast the dot if the eclipse matches you, MF = +30% SF eclipse, IS = +30% Wrath eclipse skip_on_eclipse=1: If any of the two eclipse buffs is up, skip the cast For example: IS, but only when no eclipse is up .../insect_swarm,skip_on_eclipse=1/... MF, not when +30% wrath eclipse is up, and on +30%SF eclipse only when at least 10 seconds are left .../moonfire,skip_on_eclipse=-1,eclipse_left>=10/... |
New windows-executable for download is up at Downloads - simulationcraft - Google Code
Has everything working I mentoined in the thread. PTR: Everything except PTR Patch 3.1.0 - Build 9658 changeset is in. If you wanna run the sim with PTR changes just add patch=3.1.0 in the config, currently "patch=3.0.9" is the default I encouraged the projectleader to make at least weekly releases, because i know there are many people who would use it if they don't had to checkout the source and compile it. Hope the downloadable executable for windows will from now on be more up2date for you :) -------------- Ok, so everything from the PTR is in, also the new Nature's Grace and i ran SimulationCraft with the standard PTR profiles Gear: Code:
gear_stamina=1000Profiles: Druid_T8_58_00_13_SF http://simulationcraft.googlecode.co...13_SF.simcraft Druid_T8_58_00_13_W http://simulationcraft.googlecode.co..._13_W.simcraft T8 here refers to PTR mechanics, SF/W is the main nuke Output: Code:
SimulationCraft for World of Warcraft build 3.1.0 (iterations=5000, max_time=300, optimal_raid=1)10 Stars + 10 Splash damage = 20 hits per Starfall, also it's set to skip if eclipse buff is up. So you have 3-4 Starfall casts in a 300 sec fight. :) |
http://simulationcraft.googlecode.co...16-windows.zip
This includes all the changes from PTR Patch 3.1.0, including Build 9658. IS spellpower coeff => 0.2 per tick New NG mechanic Starfall cd => 90sec Fearie Fire duration => 300sec, with this i also deleted the _FF profile, which should just show how much of an impact it was when a moonkin had to deliver the 3%-hit-debuff. I don't think 1 gcd every 300 seconds is worth an extra profile. If some ferals could look at the rotation used for the druid in Druid_T8_00_55_16.simcraft, if it is reasonable or way off, feedback is welcome. CatForm dps should be rather complete now. Latest downloadable executable, with all profiles run on patch=3.1.0 (on detailed reports only druids are shown) Code:
DPS Ranking: |
Previously, you did 2 simulations for Moonkins with one having the Moonkin applying Faerie Fire and one without. Could you do the same for cat dps except with the mangle/trauma debuff? I'm trying to work out why your numbers seem so low for feral dps compared to Nightcrowler's, Toskk's, and Rawr's numbers. They all had feral dps at ~5500 dps on live. With the changes on the PTR so far, they were seeing over 6000 dps sustained. Does the simulator take into account the way the new sunder/ff interact with the new boss armor values? Also, does it take into account the +25% ArP and +30% melee haste from the latest build? Sorry for all the questions. Trying to figure out why the numbers seem off when compared to other people's numbers.
Edit: I finally got to check out the spec used for feral dps. 2 points should be moved out of Furor preferably into Improved Mark of the Wild for a boost in stats. Also, why does TickCrit have 2 values? Is it lower and upper bound instead of average TickCrit? The recommended glyphs are rip, shred, and savage roar instead of the default shred, mangle, and savage roar. Is it possible to have the model use the agility food instead of fish feast? Edit 2:In the line, actions+=/rip,cp>=5,rip<=6, is rip clipping itself or is the simulation waiting for rip to fall off first before reapplying it? Also, I may be wrong, but it seems the shred glyph maybe modeled improperly? It has 10.0 applications with 105 ticks. That would be an average of 10.5 ticks per rip, putting its average duration at 21. Shred glyph adds at most 6 seconds (3 ticks) to the duration of rip, and that's if you get 3 shreds per rip in. Base duration of rip is 12 seconds. 12+6 =18 second duration at most. I'll try and look at the source code tomorrow morning if I see any other discrepancies. Edit 3:The line for rake is odd. It only rakes if there is less than or equal to 4 combo points. There also is no parameter to check rake duration. I think it would be more ideal to rake if rake has fallen off instead of when there is less than or equal to 4 CPs. Edit 4: Hrm, it looks like berserk, the 51 pt talent hasn't been modeled yet? Or is it tied into something else for the time being and just works behind the scene? Edit 5: Last edit before I go. Are the base melee stats in T8_Melee_Base just stats purely from gear which is then modified by talents and buffs? And I assume since it's only hunters, rogues, and druids at the moment, that AP includes AP from agility? Also, do druids get the base +160 AP from being in cat form? Finally, the stats displayed on the break down under core stats. Are these values buffed or unbuffed? |
Quote:
+30% haste, +25% ArPen are both in. target_armor=10645 is used in 3.1.0 simulations. Quote:
Multiple values always refer to Average|Max. Glyphs will get changed, food also. Quote:
actions+=/rip,cp>=5,rip<=6 It's more like, when the druid stops to spend combo points on other finishers. So if rip<=6sec he won't spend any cp on ferocious_bite. Quote:
Quote:
Quote:
I'll post a new sim with your changes when I'm ready with them ----------------------------- New druid profile according to your suggestions: Code:
#!simcraft max_time=300 optimal_raid=1Code:
|
- Use rip glyph instead of mangle glyph. Also, change the mangle refresh to 1.
- Does your formula use any kind of energy pooling? The mangle uptime is low and I can't tell if there's advanced prediction energy holding, but a common decision a feral makes is "I have 45 energy and mangle expires in 3 seconds, do I shred?" The answer is no, but the simulated feral would do it. Ferals also want to pool energy if they have 5 CP instead of wasting the combo point. The code for shred might be: Code:
actions+=/shred,berserk=1- every DPS feral will have Feral Swiftness, at the expense of Feral Aggression or Improved Leader of the Pack. |
Quote:
|
Given shred glyph, rip glyph and T7 bonus, 13 tick rips are possible.....
rip and rake cannot be clipped without beating on their ready() methods pretty hard.... So.... that rip<= conditional in the rip attack isn't doing anything. Given the I've moved faerie_fire and mangle to the Globals_T8.simcraft on the assumption that a feral tank will be keeping them up...... but it would be nice to have a Cat profile that manages these just so we can see how the AI changes. EDIT: The level-based AP is included: Code:
double druid_t::composite_attack_power() |
Sorry about that Starfox, I hadn't seen your reply to Ramenchef before I posted. That teaches me to preview my posts!
I think I misunderstood (and still am not quite clear about) how that profile works-- I looked at the Google Code page and didn't see any advice on what was going on... You said it "crawls down until it finds something it can do" in the first post here, so I assumed it just processed the whole thing trying to find an attack it liked. But to Ramenchef you're saying, it'll wait at a line if it meets the condition but the debuff isn't down yet? Specifically: -Savage Roar-- once you get 4 seconds away from the debuff dropping, it doesn't do any other attacks? Or does it only stop finishers (i.e., it will refresh mangle/rake, or shred if it gets to 42 energy)? -Mangle-- when it's 2 seconds away from expiring, it stops doing other CP-generating attacks? (I.e., Rake expires in 1 second, mangle in 2; you have 42 energy-- does it shred immediately/mangle after energy regen, rake in 1/mangle in 2, or mangle in 2/rake in 3?) Some other advice, regardless of the answers to the above: -Ferocious bite consumes all your energy, every time you use it (and converts extra energy to damage, but poorly.) So, you don't want to use it above 40-45 energy (when berserk is down), and never when berserk is up (lag will make it waste too much energy). -Something needs to be done about rake uptime-- it's applied every 12 seconds in the mangle model, which is below what most ferals average, and given the DPE, you can probably reduce shreds. Increasing rake uptime should show an increase in damage. - DOT tick crits: These are consistently about 5% above the crit percentage of normal attacks. Is this how it works in-game? If it hasn't been researched, I'd probably go with the assumption that they suffer the same crit penalty that normal attacks do. |
| All times are GMT -4. The time now is 4:06 AM. |
Forum Infrastructure by vBulletin 3.6.12 ©2000-2007, Jelsoft Enterprises Ltd.