Altered Time - your #1 World of Warcraft Mage resource!

This thread is intended to serve as a place for people to leave their comments in on the current state of the Mage Module for Simulationcraft. It can also serve as a sort of direct line of communication between the community and the people who work directly with and on the module.

Things like APL discussion, improvements to be made to the default profile gear (BiS), and any sort of bugs that are found in the module should be post here for discussion.

Additionally, things like major discrepancies between sim outputs and in-game logs can be discussed here, but they'll need to be supported by a link to the logs being discussed, sim outputs, and the post should try to steer discussion towards a specific problem. Something like "frost sims are doing too much damage" is probably not appropriate; but "frost sims seem to be producing 10% more FoF-IL damage than in-game logs" with a link to some logs to support that, might be.

Below is a link to my directory for Simulationcraft. In it you can find the current Highmaul Only gear profiles, latest Trinket sims for Fire/Frost, talent comparisons for the two specs, racial comparisons, and multi-target sims. These can serve as "default" outputs for people to work from.

http://downloads.simulationcraft.org/Frost/" target="_blank

If you're new to simulationcraft and want to get involved, a good place to start learning is my guide here:http://altered-time.com/forum/viewtopic.php?f=2&t=620

Recently I've been looking into how we can improve banking in SimC for BF and FoF - mainly to see if it actually is a DPS gain or not.

I started by doing both FoF and BF banking for Mark of the Frost Wolf.

Here are the relevant changes to the APL for Ice Lance and BF-FFB use:
The end result after my attempts is this:



Banking:



Non-Banking:



So it's pretty clear that it's DPS loss here. FoF-ILs are being munched / not produced for the banking strategy.

As for why BF-FFB munching occurs, I'm not sure. Others have questioned this as well. I can see why FoFs would be munched. Say you just cast a BF-FFB, FoF procs, but you've already queued a frostbolt - and it also procs FoF. That munches a proc (or reverse the spell order). BF procs though should be much less open to munching (which, I think we see in the sim). Robomage shouldn't have these problem though, since I'm letting him cheat. In reality though, people absolutely will have this problem.

So, I then went on and just let robomage use FoF-ILs like normal, but bank BF-FFBs for MotFW procs.

Relevant changes to APL:
End result:



Banking:



Non-Banking:



Everything returns to being equal, except BF-FFB executes. To me, this is telling. it means we don't want to bank procs for MotFW! You actually want to be casting frostbolt more than anything else during these periods of time, because the damage gain from procing an additional BF-FFB (or sometimes procing one, which is what that fractional BF-FFB execute is representing I think) is worth as much as the damage gain from banking a BF-FFB to use during the period of higher MS.

The HTML outputs are both contained in the link in the OP to my SimC directory online.

AoE thresholds in current APL.
Currently on 4+ targets and robomage starts AoEing, this 4 target threshold just works with Frostbomb. I've tested some AoE stuff with robomage:

With Frostbomb:
4+ targets: start AoE rotation, while using Waterjet before blizzard. This is a little increase on 4 Targets, almost equal on 5Targets and a negative dps increase on 6+ targets. Depending on simcraftbuild, you need to change the default WJ on opener part:
This is the aoe rotation. active_enemies<5 is recomended, may need to change that, for testing purposes.
Without Frostbomb:
Well, IN and UM does not scale very well. On less then 6 targets any AoE rotation should be a dps loss, compared to singeltarget rotation. I'm not interested in figuring out the treshold for AoE with these talents.

Without TV:
CmS should not make any problems. PC results are weird, not really interested to debug / change em.

I've talked allready with komma about the current AoE-rotation. A correct APL for every talent on AoE would require a lot of duplicated lines, which makes it even harder to read. I would recommend a working apl just for frostbomb + tv / CmS. (PC is ... a little bit toooo anyoing)

Results: http://www.file-upload.net/download-100 ... E.rar.html" target="_blank


1-3 Targets, Frostbomb:

Currently robomage does not cast Frostbomb with less than 2 charges of FoF. He even accepts to cast IL without Frostbomb, if procc runs out. This is actually ~200 dps better then casting it with 1 Charge. Results vary slightly depending on gear, e.g. with lowgear on 3 targets there may be some cases where it is equal.
err... as komma mentioned, I've failed here. Basically, casting frostbomb allways, when you've 1 charge of Fingers of Frost should be a dps loss. The following line handels frostbomb casts for 1 charge of IL - if IL runs out. You're able to cut that part out of the rotation and achieve "equal" dps.
TL;DR: this part of the Apl is fine for 1 to 3 targets.
Edit: Forgot to mention endless IV on AoE Situations:

APL changes: Basically, just extend IV before chaining Blizzard. It actually includes the waterjet changes listed above, to check it vs 4 targets. Results:
4 Targets: dps loss. No go.
5 Targets: Equal ~ 86 dps increase
6 Targets: Equal ~40 dps behind

And here is the interesting part:

"How long should I keep IV up?"

Varying booleans did not change anything, e.g. the following "changes", to control how long IV will be extended, had almost no influence for DPS. "cooldown.icy_veins.remains>60" yielded a quite good result. Actually, just the amount of targets for improved Blizzard mather. 5+ and you're fine.

This has a big impact on actuall raid dps. E.g. Chogall p4. You're able to keep IV up endlessly in that phase, by spamming nonbuffed IL's onto chogall, waiting for 10 adds in range. Downtimes in addcount do not change our AoE dps :--)
Results: http://www.file-upload.net/download-100 ... V.rar.html" target="_blank

WJ on Pull with PC + IN

Did some math on it, which shows on a simplified problem, that WJ on pull is a dps increase. Fuck simplified shit, here comes simcraft:

With IN at 250k iterations it is equal. I can not really give you a advise here. More detailed: you'll see on both IN profiles a dps error of 8 and a dps difference between both profiles of 14 dps. Double the error : 16 > 14. I would need more iterations to give a clear answer. However the variance / normal distribution points out, that using WJ on pull increases the dps range significantly, but lowers min / max posibilities. Simcraft fails to provide tools to analyze dps range. I can not tell, if WJ on pull just cuts "bad"-rng partly or if it is just worse on average.

WJ on Pull with PC + FBomb

With FB the playstile is... horrifying. Anyway it is definitly a dps increase.

For both IN + FBomb I needed to rewrite the APL.
The basic idea of that APL is to:
- Frostbolt + WJ usage
- dump IL's to 1 charge before PC.
- if there is some spare time: dump FFB's to 1 charge before PC, trying to avoid FFB munching from frostbolt casts.
basically: keep 1 charge of FoF and 1 or 2 charges of FFB for PC.

Mostly I had to add 1 Line to somespells, which says: Ignore this spell, if PC is ready. Full list of changes: It should be mentioned, that without the above codehack inside WJ-APL results differ alot. Robomage likes to munch ILs, if you try to pool em. Used a fixed fighttime of 80 seconds to debug it easier. Keep in mind, that with the above changes robomage just pools for the first usage of crystal. Changing fighttime to anything longer then 80 second lower the impact of these lines, anyway the results stay roughly the same.

Results btw: http://www.file-upload.net/download-100 ... l.rar.html" target="_blank

P.S. The waterjet codehack actually increases dps for every profile. Results above were created without these line inside comparing profiles. It does not change much, lowers the gap between Fbomb exampels. That little line actually just tells robomage: Inside waterjet, frostbolts which hit the boss, will give a FoF charge, which I can dump instantly, with spellque system. No need to think for robot, no need for reaction times.


Edit:
Rerun sims with Waterjet codehack inside comparing profiles.Furthermore for Frostbomb I've changed precast to Frostbomb aiming to get this rotation done:
1. Frostbomb precast
2. Frostbolt into WJ
3. Dump to 1 IL charge
4. PC / Orb burst as usual.

I actually failed a bit with robomages rotation, that is the reason, why I'll not share the profiles. On a ~450 second fight time, the dps gains / loses from "perfect" openers are below 100 dps. Based on roughly ~32k dps averages. Waste of time to do more work here.

Anyway, results show a slight dps loss for IN rotations and a okayisch dps gain for Frostbomb.Again, results are from sims with 80 sec fighttime, 0 variance, 250k iterations. Keep in mind, that by using PC after WJ you're kind of delaying PC. Depending on fighttime this result in a slight dps loss, even for Frostbomb. (E.g. 400 sec fighttime, 20% variance = dps loss, 450 sec fighttime, 20% variance = dps gain.)

Results: http://fs2.directupload.net/images/150109/uf8kicop.png" target="_blank

With Frostbomb:
4+ targets: start AoE rotation, while using Waterjet before blizzard. This is a little increase on 4 Targets, almost equal on 5Targets and a negative dps increase on 6+ targets. Depending on simcraftbuild, you need to change the default WJ on opener part:

This has been fixed I think!

This is the aoe rotation. active_enemies<5 is recomended, may need to change that, for testing purposes.

So essentially this is a cleave APL. I wonder about the positioning for CmS and BF-FFB. Also, Wouldn't Ice Nova be prioritized higher than CmS? Also, Ice Lance's positioning should heavily depend on whether Frost Bomb is being used instead of IN. Splitting Ice seems to be an assumption as well, which is fair. They should really just make Splitting Ice baseline.

I've talked allready with komma about the current AoE-rotation. A correct APL for every talent on AoE would require a lot of duplicated lines, which makes it even harder to read. I would recommend a working apl just for frostbomb + tv / CmS. (PC is ... a little bit toooo anyoing)

I'll mention what I said in chat. To create a true AOE APL suitable for all target numbers, we would probably need some form of "dynamic priority". This means a list of actions, where each one has a computed value based on number of targets. The action with the highest value in then chosen for execution. This would avoid weird situations such as "Blizzard at 5 targets > CoC at 5 targets > CoC at 4 targets > Blizzard at 4 targets". Unfortunately, with the current APL language, this is not possible.

Currently robomage does not cast Frostbomb with less than 2 charges of FoF. He even accepts to cast IL without Frostbomb, if procc runs out. This is actually ~200 dps better then casting it with 1 Charge. Results vary slightly depending on gear, e.g. with lowgear on 3 targets there may be some cases where it is equal.

There is a safeguard for robomage to cast Frostbomb if it detects FoF with 2 GCDs remaining, but Frost Bomb not being available on the target. This prevents casting FoF-IL without FBomb. It is contained in the following line: "buff.fingers_of_frost.remains<gcd.max*2" means "FoF has less than 2 GCDs remaining".

APL changes: Basically, just extend IV before chaining Blizzard. It actually includes the waterjet changes listed above, to check it vs 4 targets.

Results:
4 Targets: dps loss. No go.
5 Targets: Equal ~ 86 dps increase
6 Targets: Equal ~40 dps behind

This looks significant. It is curious that infinite IV doesn't consistently increase throughput even with Frost Bomb on AOE. I wish there were some way to identify "damage done to main target", because that is much more interesting to me than overall throughput. After all, it's the damage done to Cho'gall that matters, not to the night-twisted faithfuls. Due to lacking detailed target damage breakdown, it seems like the sim isn't offering us these details.

WJ on Pull with PC + FBomb

With FB the playstile is... horrifying. Anyway it is definitly a dps increase.

I assume by "horrifying" you mean that it is excessively busy and complicated due to endlessly playing catch-up with procs.

For both IN + FBomb I needed to rewrite the APL.
The basic idea of that APL is to:
- Frostbolt + WJ usage
- dump IL's to 1 charge before PC.
- if there is some spare time: dump FFB's to 1 charge before PC, trying to avoid FFB munching from frostbolt casts.
basically: keep 1 charge of FoF and 1 or 2 charges of FFB for PC.

I'm not sure if the current APL dumps FoF to 1 charge before PC. That sounds like something that needs to be done if it isn't. The proper implementation would be done in actions.crystal_sequence.

I've already played around with banking BF-FFBs for Prismatic Crystal. In general, they never seemed to yield noticeable gains, which was a surprise to me. I'm not sure why, but I guess it might be due to PCs being overly busy with IN/FB/FO/FoF-IL usage already. Maybe it needs to be gated by talent selection?

P.S. The waterjet codehack actually increases dps for every profile. Results above were created without these line inside comparing profiles. It does not change much, lowers the gap between Fbomb exampels. That little line actually just tells robomage: Inside waterjet, frostbolts which hit the boss, will give a FoF charge, which I can dump instantly, with spellque system. No need to think.

This shouldn't be necessary. The original code already includes this: This is already the "dump IL if Frostbolt is going to generate FoF overflow" line. Maybe the condition isn't correct with react=2? I'm not sure. But whatever "dump IL after WJ to avoid proc capping" behavior should be included here through condition changes.

I appreciate your work!

The results are quite interesting. Banking was allways something which i had in mind could result in a DPS Increase in Simcraft. I can understand why banking for Mark of the Frostwolf doesnt work out. I wonder what the result is to only bank a maximum of 1 FoF.

It is not that easy to implement banking though. The intuitive playstyle (of me) would be to only Bank right before major Spells are about the get ready. It is also quite intuitiv when i would bank 1 FoF (most of the time) or even 2 FoF Procs right before e.g. PC.

As soon as my semester ends, i will take a look at how the Coding of the APL works and will do my own testings. It is about time to contribute some testing myself.

I appreciate your work!

The results are quite interesting. Banking was allways something which i had in mind could result in a DPS Increase in Simcraft. I can understand why banking for Mark of the Frostwolf doesnt work out. I wonder what the result is to only bank a maximum of 1 FoF.

It is not that easy to implement banking though. The intuitive playstyle (of me) would be to only Bank right before major Spells are about the get ready. It is also quite intuitiv when i would bank 1 FoF (most of the time) or even 2 FoF Procs right before e.g. PC.

As soon as my semester ends, i will take a look at how the Coding of the APL works and will do my own testings. It is about time to contribute some testing myself.

I went from FoF/BF banking -> only BF because the primary loss of DPS in BF/FoF banking is from FoF proc munching. Unfortunately, you just cannot bank 1xFoF without losing munching. This is primarily due to the nature of how quickly the frost spec plays, how all of it's spells can be cast in a non-proc state, and in just the nature of reaction times.

Arcane avoids this problem by having AM not be castable unless you have the buff. So you can kind of "roll" your finger from AB -> AM, and if the AB you are casting proc'd an AM, an AM will immedietly fire. If you do this sort of predictive play with BF-FFB or FoF-IL, you end up queue a cast-time FFB, or non-FoF-IL.

What you propose (only banking when you know certain periods of power are coming up) is actually a much smarter way to do it (and may well avoid the problem of munching because you are banking all the time). The problem is that this is also much more complicated to implement!

If you're interested in the APL, you're welcome to hop into the IRC and have a chat with Komma / myself, and read through the documentation on action lists here: https://code.google.com/p/simulationcra ... ctionLists" target="_blank . I think we're pretty open to new people signing on and helping

( also, I haven't ignored your posts wand. They just deserve more attention / time than I have to give right now <3 )

For both IN + FBomb I needed to rewrite the APL.
The basic idea of that APL is to:
- Frostbolt + WJ usage
- dump IL's to 1 charge before PC.
- if there is some spare time: dump FFB's to 1 charge before PC, trying to avoid FFB munching from frostbolt casts.
basically: keep 1 charge of FoF and 1 or 2 charges of FFB for PC.

I'm not sure if the current APL dumps FoF to 1 charge before PC. That sounds like something that needs to be done if it isn't. The proper implementation would be done in actions.crystal_sequence.

I've already played around with banking BF-FFBs for Prismatic Crystal. In general, they never seemed to yield noticeable gains, which was a surprise to me. I'm not sure why, but I guess it might be due to PCs being overly busy with IN/FB/FO/FoF-IL usage already. Maybe it needs to be gated by talent selection?

Well, including codehack inside the comparing profile for Frostbomb shows a 50 dps increase. Which is.... minor. Double checking WJ voodoo may yield better results.

P.S. The waterjet codehack actually increases dps for every profile. Results above were created without these line inside comparing profiles. It does not change much, lowers the gap between Fbomb exampels. That little line actually just tells robomage: Inside waterjet, frostbolts which hit the boss, will give a FoF charge, which I can dump instantly, with spellque system. No need to think.

This shouldn't be necessary. The original code already includes this:

Code: Select all

actions.water_jet+=/ice_lance,if=buff.fingers_of_frost.react=2&action.frostbolt.in_flight

This is already the "dump IL if Frostbolt is going to generate FoF overflow" line. Maybe the condition isn't correct with react=2? I'm not sure. But whatever "dump IL after WJ to avoid proc capping" behavior should be included here through condition changes.

talked with you about that part.

Mid raid atm, but legendary ring proc banking. (10% int is pretty huge to bank for)

Mid raid atm, but legendary ring proc banking. (10% int is pretty huge to bank for)

Becomes 15% with the next upgrade. Seems like it's worth looking at, but I'm not sure how feasible it is. At some point, we'll have to start creating an AI to generate APLs...

Edited my post about WJ on Pull. May be interesting for some of you guys, who like to do more work on this topic. I'm done with it. Recommend reading the fullpost.

Edit:
Rerun sims with Waterjet codehack inside comparing profiles.Furthermore for Frostbomb I've changed precast to Frostbomb aiming to get this rotation done:
1. Frostbomb precast
2. Frostbolt into WJ
3. Dump to 1 IL charge
4. PC / Orb burst as usual.

I actually failed a bit with robomages rotation, that is the reason, why I'll not share the profiles. On a ~450 second fight time, the dps gains / loses from "perfect" openers are below 100 dps. Based on roughly ~32k dps averages. Waste of time to do more work here.

Anyway, results show a slight dps loss for IN rotations and a okayisch dps gain for Frostbomb.Again, results are from sims with 80 sec fighttime, 0 variance, 250k iterations. Keep in mind, that by using PC after WJ you're kind of delaying PC. Depending on fighttime this result in a slight dps loss, even for Frostbomb. (E.g. 400 sec fighttime, 20% variance = dps loss, 450 sec fighttime, 20% variance = dps gain.)

Results: http://fs2.directupload.net/images/150109/uf8kicop.png" target="_blank

Might as well post this here instead of creating a different thread. I tried simming casting Frost Bomb with 1 or 2 FoF procs, by changing the line:
and replacing buff.fingers_of_frost.react=2 with react=1. Does that make sense? Did i go wrong anywhere? The end result was that the difference between the two is negligible, while every other source i've found says you should FB only with 2 FoF procs.

Results:

With just one proc:


And with two:


I ran 25k iterations, 300s single target fight with no fight duration variance, using MI/TV.

When the Fbomb logic was originally settled on, we had not yet wrangled in the monster that is WJ. I think it's possible that now robomage is using WJ smarter, you can cast Fbomb more and be ok.

It doesn't look out of sorts immedietly. How does the ability use breakdown change, if it does?

Ability breakdown is basically identical:

one proc is this



and two procs is this



The only thing that seems to change is variance; casting Bomb with one proc seems to lower it a little bit, which makes sense since i guess it makes munching less likely from something like get one FoF, queue Frostbolt, get FoF, another Frostbolt already queued, cast Frost Bomb, meanwhile the second Frostbolt procs a third FoF before you have a chance to get rid of the first two.

For reference, this is my armory http://eu.battle.net/wow/en/character/c ... ema/simple" target="_blank

Now that progression is over, I can actually go back to fixing stuff.

P.S. The waterjet codehack actually increases dps for every profile. Results above were created without these line inside comparing profiles. It does not change much, lowers the gap between Fbomb exampels. That little line actually just tells robomage: Inside waterjet, frostbolts which hit the boss, will give a FoF charge, which I can dump instantly, with spellque system. No need to think for robot, no need for reaction times.

This is now included as part of the sim.

The only thing that seems to change is variance; casting Bomb with one proc seems to lower it a little bit, which makes sense since i guess it makes munching less likely from something like get one FoF, queue Frostbolt, get FoF, another Frostbolt already queued, cast Frost Bomb, meanwhile the second Frostbolt procs a third FoF before you have a chance to get rid of the first two.

Right now in the regular rotation, Frost Bomb is used when any of the following conditions are met:
1. Before water jet.
2. For Frozen Orb/PC.
3. If the player has 2 FoF.
4. If the player has 1 FoF but is using TV.
5. If FoF will expire soon.

Most of these have been verified pretty thoroughly to be gains. If you guys feel like any scenario is missing, feel free to suggest them.

The only thing i can think of is that holding on to 2 FoF procs before casting Bomb might lead to some munching. As i said, i tried changing that last line from buff.fingers_of_frost.react=2 to =1 and nothing seemed to change dps wise, but frankly i have no idea why that is happening.

My idea originally was simply to check how much dps i had lost after one night when i kept playing badly and casted a few Frost Bombs with only one FoF charge

I found what appears to be a couple of significant bugs with the code that computes icicle damage. I was looking at some sim results and wondering why icicle damage did not match the expected [ie. (FB+FFB) * mastery%]. In particular, icicle damage was noticably higher than expected, which should be impossible with no dynamic mastery buffs.

The 2 issues I found were:

1) Versatility is being applied to icicles. Icicle damage is computed based on final FB/FFB damage, but when calculate_direct_amount() is called for icicles, it applies the versatility multiplier again. One easy fix is:
2) This one is trickier and harder to explain. If you have 5 icicles and a Frostbolt multistrikes, the sim will schedule events to immediately fire off the oldest 2-3 icicles. The problem is, there's only 1 instance of the icicle damage object in the events. That means that instead of each launched icicle doing their respective damage, all 2-3 of them will do the same damage. The relevent code is here:
As a proof of concept, I modifed the code to allow separate icicle damage objects in the events. With both fixes in, the icicles were being fired with the proper damage, and overall icicle damage matched the expected formula.

Implications: Frost DPS currently inflated by ~1-2%, versatility likely overvalued.

Good catch. There are a few other known bugs, which also relate to Prismatic Crystal. I'll look into these two.