Lhivera - just to clarify. You are using, 3.2 mechanics right? (3/3 Enduring Winter - no replenishment munching, and -0.2 seconds for Empowered Frost Bolt vs. 4% Crit)

You lost me on your builds though. If we're going for a max DPS build (ignoring things like Ice Barrier and Actic Reach) then I only count 51 DPS talent points on the Frost Side.

If you're specing 3/3 BF then you have to put 52 points into the frost side, so you have 1 free point to spend lower in the tree (Arctic Reach, Ice Barrier or Improved Blizzard are good choices) - but that leaves 19 points for Arcane. This will give you 5 points to spend between SotM and Spell Impact. I'd have to check the calculators again with your BiS gear - when I run RAWR it tells me there is 1DPS differece between 2/3 spell impact with 3/3 SotM and 3/3 spell impact with 2/3 SotM.

Something like this

It's probably only like 2-3 more DPS than what you were looking at with 18/0/53a, but it may help.

Correct. The run commands look something like this:

./Mage_T8-wrd.simcraft patch=3.2.0 max_time=900 raid_events=moving,cooldown=30,duration=3 iterations=20000 html=../frostopt/mage32_900m.html

The patch=3.2.0 argument triggers the change to Empowered Frostbolt, and all four specs include 3/3 Enduring Winter.

Before I go posting the complete results, I just want to make sure that the gear set in the Fire file is comparable in quality to the one in the Frost files.

The one at the top, "2.2.6 - Ulduar - Normal and Heroic Hard Modes" that has a solution dps of 7780.29. I doubt that the one already in simcraft is optimal, 7200-ish dps seems rather low for a perfect Ulduar hardmode set.

OK, new results using Sancus's gear list (I did not use the Jewelcrafting gems, since the Fire set doesn't).

For those who have lost track:

- 18/0/53a has 3/3 Spell Impact, 1/3 SotM, 3/3 Brain freeze, and uses BFF's and Ice Lances on ghost charges.

- 18/0/53b has 0/3 Spell Impact, 3/3 SotM, 3/3 Brain Freeze, and uses BFF's only when moving.

- 18/0/53c has 0/3 Spell Impact, 3/3 SotM, 3/3 Brain Freeze, and uses BFF's but not Ice Lances on ghost charges. This is the "what if I put a point into Magic Absorption and ignore Ice Lance on ghost charges" spec.

- 19/0/52 has 0/3 Spell Impact, 3/3 SotM, 0/3 Brain Freeze, and thus does not use BFF's at all. This is the "what if I spend those points on more range or some kind of Frosty flavor" spec (although in my example, I use one of the three points to max Magic Absorption).

5-minute stationary fight

Spec	DPS	% below Fire	% below top Frost
20/51/0	7337	n/a	n/a
18/0/53a	6437	13.98%	n/a
19/0/52	6426	14.18%	0.17%
18/0/53b	6407	14.52%	0.47%
18/0/53c	6372	15.14%	1.02%

5-minute fight, 3 sec movement per 30 secs

Spec	DPS	% below Fire	% below top Frost
20/51/0	6829	n/a	n/a
18/0/53b	5755	18.66%	n/a
18/0/53a	5742	18.93%	0.23%
19/0/52	5684	20.14%	1.25%
18/0/53c	5639	21.10%	2.06%

7-minute stationary fight

Spec	DPS	% below Fire	% below top Frost
20/51/0	7284	n/a	n/a
18/0/53a	6320	15.25%	n/a
18/0/53c	6291	15.78%	0.46%
18/0/53b	6247	16.60%	1.17%
19/0/52	6220	17.11%	1.61%

7-minute fight, 3 sec movement per 30 secs

Spec	DPS	% below Fire	% below top Frost
20/51/0	6781	n/a	n/a
18/0/53a	5621	20.64%	n/a
18/0/53b	5610	20.87%	0.20%
18/0/53c	5550	22.18%	1.28%
19/0/52	5492	23.47%	2.35%

10-minute stationary fight

Spec	DPS	% below Fire	% below top Frost
20/51/0	7201	n/a	n/a
18/0/53c	6196	16.22%	n/a
18/0/53b	6175	16.62%	0.34%
18/0/53a	6152	17.05%	0.72%
19/0/52	6079	18.46%	1.92%

10-minute fight, 3 sec movement per 30 secs

Spec	DPS	% below Fire	% below top Frost
20/51/0	6707	n/a	n/a
18/0/53b	5549	20.87%	n/a
18/0/53a	5462	22.79%	1.59%
18/0/53c	5458	22.88%	1.67%
19/0/52	5337	25.67%	3.97%

15-minute stationary fight

Spec	DPS	% below Fire	% below top Frost
20/51/0	7130	n/a	n/a
18/0/53c	6087	17.13%	n/a
18/0/53b	6078	17.31%	0.15%
18/0/53a	6065	17.56%	0.36%
19/0/52	5987	19.09%	1.67%

15-minute fight, 3 sec movement per 30 secs

Spec	DPS	% below Fire	% below top Frost
20/51/0	6657	n/a	n/a
18/0/53b	5452	22.10%	n/a
18/0/53a	5382	23.69%	1.30%
18/0/53c	5371	23.94%	1.51%
19/0/52	5263	26.49%	3.59%

Some things of note:

- Some accuracy checking is still needed. I think the sim zeroing in on Frost, but it'd be nice to figure out why 19/0/52 and 18/0/53b don't perform identically on stationary fights, and the Fire numbers at least are wildly different from Rawr's. I have no idea how similar the Frost numbers are to Rawr's. Everything below here is somewhat contingent on the numbers above being at least reasonably accurate.

- The total spread of all four Frost specs and DPS methods ranges from 1.02% on a 5-minute fight to 1.67% on a 15-minute fight. Using Brain Freeze and Ghost Charges strike me as strictly optional; human reaction time will make a larger difference.

- The total spread on movement fights ranges from 2.06% to 3.59%, again based on fight length. Is that worth three talent points that will likely come out of Arctic Reach, Ice Barrier, and/or Ice Floes? If movement is common, an argument could probably be made, although losing range is counterproductive to improving performance on movement fights.

- Frost's efficiency is not particularly helpful, it seems, since Frost has to evocate more frequently than Fire, and the DPS gap increases by about 3.6% as we move from 5-minute to 15-minute fights.

- Frost's short cast times are not particularly helpful on mobility fights, since the DPS gap increases by about 5-6% when 10% movement time is added.

Those last two points are, I think, particularly worthy of note because one of the most common arguments against making Frost DPS comparable to Fire (even when ignoring PvP) is that it's loaded with advantages that would make it the only reasonable choice. But now we can see that even if Frost were to put out identical DPS to Fire in a 5-minute Patchwerk fight, it would drop behind as fights get longer or involve more movement.

Frost would be the superior choice in:

- Encounters in which using Ice Barrier is useful enough to reduce DPS by spending GCD's on it.

- Encounters in which a 35% AOE snare is useful.

- Threat-sensitive encounters (having a percentage of damage come from the pet means around 17% threat reduction vs Fire's 10%).

- Encounters in which active mana regeneration effects don't work.

- Encounters in which having an Ice Block every 4 minutes is better than having one every 5 minutes.[/ul]

Fire would be the superior choice in:

- Any encounter longer than five minutes.

- Any encounter that contains movement.

- Any encounter that kills pets.

- Any encounter in which DPS near the end of the fight is more important than DPS in the rest of the fight.

- Any encounter in which passive pushback protection is useful.

- Any encounter in which having a ranged, single-target snare is useful (Fire has a 1.5-sec Frostbolt; Frost has only a 3-second FFB due to Glyph of Frostbolt).

Frost does indeed have Replenishment. But if it's much more than 5% lower than Fire in a short Patchwerk-style encounter, then in the majority of encounters, it will remain too low for most people to consider it a viable choice, I think.


Links to output; the links in the files to the configs will work:

5 minute stationary
5 minute movement
7 minute stationary
7 minute movement
10 minute stationary
10 minute movement
15 minute stationary
15 minute movement

Some questions I had after reviewing your sim output:

• Evocation appears to be handled very strangely by this sim. I'm seeing mana return values of 11.2 and 355.4 on the 5-minute movement fight (with no overflow). Why is the sim trying to evocate for minute amounts of mana that do not even prevent an OoM state (since the frost models don't appear anywhere close to running out of mana with that fight duration)?
• There seems to be something wrong with the crit calculation for Pyroblast in fire as it is over 40% behind all of the other spells in crits.
• Fireball does not seem to be ticking its DoT in the frost models.
• While Frost seems to have an intelligent movement rotation, Fireball is limited to Fireblast. Is there a reason for not casting Living Bomb/Hot Streak Pyroblast during movement? Even Ice Lance would be better than dead air.
• The "moving=1" conditionals don't appear to be present in your configs. Aren't they required?
• Why are you using the Glyph of Ice Lance on the 19.0.52 spec if it is only supposed to be casting Ice Lance when moving? Is it still more dps than Water Elemental?
• Ghost Charge Fireball crit rates are way below the expected value. If you have Shatter (50%), Focus Magic (3%), Scorch (5%), Master Poisoner (3%), Moonkin Aura (5%), you have a 66% base chance to crit before adding in gear/Molten Armor/Int. The crit rate should be over 80%. Instead, depending on the amount of points in Student of the mind, its somewhere in the 55-60% range. That's a pretty significant difference.
• Looking at the fire mana usage graph, it looks like it is assumed that mana gems have unlimited charges. Obviously this is not the case and the logic for reconjuring gems is going to be needed (and I expect it's going to cause some serious dps shifts in the longer fights as evocation is going to be needed a lot more).

I don't see a pyroblast dot either, so maybe the crit % includes the non-critting DOT component?

Indeed, if brain freeze fireballs are used only on ghost charges, the crit rate should be equal to that of ice lance (about 95% using the original gear). If brain freezes are used when available + on ghost charges, then the crit rate will be approximately the same as for frostbolt (this is just an interesting result of FoF ghost charge probability vs. brain freeze probability after the two FoF charges have been used on frostbolts).

The fireball rank table is missing DOT components. Comparing FFB code with fireball code, a few other lines are missing as well that would deal with the DOT component.

Fireball top rank (0 tick value):
{ 78, 16, 888, 1132, 0, 0.19 },

FFB entry (30 per tick):
{ 80, 2, 722, 838, 30, 0.14 },

Pyroblast includes the DOT component (75 per tick+ 20% of spell power over the full duration added elsewhere in the code):
{ 77, 12, 1190, 1510, 75, 0.22 },

The missing DOT relatively little to the damage of fireballs because it doesn't scale (unlike the pyroblast DOT). What's needed is the DOT implementation and then a check for the fireball glyph to remove the DOT when the glyph is present.

The numbers would fit pretty well if pyroblast ticks were counted as direct damage executes, but then the printed count (29.7) would also include them, so the pyroblast crit rate really must be very low for some reason...

The average pyro crit was 12180 and the average hit was 6707, which allows us to calculate the DPE (without DOT): 21.8% * 12180 + 78.2% * 6707 = 7900. The damage per execute is however 11124. Even with the DOT, it's too much of a difference. Something is fishy there.

The sample output on the SimulationCraft wiki shows a similarly low crit chance on pyroblast... Looking at real logs on WMO, the crit chance is of course much higher, as expected.

I was going to quote for context, but the forums seem to be having difficulty, so I'm going to try a straight post.

I'm going to leave most of this for dedmonwakeen to answer. Bear in mind that he's received very little peer review on his Mage model until the past few days, so we probably shouldn't be surprised to find some problems.

I'm not sure about the moving=1 conditionals. The default files simply include stuff that's only supposed to be cast while moving at the bottom of the priority list, without the conditional -- so my assumption was that the sim just looked, during movement phases, down the list until it found an instant-cast spell that was available to cast. If this is incorrect, I can change those.

As for the Ice Lance glyph -- oops! Bad Lhivera, no biscuit. I shall fix and re-run.

ETA: I have uploaded the new output files. I will update the charts in my earlier post shortly (though I anticipate having to update them again in the future as the sim is refined).

Hopefully this makes it through....

(1) Evocation: Small gains imply that there were few iterations (out of the thousands) that required an Evocation. The fight lengths per iteration can vary significantly depending upon good/bad "luck". The fight-length request is not a hard limit. The sim does not end until the target dies. As the iterations progress, the sim gets a better and better idea of the amount of target health required to get the requested fight length. However, the length of each iteration can still change depending upon procs, crits, etc. However, the total amount gained through evocation is divided by the total number of iterations even if 99% of the iterations did not require it. Obviously, this reporting mechanism should be improved.

(2) The Pyroblast cris were occurring and the final dps value included them. The problem was in the reporting: There was a bug in the manner in which DoTs were being overwritten that was interfering with the reporting.

(3) I wimped out and never implemented the Fireball DoT. I was under the (mistaken?) impression it was in the noise region.

(4) Don't be distracted by the "actions while moving" comment. The entire priority list is examined when players are in a moving state. Any action that is instant (and ready) will be used. This means at Living Bomb will get refreshed. Hot Streaks will get used, etc. However, it does NOT mean clever things like refreshing Living Bomb early while moving will happen. That being said, "dead air" is no good, and we should populate the end of the list with whatever we can.

(5) The "moving=1" conditional is not required. It is used to put an action in your priority list that you want to ONLY happen while moving. The Warlock moving "filler" is crazy-complicated because they come to the end of their priority list quite frequently and need to distinguish between an OOM situation and a movement situation.

(6) Ghost Charge Ice Lances have near 100% crit rates, so if BFF crits are not the same we need to examine the action list to make sure that they are only triggering when "frozen=1" and that the code respects that.

(7) The mana gems are (incorrectly) coded with infinite charges. What is the exact behavior you would like to see?

I appreciate the detailed review. I haven't really pushed for this in the past because I have such a high regard for Rawr.Mage and did not want to distract from that effort. I assure you that you have my complete attention: Any problems you highlight will be resolved immediately.

dedmonwakeen:

3) It would be considered noise for most Mages, because they're using the Fireball glyph which eliminates the DOT. But the DOT is more significant than most people give it credit for, particularly for Frost because it's likely to get most or all of its ticks. With +13% spell damage on the target, the DOT can account for an extra ~131 damage each time you lob a BFF. It's probably worth accounting for.

4) I'll add nonfrozen Ice Lances to the bottom of the Fire action list. Otherwise I think it's covered, since the whole list is checked for available instants.

6) Did you add frozen= support to fireball? If you told me that, I missed it. I'll make sure it's included where it matters.

7) Three charges. I guess the real question is when it's best to use the first charge -- earlier to kick off the cooldown faster, or later so you don't waste some of the charge's mana? And when (if ever) you want to take time to conjure a new one. I don't know the current best practice on this.

ETA: I'm going to wait on updating the tables above, but I'll keep the actual output files current.

(1) Added support for DoT component of Fireball and Frostfire Bolt. In both cases, it will not wait for the DoT component to complete.

(2) Added ghost_charge= option to several mage spellls to ensure the action priority list is doing exactly what the user wants

(3) Limit mana gems to 3 charges. They will be used when the drop from max mana equals the "trigger" value as specified on the action. If not specified, it defaults to 3500.

(4) Allow brain_freeze to affect frostfire_bolt for theorycraft experiments

EDIT: This is all in revision r2865

At a quick glance, on the longer fights, we are seeing a hard OoM condition. The sim is then just having the mage stand around doing nothing. So, we are going to need 2 things (1) code for wanding with JoW procs and (2) code to perform the following: IF ((mana_gem_charges==0) && (current_mana>mana_gem_cost) && (mana_gem_cooldown==0)) THEN recast_mana_gem.

Also, something appears wrong with the optimum set for frost. It is showing 12.85% hit which is almost 2% over the cap. I'll check Rawr later, but I don't think that would be the optimal set.

Something is also odd with the mana gem usage. They are being used prior to a mana debt of 3500 causing large amounts of overflow. Since there's no T7 bonus, they should be held until and used when there is enough actual mana debt large enough not to overflow.

Edit: On the mana gem usage: It looks like it is cracking one right at the start of the fight after the first cast.

Looking at the 15-min movement fight, 19.0.52 is spending significant portions of time OoM and not casting. But the mana regen fron Int/Spirit isn't reflecting this Oo5SR time compared to the specs that aren't spending time not casting. Is there a way to check the model to make sure that the Oo5SR ticks are happening?

I'll add a Conjure Mana gem action with enough conditionals for people to experiment with.

Regarding the usage: Check the trigger= value on the mana_gem action to make sure it is large enough to prevent overflow.

Unfortunately, SimulationCraft is not clever at all when it comes to optimizing Oo5SR regen: It will keep casting as soon as there is sufficient mana instead of pooling mana. I'll look into creating a "wand" action that will be smart enough to pool mana based upon time_remaining() in order to maximize Oo5SR regen.

There's 2 things I was looking at here:
1) I wasn't concerned about it optimizing Oo5SR time. What I was concerned about was the fact that one spec had 2.5% waiting time and another had 5.5% and they had nearly identical int/spirit regen values. That struck me as odd and a possible indicator that no Oo5SR ticks were happening.
2) Even just adding a wand action without any real A.I. to it would probably be enough to fake it for now. Basically, just letting it wand when there's not enough mana to cast anything would reduce the actual standing around time to almost nothing.

I really dunno what the point of 10+ minute simulations are. To be perfectly honest, even 10 minute is ridiculously long. There's literally never been a fight in the history of the game that I can think of that's been more than 10 minutes long without significant phase/movement breaks where you regen mana/evocation cooldown ticks down.

The longest fight we've seen in WotLK was XT-002, which pre-nerf came close to 10 minutes straight of almost chain nuking if your raid dps wasn't great, but there's definitely no precedent for anything longer than that. All the other long fights have a lot of movement time or long phase breaks like Mimiron.

I don't think comparing specs in mana starved, standing-around-oom conditions is useful because if that ever happens the Mage should be swapped out with some other class or Innervated, anyway.

While there aren't that many fights with such a lengthy duration outright, mapping out these types of fights gives us a better insight on some other fights where mana becomes a concern. An example of this would be something like KazRogal. The fight duration wasn't exceptional, but the mana issues were very real.

Looks like a nice list. If you're accepting wish list items, there's a few things I'd like to have a nice conclusion on.

How much DPS is each point in Ice Floes, Spell Impact and SotM with your BiS T8? For a lot of us it comes down to some choices to grab everything we want (Artic Reach, Ice Barrier, Frost Bite, Imp Blizzard, etc). A definitive answer on how much DPS these talents give us may help us budgeting for builds.

As a side note - if the BiS glear really does put us ~1.7% over the hit cap - you can probably safely pull a point out of precision. Unless the 1% mana efficency comes into play for one of your simulations, it shouldn't matter. Of course if you don't have another DPS talent to spend the point on instead, then it doesn't help you anyway.

The gain per talent point is a really good idea, but until we've narrowed the specs down to a very few top choices, it'd be too big a job. I'm doing some preliminary work to try to reduce the list before I post results -- for example, I've verified that, yes, Spell Impact is very very slightly better than SotM if you're executing both Fireball and Ice Lance ghost charges, and that they two have virtually identical value if you're only executing Fireball ghost charges, so that pretty much eliminates the need to produce results for both versions of those two DPS methods.

It may be a legit option to yank a point from Precision and move it into Arctic Reach or something. In fact, heck, why not two points? Three yards range may well be more useful than 0.3% hit. I'll see if doing so has any effect on the numbers.

Edit: Some info on ghost charge usage with the improved code:

6553: BFF > Ice Lance, 100% accuracy
6503: BFF only, 100% accuracy
6502: BFF > Ice Lance, 85% accuracy
6500: BFF only, 85% accuracy
6499: BFF only, 75% accuracy
6489: BFF only, 25% accuracy
6470: BFF > Ice Lance, 75% accuracy

Using Ice Lance gains you about 0.8% DPS if you have absolutely perfect accuracy. At 85% accuracy (which is probably more realistic for most people who know about and can execute ghost charge casts with any reliability), performance is virtually identical to attempting to use BFF on ghost charges and using the Water Elemental glyph. A drop of another 10% accuracy results in a relatively precipitous drop in Ice Lance performance -- at 75% accuracy, using Ice Lance isn't as good as using only BFF at only 25% accuracy.

Using Ice Lance: -25% accuracy = -1.3% DPS
Using BFF only: -75% accuracy = -0.2% DPS

Given the miniscule advantage that even perfect play produces with Ice Lance, and given how much more tolerant of error and latency fluctuation the BFF-only method is, I propose that we drop the Lance method from consideration.

I think for the "attempts to use ghost charges" spec, I'll use the BFF spec with SotM, Water Elemental glyph, and 85% accuracy. We can basically just footnote the facts that if you can achieve 100% accuracy with ghost charges, you can gain about 3 DPS with this spec, or about 53 DPS if you switch to the Ice Lance glyph and use Ice Lances on ghost charges when BFF isn't available.

Do you all think 25% is a reasonable "accidental ghost charge usage" percentage for someone who does a decent job of chain casting but isn't aware of/doesn't deliberately aim for the ghost charges?

Edit: Some results

OK, I've narrowed things down to three specs (see above for rationale on eliminating some of the options):

18/0/53 ghost BFF: This spec will cast BFF whenever available, unless FoF is up, in which case it casts Frostbolts and then attempts to hit the Ghost Charge with the BFF cast, with an 85% accuracy rate. Ghost charge accuracy rate has a miniscule effect on this spec's DPS.

18/0/53 BFF: This spec will cast BFF whenever available, unless FoF is up, in which case it casts Frostbolts and then casts the BFF. It does not aim for ghost charges, but we assume it hits them accidentally 25% of the time.

(Note: Both these specs have 3/3 Student of the Mind and 1/3 Spell Impact. The value of the two is virtually identical, but Student of the Mind is a tiny bit more robust -- single-digit differences -- when Ghost Charge accuracy declines.)

18/0/52 no BFF: This spec spams Frostbolt. It does not have Brain Freeze, which hurts it on mobility fights. It does have max Arctic Reach, which helps on mobility fights (though the sim doesn't tell us how much), and a point in Magic Absorption, and one point to spare.

I'm only going to post the 5- and 7.5-minute encounters for now, to give people a chance to look at the results and the config files to see if they can spot any obvious problems. I'm sure Tigafin, for instance, will be interested in examining the stationary results to see if he can spot a reason why the BFF results are lower than the Frostbolt spam results. (The answer seems to be related to the fact that the no-BFF spec's Frostbolts have a lower average cast time and therefore higher average DPET than the BFF specs' Frostbolts do. This doesn't mean there isn't something wrong, though. I just haven't spotted it yet, and I'm hoping someone else will look while I'm testing DPS values of various low-value talents.)


5-minute encounter, no movement

Spec	DPS	% Behind Fire	% Behind Top Frost
19/52/00	7577	n/a	n/a
57/3/11	7220	4.94%	n/a
18/0/52 no BFF	6521	16.19%	n/a
18/0/53 ghost BFF	6501	16.55%	0.31%
18/0/53 BFF	6487	16.80%	0.52%

5-minute encounter, 3 sec movement per 30 secs

Spec	DPS	% Behind Fire	% Behind Top Frost
19/52/00	7007	n/a	n/a
57/3/11	6662	5.18%	n/a
18/0/53 ghost BFF	5722	22.46%	n/a
18/0/53 BFF	5707	22.78%	0.26%
18/0/52 no BFF	5663	23.73%	1.04%

7.5-minute encounter, no movement

Spec	DPS	% Behind Fire	% Behind Top Frost
19/52/00	7392	n/a	n/a
57/3/11	6789	8.89%	n/a
18/0/52 no BFF	6317	17.02%	n/a
18/0/53 ghost BFF	6305	17.24%	0.19%
18/0/53 BFF	6293	17.46%	0.38%

7.5-minute encounter, 3 sec movement per 30 secs

Spec	DPS	% Behind Fire	% Behind Top Frost
19/52/00	6815	n/a	n/a
57/3/11	6126	11.25%	n/a
18/0/53 ghost BFF	5588	21.96%	n/a
18/0/53 BFF	5576	22.22%	0.22%
18/0/52 no BFF	5466	24.68%	2.23%

All results are available at:

Index of /misc/simcraft

I'm enabling directory listing, rather than linking all the files. The .simcraft files are in there too.


Brain Freeze

This talent is worth less than 0.15% DPS per point in stationary fights of up to 7.5 minutes. In longer fights, when efficiency becomes an issue, it becomes more valuable, adding up to around 3% DPS per talent point. In short mobile fights, it adds a small amount of DPS, around 0.67% per point or less.


SotM

Spec	Total DPS Change	Total % Change	DPS/Point	%/Point
18/0/52 no BFF	31	0.48%	10.33	0.16%
18/0/53 ghost BFF	32	0.49%	10.67	0.16%
18/0/53 BFF	33	0.51%	11	0.17%

Notes:
- Obviously the first point is worth more than the second and third, but the values are so small it hardly seemed worth figuring them separately.
- The value of Spell Impact is virtually identical.


Precision

I am now using the BiS sets included in the .magecraft files updated today. This puts us at 10.9% hit, which unless I'm mistaken is 0.1% under cap. I have therefore not experimented with removing points from Precision.


Ice Floes

A full 3-points in Ice Floes provided no DPS gain in 5 and 7.5 minute encounters, a 3-10 DPS gain in 10 minute encounters (depending on Spec), and a 5-16 DPS gain in 15 minute encounters (again depending on spec). Once again, this hardly seems worth calculating point-by-point; if you take Ice Floes, you're taking it for quality-of-life while doing something other than raiding.

Shatters and others are reporting an apparently Frostbolt coefficient increase on the PTR:

World of Warcraft - English (NA) Forums -> Stealth Frost buff 3.2?

S/he's estimating a 100% coefficient with Empowered, but is using the average rather than the min and max hits. Removal of the snare penalty seems more likely than 100%. I haven't copied over myself, but if anyone here has, some testing may be in order.

Just before work this morning, I did a very quick test on the PTR and on live using the same spell power (1931). It seems to confirm that the "snare tax" is gone. Lowest hit was increased by 148 damage, which would be a 7.66% scaling increase including talents. I used my "for fun" spec, which only has 7 arcane, so the talent scaling factor is 1.1 * 1.06 * 1.05 * 1.05 = 1.285515. This brings it down to an untalented scaling increase of 5.96%. 0.95 * 1.0596 = 1.0066, so even for that quick test, the result was within 0.6% of the 3/3.5 base scaiing factor.

It's probably worth testing with 0 spell power and about 1000 spell power to see if there are any base damage changes (tooltips were identical, but they can be wrong, as was found to be for ice lance tooltips recently) and that it does indeed scale purely with spell power.

I think that's all frostbolt needs really for 3.2. For PvE, I still have better brain freeze (spirit to DOT) and water elemental (haste and crit) scaling on my wishlist.

If I can stay awake long enough to update my PTR client, I'll do some testing with 0/0/0 talents and post the results. I ran my simcraft script with the 0.95 multiplier removed, and it closes the gap between Fire and Frost to 12.36% at 5 minutes and 13.16% at 10 minutes (stationary). It also nearly eliminates the difference between the BFF and non-BFF specs (non-BFF having a 0.6% advantage on stationary fights, BFF/ghost having a 0.7% advantage on mobile fights). I only ran the short fights, though -- BFF will help more on longer fights.

I agree that Frostbolt really doesn't need more scaling. Remaining improvements should be made via the Water Elemental and introduction of an additional spell to the rotation (though the second will be somewhat difficult). The Water Elemental will gain some more DPS with the general pet improvements still to come -- scaling with Crit and Haste rating.

I had not read about the Ice Lance tooltip problem. Can you provide the correct damage range so I can double-check the data in simcraft's files?


ETA:

Min Base Dmg: 799

Current Spell Power Bonus: (3 / 3.5 * 0.95 + 0.1) * 1931 = 1765
Normalized Spell Power Bonus: (3 / 3.5 + 0.1) * 1931 = 1848

Multiplier: 1.06 * 1.05 * 1.05 = 1.16865

Predicted current min hit: (799 + 1765) * 1.16865 = 2996
Predicted normalized min hit: (799 + 1848) * 1.16865 = 3093
Increase: 3093 - 2996 = 97

Your increase is about 50 points more than I'd expect given your spell power. Hurm. It actually nearly fits a coefficient of 3.08 / 3.5 + 0.1, which is quite strange:

Spell Power: (3.08 / 3.5 + 0.1) * 1931 = 1892
Predicted min hit: (799 + 1892) * 1.16865 = 3145
Increase: 3145 - 2996 = 149

If your sample size was small enough that the increase could be off by a few points, it could be as high as 1.0 before Empowered.

Oh, I should note also that 0.98 closely matches what Shatters estimated, though I don't know enough about his spec/glyphs to have checked the math. And if it is indeed 0.88 + 0.1 for Empowered, that would put us within 10.5% of Fire on a 5-minute fight.

Crazy theory here -- if they (a) normalized the coefficient, and (b) replaced the 3% crit on Winter's Chill with +3% spell power on Frostbolt, that would result in a fully talented coefficient of 0.987, while also reducing talented crit and thus improving the value of crit rating for the spec by a little bit. I'll test untalented, with Winter's Chill, and with CttB, just in case. But I won't be able to do it until tomorrow night, so someone else may get to it first.

Here's the ice lance topic from the EU mage forum:

World of Warcraft (en) Forums -> Ice lance bug?

Essentially, when playing with zero spell power (naked), none of the ice lance tooltips matched the actual output. We didn't investigate thoroughly. It could simply be a tooltip display calculation bug.

For a long fight, It would probably make sense to make full use of brain freeze fireballs for maximum efficiency. If SimulationCraft ends up standing idle for mana regen, the numbers are probably not very meaningful anyway.

I still need to figure out why I'm one of the few people who believe that brain freeze is a DPS increase (not a major one and getting weaker with each frostbolt buff, but still a DPS increase). I have the next two weeks off from work, so I might finally have the time and energy to run some simulationcraft tests of my own to see how its parameters differ from my own calculations and simulations.