Glacial Spike on Quick Sim - Raidbots

Classes Mage
1

Edit: TLDR: Quick Sim is only casting GS after Flurry proc, this leads to (in some cases) strings of 7+ icebolts cast with 5 icicles up. This seems to be causing a 25-35% reduction in GS build sims. Quick Sim shows Thermal Void (on my toon) doing 15-20% more than GS build, but when I test, I get GS build ~ 10-15% higher than TV build.

I checked out a bunch of spec issues over the last few days, primarily driven by posts where I saw “there’s zero reason to take anything other than Thermal Void”. I don’t have the icebolt legendary, but everything else seemed sound. I’m also being pressured to look at fire, but that’s going to require a covenant change and major changes to playstyle (my brain has problems with 20 spell casts in 10 seconds).

I simmed the icy veins recommended frost build and my current build on raidbots (using quick sim). I also simmed the icy veins recommended fire build (again, with the wrong leggo). This is what I got:

GS build: 4320 dps
TV build: 4840 dps
Fire build: 4750 dps

Fine… but then I tried GS vs. TV on training dummies:
Keep in mind: no lust, no flask, no food, no oil, no combat pot…
Boss+add training dummies (just 1 add):
GS: 5200
TV: 4700
I figured - ok, maybe GS build cleaves better…

Boss solo:
GS: 3800
TV: 3200

Now, I may cast some things wrong, but this is night and day on the relative side. Instead of TV being 15-20% higher dps than GS build, it’s 10-15% lower. The thermal void rotation is absolutely brain dead, so I don’t see how that’s the problem. It’s literally only 3 spells (frostbolt, flurry, icelance) + occasional buffs and 1 minute/+ c/ds - and all of the stuff outside those 3 spells are the same for GS.

Between buffs and an overall problem with how I cast (i.e. when to queue spells), I am not worried about the baselines being off by ~33% (i.e. I get 3200 when Quick Sim says I should get 4800).

I looked a little closer at what Quick Sim gives as the spell rotation for my GS build. Every, literally every GS is cast after Flurry, and before Ice Lance. Now that’s not particularly terrible, but there’s one point in an example parse from them where they cast, with 5 icicles already up, 7 icebolts in a row. Presumably this is to proc Flurry… but this is insane. For one thing, I tried doing flurry - GS - ice lance and I can’t even tell if the ice lance is getting the frozen target bonus with GS stuck in the middle (doesn’t look like it). And GS doesn’t seem to hit harder this way. For another, apparently when you cast icebolt with 5 icicles up, it tosses an icicle out at the target (maybe, I didn’t really pay attention)… fine but that icicle is garbage (couple 100 damage). My base GS is 10500+/-1200, with almost identical 2x crit. If I had cast GS first (without flurry), then the 7 frostbolts, it would cost me 2.7s to add an average of 15000 damage, or 5500dps, compared to the lame dps of frostbolt [1500] + icicle [800] doing about 2700/1.8s or 1500 dps.

I’m getting pressured to switch to fire, but as far as I can tell, the frost spec that people are comparing to may be getting handled in a peculiar way at the theorycrafting angle.

The only thing I can see as a benefit on fire is that it peaks in the 10k+ dps range for 10s out of every combustion cycle. But then it’s trash (literally below 65% of average for 24 out of 60 seconds). This has potential benefits. But so far I’m not even convinced that my potential dps throughput on fire will be remotely as good as my frost dps, when built properly.

Is this type of bizarre-o thing with Quick Sim a problem? Or is there something strange I’m not getting right about TV build? I ran a few 1M dps samples against the raid tank boss with no group or consumable buffs. At 209 ilvl, with an aoe leggo, and I could only hit about 3200-3300 dps with TV build. Do others get similar?

2

So one thing you aren’t including here is your conduits/fight time. The main reason TV is used over glacial spike is it has huge synergy with icy propulsion, the best frost conduit. Glacial spike has 0 synergy with any conduits and makes icy propulsion worse. And the fight time also changes how many icy veins you get which is the main reason to run TV over GS.

3

I see, yeah, GS doesn’t work with any conduits directly. I’m running Unrelenting Cold and Ice Bite. I’ll swap probably Unrelenting Cold (since it buffs orb the same way regardless of build) and re-run 1M sim damage against the raid boss dummy.

For reference, these trial runs are running about 8 minutes. Quick Sim seems to run 6 minutes. My approach typically gives Icy Veins one more try.

I had forgotten about Icy Propulsion. The Quick Sim difference is including my conduits, so it’s not exactly addressing the issue that Quick Sim seems to be totally out of whack for GS builds. But I’ll test with Icy Propulsion (and leave it there for my GS build) and do two quick 1M runs. Gimme about 15-20 mins and I’ll edit/report.

Edit: Report…
I switched to Ice Bite/Icy Propulsion on conduits (using Theotar, with 2 defense conduits because I don’t get heals sometimes D-: )
GS build: 3850 dps (2 Icy Veins, 3rd about to hit right as I hit 1M)
TV build: 3350 dps (3 Icy Veins…)

I was wrong about duration. The duration on my GS test was about 260s, duration on TV test was about 300s.

For what it’s worth, my crit is crap. This is, of course, because when I hit the vault (as frost spec), I get a 223 ring (this week, but this happens every week) with +135 haste and +75 vers.

So it’s still off, by a lot… TV is way worse than GS build. But I bet if I simmed all this on Quick Sim, it would have TV up by 15% of GS. I’ll try to run those, but they take longer due to queue time.

4

Isn’t glacial spike worse than not picking anything on the last row? Or did they fix that?

5

The advantage of simulations is that they can re-run the same combat scenario many times over and average out the effects of RNG, as well as consistently execute a rotation in order ignore player error. These are the main problems of running a few dummy tests. Did you execute each test in the same way? Did you have the same kind of procs? Did the procs occur at the same time? Did you crit in a similar manner? These factors are averaged out across the iterations that simulationcraft runs. These factors in combination with not matching sim vs dummy combat environments most likely accounts for the discrepancies in what you see in the sim vs dummy test.

GS is not cast as often as possible in the sim because it’s a DPS loss. You can make a small modification to the mage rotation in simualtioncraft in order to cast GS more often and you’ll see that it’s at best DPS neutral and most of the time a DPS loss. The unfortunate fact of the matter is that GS is a terrible DPS talent so far this expansion.

1 Like
6

I assume by “TV build” you’re running LW, SI and CR? If you sim and only change 1 talent, it’s not going to show the full benefit of it.
I also hope you’re running the Slick Ice legendary when you sim TV, because higher IV uptime means you gain more benefit from that.

From all accounts in both sims and from the top Mages in the world, GS is a bad talent to take. As the poster above says, it’s neutral dps at best and a loss at worst. It has no synergy with any legendary or conduits.

As for the sims, they only cast GS after Flurry because it’s a damage loss otherwise. GS has to crit to make it worth the cast time, and the only way to ensure that is Flurry and a decent crit rating.

7

Edit: It is annoying I can’t respond to my own response when someone makes a point that I have not addressed in between…

My reply to Tuckletuts is here: Followed by additional info on what I’ve found by more testing and replies to others above…

I’m using the icy-veins recommended build of LW, SI, and CR. And I’ve changed conduits to have Icy Propulsion.

My crit sucks (23%), but I average about 50/50 on Glacial Spike crits without trying to get them on flurry procs. The increased damage from regular to crit is about 10k, so if you only cast on flurry proc, you eat 1 4k increased ice lance (already reducing the benefit to 6k), and if it would appear to be around 2 or 3 additional frostbolts where if you had just cast the GS, you would be back for more within 2-3 seconds.

I understand it will “get better” if my build is right. I’m not worried about that - I’m working on it. What I don’t understand is how to even test what I’m doing if the sims that say I should be 10% over GS build are giving me 10% or more below the GS build when I run TV.

======== Original Reply =========

I understand that the simulations remove a lot of the randomness. I also know that, based on what Quick Sim gives as 95% range on simulations, it means 2-sigma on each side of the mean. I also know that 2/3rds of the time (roughly), my results should fall within 1-sigma.

One thing I did figure out, looking at Quick Sim’s casting approach is that I totally did not understand that you should icelance 2 times after flurry. That’s bumped my dps up about 2-3%, apparently on both my GS and TV build.

But Quick Sim says I should get:
TV Build: 4975+/-204
GS Build: 4414+/-135

Rerunning again, on the raid tanking dummy (which has no debuffs or anything else strange), but not using raid buffs, time warp, or consumables, I just now got (for 1.5M damage runs, which take at least 6 minutes)):
TV Build: 3450
GS Build: 3800

I understand individual runs of something like this can differ. But look, I’m getting 3800 (with no buffs) while Quick Sim says I should get 4400+/-135. To me, this is about right.

But then it turns around and says I should get 4975+/-200 on the TV build, while I’ve gotten everything (with slight mods to build) from 3200 to 3450 now. I’m inclined to feel the 3450 is a little high because it was tracking much lower then recovered during an icy veins. But if we assume the buffs are good for about 15% of the overall damage (i.e. I’m doing about 87% of possible raid damage - so the 4400 literally comes down to 3825), then the 4975 comes down to 4328. But I’m only getting about 80% of this, at 3450.

What’s interesting, still, is that Icy Veins only casts Glacial Spike after Flurry - maybe this is because of what you guys are indicating regarding GS being terrible. But they’re literally throwing away 3500+ dps by not casting it (i.e. every 2 frost bolts for a total of 3000 damage could have been a GS for an average of about 15000 [for my build], so losing 12000 extra damage but only potentially saving 2 icicles.

My main issue is, however, what could I possibly be doing wrong on the TV build??? Is it any more complicated than:
If fingers of frost, cast icicle
If brain freeze, cast flurry, then icicle, then icicle
Else cast frostbolt
Cast icy veins, frozen orb, mirror images, and trinkets on c/d

What am I missing? Why am I getting 20% (at least) lower dps with TV build than I should be getting?

8

Are you doing your testing at a dummy that’s close enough to others for GS to cleave to via splitting ice? I would imagine a GS build would benefit more from a second target than TV and might be skewing your numbers.

IIRC, there’s a dummy in Ironforge that’s far enough away from the others to not cleave.

9

I try really hard to use solo dummies. The turnip toy sometimes works for this but you can only use it once an hour for 5 minutes. I’ve been using the tanking raid boss in Venthyr hall, which seems to be intended to be a solo target (though there is a 62 tanking boss dummy near it). My usual check is if glacial fragments hits anything.

I think there’s a trick that is not well documented. I think there might be two tricks, actually.

I think one trick is related to getting flurry to have a potential brainfreeze proc. It appears you can do this by casting frostbolt with brainfreeze already up, then take flurry. When you do this, sometimes flurry procs another brainfreeze, even though flurry doesn’t actually participate in brainfreeze. But… if you do this, you MUST cast flurry immediately after casting frostbolt while in brainfreeze, or you still lose it.

The other trick seems to be that you should cancel frostbolt any time that you have 2 fingers of frost, and immediately consume your fingers of frost. Method has this noted on their frost mage guide.

By using the first trick, my TV build is getting really close to the dps of my GS build. Maybe using the 2nd trick will fill in some of the gaps -but it looks like its going to only add 1-2%, so it’s still going to be 8-9% below where it should be.

Edit - incomplete sentence at the end of the Method reference…

Edit 2: one thing I’m noticing in my logs compared to Quick Sim is that my ratio of Flurry to Frostbolt sometimes runs low. As my TV build has improved overall dps, the last couple runs have resulted in 27.5 and 28.6% flurry/frostbolt ratio. However, my last GS test has 34.3% flurry/frostbolt ratio. These ratios are not cause for concern on this few casts (each is 50-100 flurrys, 150-300 frostbolts), but I will watch to see if the trend holds up.

Edit 3: I run Ebonbolt on my GS build, so I have to account for that… the 34.3% flurry/frostbolt on my GS build becomes 25.2% after removing Ebonbolt. I reran my GS build again to try to see what’s going on here, and I got another run with 41.2% which corrected down to 32.1% after removing Ebonbolts. I do believe Ebonbolt, if cast on a frostbolt that brainfreezes, saves the brainfreeze as log as you cast flurry immediately after. I.e. frostbolt->procs brainfreeze and you cast ebonbolt->then immediately flurry-> proc’s brainfreeze again. Also… I think that if brainfreeze procs on the last frostbolt before you GS, you should cancel GS and hit the flurry proc first. I seem to be reliably hitting 3950 on my GS build now. But I have to pay a lot of attention or it slips to 3800.

Edit 4: Also, mirror images only seem to add 3000 damage per cast (right around 3000-3500 dps), so there’s no need to do it, ever (though it reduces incoming damage). Quick Sim never casts Mirror Images. I removed that from my rotation about 6-8 2M+ tests ago.

10

This is part of a few changes the devs worked on a few years back to make frost feel better to execute the shatter combos. They do a good job of explaining it, so I’ll just quote it and provide a link to the page. Keep in mind the quote below is from the Legion expansion, so the double ice lance after flurry was not officially supported back then.

Brain Freeze procs gained while you already have a Brain Freeze active and unused will now be delayed slightly.

  • Developers’ notes: The delay on the second Brain Freeze will give the Mage time to use the first Brain Freeze proc. For example, the Mage might cast Frostbolt (proc Brain Freeze), then Frostbolt (proc Brain Freeze again), then instant Flurry, then Ice Lance, then Instant Flurry again, then Ice Lance.

Scroll down to January 31 for the quote. This particular link just happened to come up first for a search.

11

Thank you! - For anyone else’s reference, the entry is from Jan 31 in that log.

But there’s something more subtle that might be totally messing up my dps here. and it comes from this entry on the same date.

  • The first missile from a Flurry cast should now hit sooner.

    • Developers’ notes: This allows a Frostbolt or Ebonbolt cast immediately before a Brain Freeze Flurry to more consistently benefit from that Flurry’s Winter’s Chill effect, even if the mage isn’t standing at long range. This allows the Frost Mage who has an active Brain Freeze to cast Frostbolt/Ebonbolt, then an Instant Flurry, then Ice Lance while standing 15+ yards from the target. Previously, the Mage had to be about twice as far away.*

I have been intentionally standing within 15 yards of my target dummy because I have Soul Ignitor, and I’m dumb. I just double-checked the tool tip, and although the graphic is very short range, it says the area of effect is 40 yards.

Here’s the thing: Flurry sends out 3 bolts, but it only stacks to 2 on Winter’s Chill. I had assumed it was only intended to increase crit chance for 2 spells. But, clearly they want you to shatter a frostbolt or ebonbolt with the first one (for some reason you can’t shatter a GS, there are a lot of strange discussions of this). Because I’m standing so close, it appears to not shatter my preceeding frostbolt, and then stacks to 2, because they all hit very quickly.

I bet I’m losing 1/3rd of the crit benefit of flurry sometimes (though only affecting icebolt). My tests show 30% crit on frostbolt, but Quick Sim gives 43% crit on frostbolt for my TV build. It could be because I’m not trying to shatter a frostbolt after icelancing from fingers of frost, OR because I can’t shatter it because I’m standing too close.

I’ll try standing further away (20-30 yards), and make sure I put a frostbolt between any icelance and flurry proc. If this increases my frostbolt crit%, it might resolve getting 3 crits off flurry.

… I think maybe I’m starting to home in on my issues here. LOL

12

Less than 15 yrds and timing gets fuzzy for shattering the pre-flurry spells, but something I’ve experienced with shatter in and of itself, is unless you’re within 5% of soft-cap, you’ll always have a 10-20% chance to miss a crit, which ultimately ends up being GS or Frosbolt for some reason.

Which is compounded with Blizzards inability to let Frost spells play off each other. That, and the lack of available stats causes any builds that reliably rely on two particular to bump up in power over-all.

That and Icy Propulsion + TV giving you lust for 30% of the fight duration doesn’t hurt.

13

– Personally, I can’t help wondering if I picked Arcane or Fire loot specification, if my “Frost” loot would have better stat allocation.

…
One difference I’ve found after tons of checking, and a lucky hint by another frost mage in Venthyr hall: Mirrors of Torment gives an additional Brain Freeze proc on average every 30 seconds (in practice is gives 3 in about 15 seconds, then it’s down for 90 second c/d). But Mirrors don’t proc off training dummies.

MOT has an interesting property that it theoretically results in an additional 9 seconds of Icy Veins due to Brain Freeze giving 3 more crits each time. But it’s not clear how to test this, and it’s not clear why this would be more than a few % more improvement to TV over GS, and I’m still sitting at about 6% higher GS build to TV.

I may have to take it to a dungeon to see if TV is overall easier to deal with while producing similar, if not better, damage to GS.

The MOT impact could be rather large… But the enemy has to attack… which training dummies don’t really do (though I could have someone else maybe attack a tank trainer while I nuke it)?