[2.1 Mechanics] Spirit Generators: Quick Ref.

Monk
Prev 1 24 25 26 Next
02/26/2017 10:37 AMPosted by StoleOwnCar
Either way, this is good. I like having a choice in the gen to use. Good SoJ and elemental TP's are difficult to get, so the more elemental choices that are viable the better it is for players. Looks like both cold and lightning are now quite viable at least.

I agree. R6 group gen is the most interesting spec in the game at the moment because there are so many open questions, variables and configurations you can try.


Me too. I love the skill cap on R6, the flavor, the variation in skills (lots of gens are viable; BOH, FITL, Serenity, and even Implosion are all competitive sub-ins). At the other end, the Dash dmg R6 specs are also very interesting. Best-designed spec in game IMO.

02/26/2017 01:09 PMPosted by Davlok
TC is only the top generator for SINGLE target. As soon as you are in combat with more than 1 target, you're better off using any other generator. So I'd still be in the Q/WF/SG boat overall.


Good barbs and zmonks will facilitate moving adds away from the RG so that your 1-target time is maximized. I can see TC shining as the new dps gen.

Davlok - can you update the single target dps chart to show TC? It would be helpful if you also showed a TC - 2 targets for compare.

The top seasonal clear over weekend was a TC/Forsight/Blazing Fist tri-gen. If TC is substantially greater dps, he's probably using those alt gens just as buffs (as opposed to buffs + COE element dps). As I noted in original combo-gen contemplations, Blazing Fists is just a powerhouse supporting gen. It's interesting that he's gone Blazing Fists over high dmg WFF.

I intend to redo some 1 min stricken tests as suggested by nand. Possible that TC/Blazing Fists are producing much greater Stricken.
02/26/2017 10:46 PMPosted by nand
The search space of possible generator combinations is in the ballpark of Sum[ n∈[1,4], binom(4,n) * 5^n * n ] = 4320 setups. This would be trivial for a computer to brute force with a simulation, but very difficult to do by hand with d3planner.


You might as well use MATLAB or similar for that. Writing it in Python or Perl or whatever is kind of making things unnecessarily hard. MATLAB in many ways facilitates these type of calculations and in fact includes some hooks into Excel, as well as its own plotting utlities and such. If you so desire you can create crazy 3D heatmaps or whatever based on many variable analysis.

I'm not sure what you're using the binom option for (been a while since statistics), why aren't you using Combination or Permutation terms? For this, permutation terms are probably more valid, but that's assuming you consistently had a certain amount of slots you wanted to take up. You would also need to simulate not having any but one generator at a time with its supporting skills vs using multiple generators...

I mean this a bit more complex than brute forcing a simple formula. There are always certain types of engagement strategies, such as using generators as primarily buffs vs using them on their own cycle (or maybe a bit of both, if the cycles are adjacent).

That's why I think that d3planner is great, because it does a lot of the legwork and then you just need to fill in your strategies. I think if anything what'd be wonderful is if we could automate on top of d3planner. That is, programmatically fill in combinations of scenarios and then have it do its simulations normally. And technically, you can. It's open source. So again I'd just nab it and hack it for your purposes. Maybe create an SQLite database for recording results. I just think reinventing the wheel in this case is a bit pointless.
02/27/2017 11:57 AMPosted by Vox
Davlok - can you update the single target dps chart to show TC? It would be helpful if you also showed a TC - 2 targets for compare

Already did:
02/26/2017 06:24 AMPosted by Davlok
2.4.3 Single Target DPS Chart
http://i.imgur.com/t2yCyPd.png

I'll see how a TC vs 1,2,etc chart looks, but most likely it will just be simple bar chart @ 5 aps.

02/27/2017 11:57 AMPosted by Vox
The top seasonal clear over weekend was a TC/Forsight/Blazing Fist tri-gen.

Nice! I'm also curious as to why CW seems to be stacking stricken the fastest, and Wothf seems to be stacking Stricken faster than it should be. I suspect it might have something to do with the 2nd strike's multi-punch delay or something >_>
02/27/2017 11:57 AMPosted by Vox
If TC is substantially greater dps, he's probably using those alt gens just as buffs (as opposed to buffs + COE element dps)

CoE is a massive 300% damage buff that's multiplicative with everything. It doesn't matter how weak your generator is, with a 300% unique elemental buff on top it will be the best ability to use during its rotation.

That's actually part of the reason why R6 multigen is so absurdly powerful: We can double- and triple-dip from convention of elements.

02/27/2017 01:38 PMPosted by StoleOwnCar
You might as well use MATLAB or similar for that. Writing it in Python or Perl or whatever is kind of making things unnecessarily hard.

I'm a very experienced Haskell programmer and have written multiple special-purpose simulators for video games in the past, for example:

https://github.com/haasn/DiabloM6 (season 1 M6 sentry DH)
https://github.com/haasn/heroline (heroline wars starlight)

It's worth noting that both of these are significantly more advanced than what I imagine an R6 generator simulator to be, because there were more variables at the time (such as the rest of your gear) - but that's mostly static with the R6 gen. The only real variables for R6 gen are attack speed and the generator specifics (frame times, proc coefficient, weapon damage, etc)

I would be confident in my ability to write an in-depth and faithful special-purpose simulator/optimizer for R6 as well, given enough knowledge about game mechanics. It's also very easy to visualize plots and stuff with Haskell, for example as seen in https://haasn.xyz/posts/2017-02-18-the-diablo-iii-paragon-system-visualized.html

The primary focus is still testing out these game mechanics to confirm the theory matches the reality.

02/27/2017 01:38 PMPosted by StoleOwnCar
I'm not sure what you're using the binom option for (been a while since statistics), why aren't you using Combination or Permutation terms? For this, permutation terms are probably more valid, but that's assuming you consistently had a certain amount of slots you wanted to take up. You would also need to simulate not having any but one generator at a time with its supporting skills vs using multiple generators...

binom(N,M) is the number of ways you can choose M out of N. In this case, for example, binom(4,2) is the number of ways you can combine 2 generators (out of 4 being available)

binom(4,1) = 4: FoT, WotHF, CW, DR
binom(4,2) = 6: FoT+WotHF, FoT+CW, FoT+DR, WotHF+CW, WotHF+DR, CW+DR
binom(4,3) = 4: FoT+WotHF+CW, FoT+WotHF+DR, FoT+CW+DR, WotHF+CW+DR
binom(4,4) = 1: FoT+WotHF+CW+DR

This is order-invariant, since picking e.g. TC first and FoF second is the same as picking FoF first and TC second.

Permutation terms are the exact opposite: They count the number of ways you can order a pre-determined number of options, without caring about what those choices are. For example, suppose you already picked FoT+CW+DR, then the number of permutations is 3! = 6:

FoT, CW, DR
FoT, DR, CW
CW, FoT, DR
CW, DR, FoT
DR, FoT, CW
DR, CW, FoT

For the purposes of optimizing R6, it's clear that the order of abilities doesn't matter - just the choices. Not sure what you mean by “combination term”, but it's probably the same thing as binomial?
02/27/2017 02:41 PMPosted by Davlok
Nice! I'm also curious as to why CW seems to be stacking stricken the fastest, and Wothf seems to be stacking Stricken faster than it should be. I suspect it might have something to do with the 2nd strike's multi-punch delay or something >_>

As mentioned, feel free to upload a bunch of clips of you hitting mobs with random abilities/runes and a constant damage setup. I'd be happy to analyze the footage on a frame level and report my findings / confirm yours.
CoE is a massive 300% damage buff that's multiplicative with everything. It doesn't matter how weak your generator is, with a 300% unique elemental buff on top it will be the best ability to use during its rotation.


You might want to report that to the d3planner developers too, then, because from most of the simulations it seems like switching generators even on their CoE cycles does not produce the expected dps increase.

02/27/2017 02:44 PMPosted by nand
binom(N,M) is the number of ways you can choose M out of N. In this case, for example, binom(4,2) is the number of ways you can combine 2 generators (out of 4 being available)


So you're basically talking about a special case combination without repetition. Maybe it's just been too long as I've just never heard of termed a binomial coefficient. At first I thought you were somehow applying the binomial distribution here, which would be weird.

As far as permutations, ordering actually is kind of relevant, especially if we're talking about when you decide to use the abilities... but let's just put that aside. If you're confident about writing this (or rather I suppose you're less lazy than I am), I guess go for it. Though next patch gen monks may become obsolete in favor of MH Hydra Wiz anyway? At least that's what some people are saying. I'm too lazy to make this big of an effort for this game personally, so more power to you. I feel like you're putting in a lot more work into this game than the developers are.

It's just that when I look at this myself I can see of a loooot of little tweaks that you can do that you wouldn't initially think of in order to expand the simulation space. Have fun though.

I'm a very experienced Haskell programmer


Yeah, use what you know. Personally I've used a lot of languages, including at some points assembly and vhdl. I see all languages as basically tools to serve certain needs optimally but not others. So I was just recommending MATLAB as it is what I used in some digital signal and image processing classes in order to do complex computations and things like FFTs and spectral mappings. Also used it to do some interesting curve fitting exercises in my differential equations class. I've never actually used Haskell, so I'm not sure what its general use case is. Matlab just provides a lot of built in functionality for complex computations.
CoE is a massive 300% damage buff that's multiplicative with everything. It doesn't matter how weak your generator is, with a 300% unique elemental buff on top it will be the best ability to use during its rotation.


Looking at the revised STDPS chart, TC is approximately twice as strong as some gens (non-Mangle Crippling Wave for e.g.). That is basically the break even point.

Since TC does twice the dmg (100%) and benefits from 60% elemental and 30 skill% (nearly 100% more), running TC during alt gen element is essentially equivalent to running the alt gen during its element. But, I agree, your suggestion to always run the alt element during its COE nearly always true, and a good simple rule for R6 monks to use.

Exceptions occur if TC stacks stricken more effectively and/or maintains spirit more effectively than the alt gen, or if some breakpoint puts TC at substantially more than 2x alt gen dps. In those cases, it is probably better to run TC during alt gen COE phases.

In reality, having less to focus on during play is also of benefit to R6 specs that depend on positioning for survival and resource state for edps. Running TC through alt gen elements lets you focus more on gameplay to maximize survival/dps.

Personally, I'd probably stick with TC through the worst dps alt gens (non-mangle CW, non-SearingGrasp DR).
02/27/2017 03:42 PMPosted by StoleOwnCar
You might want to report that to the d3planner developers too, then, because from most of the simulations it seems like switching generators even on their CoE cycles does not produce the expected dps increase.

It seems to work as intended for me. For example, in this d3planner setup:

http://www.d3planner.com/175738592

This sims to 38.205 T DPS, and removing the conditional “deadly reach” usage sims to 36.728.

Even removing quickening and using the absolute weakest generator (deadly reach physical) in its place, casting it during the physical element of CoE brings my DPS up from 34.446 to 34.502.

02/27/2017 03:42 PMPosted by StoleOwnCar
As far as permutations, ordering actually is kind of relevant, especially if we're talking about when you decide to use the abilities...

I suppose that's sort of true, but I still think my weaker upper bound holds because the priority of which generator to use when is going to be fairly static - you're going to be holding down the generator that your gear is specialized for, until a chosen CoE rotation comes up (and you switch to that element's generator).

Also, I think in practice, for the most part you're almost surely going to want to have three different elements for your runes, so you could probably make an even weaker lower bound by assuming that you're going to pick N elements and distribute them among your runes somehow. Either way, fine-tuning the upper bound, while interesting, isn't really relevant - my point was to demonstrate the order of magnitude (several thousand).

02/27/2017 03:42 PMPosted by StoleOwnCar
Maybe it's just been too long as I've just never heard of termed a binomial coefficient.

https://en.wikipedia.org/wiki/Binomial_coefficient

It's very tightly related to the binomial distribution, because the binomial distribution is defined in terms of the binomial coefficient. (For example, the “5 successes and 7 failures” contribution in the binomial distribution has weight binom(5+7,5) * p^5 * (1-p)^7)

02/27/2017 03:42 PMPosted by StoleOwnCar
I feel like you're putting in a lot more work into this game than the developers are.

But that's the fun part, isn't it? :p

02/27/2017 04:04 PMPosted by Vox
In reality, having less to focus on during play is also of benefit to R6 specs that depend on positioning for survival and resource state for edps. Running TC through alt gen elements lets you focus more on gameplay to maximize survival/dps.

That's very true. There's always a difference to be made between what's best assuming optimal play and what's best assuming realistic play. In practice, you'll also have a hard time keeping adds away from the RG unless the zdps players are perfect as well.

Another thing to consider that only an optimal bot could do correctly is making sure always weave your buff generators in during the optimal point in the cycle - i.e. replace the first hit, not the third hit.

Perhaps you could even do some extreme min/maxing on paper where you use different generators alternatingly in order to fish for the maximum about of effective attacks per second. (e.g. switching from a 6/7/9 generator to a 6/6/9 generator for the second hit in your 3-hit combo)

02/27/2017 04:04 PMPosted by Vox
Since TC does twice the dmg (100%) and benefits from 60% elemental and 30 skill% (nearly 100% more), running TC during alt gen element is essentially equivalent to running the alt gen during its element. But, I agree, your suggestion to always run the alt element during its COE nearly always true, and a good simple rule for R6 monks to use.

You forgot to account for the fact that 30% skill damage is more like 10% DPS due to diminishing returns from all of the extra additive skill damage you're getting from group buffs. Part of the reason I always advocate including all group buffs in your DPS simulations.
02/27/2017 04:28 PMPosted by nand
You forgot to account for the fact that 30% skill damage is more like 10% DPS due to diminishing returns from all of the extra additive skill damage you're getting from group buffs. Part of the reason I always advocate including all group buffs in your DPS simulations.


I haven't forgotten. Just oversimplified.

The additive portion also doesn't add to the elemental multiplier, as my oversimplified description suggests. I think the 30% needs only be a 1.25x mod to reach the 2x overall when combined with 1.6x from elemental.

I often say the same thing regarding additive buffs (that buff X is actually buffing X/3, and therefore don't bother). But, the reality is, i'm not sure exactly how much additive is floating in zmonk/zbarb/WD meta vs. RG.

Do you know how much exactly the 30% skill damage is as an effective multiplier with standard additive buffs in play?
Another thing to consider that only an optimal bot could do correctly is making sure always weave your buff generators in during the optimal point in the cycle - i.e. replace the first hit, not the third hit.


For some, you can queue as you dash, forcing 1st/2nd hits of specific gens while not losing a beat.

This is part of the reason I like Crippling Wave Breaking Wave over Deadly Reach Forsight in practice. The opportunity cost of Breaking Wave is lower, since it doesn't require a 3rd strike (at least that's my understanding...). It can weave in following dash, replacing only the 1st strike on a guaranteed basis. (It also is 10% additive WD dps if you aren't pulling).
02/27/2017 05:29 PMPosted by Vox
Do you know much exactly the 30% skill damage is as an effective multiplier with standard additive buffs in play?

Let's do a quick estimation by looking at the current rank 1 setup. I'll use the EU season leaderboards for convenience. I count:

Witch Doctor:
Hex - Jinx: 15%

Monk (zDPS):
Strongarm Bracers: 30% (assuming it's active constantly, don't know how realistic that is)
Gem of Toxin: 10%
Mantra of Conviction: 16% (Inna-enhanced)

Barbarian:
Threatening Shout - Falter: 25%

Monk (R6):
Combination Strike: 30% (assuming you maintain buffs from all three generators)
Skill% damage rolls: 30%
Blinding Flash - FitL: 30% (current top clear doesn't use it, in favor of breath of heaven - so I'll exclude this)

On top of this, there's also the Oculus Ring proc, which I think provides additive skill damage? Not sure.

As a hard baseline, let's assume you have no strongarm and no oculus. That would add up to 96% skill damage, which about halves the effectiveness of your skill damage rolls. (So you'd get 15% more DPS overall for your “attuned” generator)

If we assume oculus is additive (again, not sure) and strongarm is up, it would be more like 211%, turning your 30% rolls into 9% more DPS in practice.

I'm actually surprised by how “strong” additive skill damage is these days. In the past, both depth diggers and simplicity's strength also provided additive skill damage, and the zmonk also had more (breaking wave, forbidden palace). Back then, additive skill damage rolls were absolute trash.

Edit: Seems like oculus is multiplicative. So 13%-15% is probably a good estimate.
Old Oculus was additive 40%. New is multi.

You have omitted at least Toxin 10% on barb and debuff EP 15%.

What's the up time on WD Jinx debuff?

Assuming 1) full uptime on Jinx, 2) 30% Combo Strike and 3) no Forsight/Breaking wave, I get total additive of 121% without gear skill bonus and 151% with the 30% gear skill bonus.

251/221 = 13.6%.

Agree, that it's typically in the 13-15% range estimate you provided, higher (15.7%) for non-combo strike variants, lower (12.2%) for variants with combo strike & Forsight + Breaking Wave.
___

In this 13.6% scenario, the main dps gen receives a net multi of 1.82x from 60% elemental + 30% worn skill.

At the low end (12.2%), the main dps gen gets 1.80x.

At the high end (15.7%), the main dps gen gets 1.85x.
02/27/2017 06:44 PMPosted by Vox
You have omitted at least Toxin 10% on barb and debuff EP 15%.

Monk in the group I was referencing was not using EP. Also, toxin doesn't stack from what I read online.

02/27/2017 06:44 PMPosted by Vox
What's the up time on WD Jinx debuff?

Oh, it's shared with CC resistance DR, right? If so, I dunno. Probably similar to the uptime of strongarm.

P.s. If we want to absolutely min-max damage, then you could fit breaking wave into the zmonk's setup. This would require either the zbarb using iceblink (+ whirlwind or something to proc it), or the zmonk using wall of wind instead of implosion.
Vox is outlining exactly what I was talking about. I did notice an increase in damage from maybe 2 gens, but only if they were like TC and WOTHF with the flurry rune. And even then the increase wasn't as big as I wanted it to be. Another thing I tried was distributing the elements and skill damages equally. IE rolling half cold and half lightning with half FOT and half WOTHF... somehow it came out to less. That was in d3planner though.

Edit: Oh I know why it came out to less. Because you're not using them 50/50. You're using one roughly 80% of the time while the other one is only used on its cycle... I guess that's it.
Posted by Davlok
2.4.3 Single Target DPS Chart
http://i.imgur.com/t2yCyPd.png


Quick clarification question. So the breakpoints are just including Transgression, FD proc, and dashing strike radiance IAS increases right? So that means the highest breakpoint for FoT thunderclap is only 4.49 and does it say 4.95 or 4.99 as the breakpoint for WFF? I was planning to take out some IAS from my gear in exchange for other stats if I only need 4.49 for FoT thunderclap. If that is the case, then what is the point of even running a 2nd generator? With a lightning build I can just run thunderclap the whole time for the RG and even replace my 2nd generator to include FitL?
02/27/2017 09:22 PMPosted by StoleOwnCar
Another thing I tried was distributing the elements and skill damages equally. IE rolling half cold and half lightning with half FOT and half WOTHF... somehow it came out to less. That was in d3planner though.

Distributing (additive) multipliers always comes out to less, unless your generators are all the same strength.

For example, suppose you have 3 generators, and you could either get 60% to one element or 20% to each. In universe A, you have 60% increased DPS (but only 1/3rd of the time), and in universe B you have 20% increased DPS all of the time.

If your generators all deal the same damage, then there would be no difference. But since some generators are stronger than others, you end up with disproportionately more damage during that 60% DPS window than you do in the other 2/3rds of the time. As a result, your overall DPS is greater than 20%.

Bottom line: Even in the ideal case, it would be the exact same as putting all elements into a single generator. But in reality, some generators deal more damage than others, and more importantly, you'll be using one generator more than others (due to CoE downtime), so it's never a good idea to spread your elements.

Spreading your multipliers is bad in general in Diablo, especially for multiplicative effects.

02/27/2017 10:33 PMPosted by BaesedGod
Quick clarification question. So the breakpoints are just including Transgression, FD proc, and dashing strike radiance IAS increases right?

Transgression, FD and Dashing Strike all increase your sheet APS, which has a hard cap of 5.0. This chart is using your sheet APS on the x-axis.

So for the purposes of that chart, it doesn't matter what attack speed effects you're using to get that sheet APS. It only shows you how much damage you'll end up doing.
Since I don't have a group of 4 players available, plus to eliminate variance as much as possible, I decided to try and simulate GR120 RG conditions using T13 ghom, similar to what Vox did. (But I wanted to be as independent of his tests as possible, so I used my own thought process)

Judging by clear videos on youtube, a RG in GR120 has about 2e16 life. In contrast, ghom on T13 has about 7e11 health. In other words, the difference in damage between T13 ghom and a typical GR120 RG is about 30,000x. (Note: You'd expect T13=GR60 vs GR120 to be about 1.17^60 = 12,000x difference, but in reality, ghom has signifiacntly less base health to begin with. So I used the actual HP numbers instead of estimating)

So in other words, if you can achieve a straight-up 30,000x reduction in damage without affecting your core stats and mechanics (attack speed, elemental damage, stricken, shenlong's etc) then you can practice GR120 RG fights on T13 ghom.

On top of this, to eliminate differences due to additive damage diminishing returns, I would highly recommend removing one skill% damage roll from your gear. (I removed 15% from my belt, leaving me with 15% overall).

Here's a list of what I removed:
1. Crit chance, crit damage. This includes green gems (replaced for life per hit) on weapons, and paragon points (except attack speed). I equipped random/badly rolled legendaries wherever possible, but I made sure to keep on attack speed, skill damage and elemental damage. (To avoid disturbing breakpoints or affecting the balance between generators)
2. Weapon damage roll. I'm using non-ancient shenlong's with 7% attack speed, life per hit, and no damage roll. I also made sure to have no damage rolls on my rings.
These two combined reduce my paper damage by about a factor of 18.8, without having to go into specifics about what exact contributions went away where - since none of these stats affect my gameplay or generator choices.

3. 44% bane of the trapped => replaced for gizzard to help survive
4. 70% simplicity's strength => replaced for invigorating gemstone (CC immune, healing)
5. I've removed the 6-set bonus. This doesn't really change my gameplay as long as I make sure to always use dashing strike every 6 seconds (just like I would in reality), so it's a straight up 13.5x damage multiplier that's gone.
6. Depth Diggers removed from cube (100%)
7. Endless Walk set (100%)
8. Rolled elite damage off SoJ (30%) and weapons (16%) = 46%

On top of this, you also lose like 2x-4x damage due to lack of group buffs / oculus ring. All said and done, removing all of these is equal to a damage gap of:

18.8 * 1.44 * 1.7 * 13.5 * 2 * 2 * 1.46 * 4 ≈ 14513

It's not quite the 30,000x I was going for, but my gear is also not as insane as those GR120 monks to begin with, so it's more or less good enough.

I will be doing some tests with this setup and recording ghom kill times under various playstyles. One thing I immediately noticed, however, is that I have absolutely zero spirit problems., without breath of heaven, and without holding down quickening. Even if I use dashing strike, I'm instantly back up to max spirit.

So I still don't understand why top groups are still using breath of heaven in particular. It just seems to provide nothing, since I'm constantly full spirit. Is there something I'm missing?

Edit: After doing some more tests, I do notice that I actually drop spirit sometimes when overdashing and FD hasn't procced. So maybe there is some merit to keeping at least quickening in there.
03/01/2017 02:48 AMPosted by nand
It just seems to provide nothing, since I'm constantly full spirit. Is there something I'm missing?

I WAS missing something. I forgot to turn off the 10% spirit regeneration from the templar. That does actually make a difference, although it's slight.

My test results. A caveat to begin with is that I was only able to have two elemental rolls of my choice, and I removed the third element to compensate. This technically slightly disturbs the balance between generators, but I believe it's close enough. For each test, I was getting 40% towards chosen element. I was also missing 1% WAS on one weapon.

Q / WFF / SG / FitL, attuned to WFF:
5:18 = 318 seconds
4:55 = 295 seconds
5:28 = 328 seconds
-> mean 313, stddev 16.9, sigma confidence 285 - 341

I was dashing every 5 seconds or so, because I don't have a good timer for when 6 seconds are elapsed. I just let the buff run out and counted a second or so. I was holding down WFF whenever not in physical or fire. I sometimes switched to quickening for a second when I noticed spirit struggling, to get it back up to full.

Q / WFF / SG / BoH, attuned to WFF:
5:41 = 341 seconds
5:01 = 301 seconds
-> mean 321, stddev 28, sigma confidence 292 - 349

I was playing more or less as before, but using BoH whenever I noticed my spirit dropping.

Q / WFF / SG / FitL, attuned to Q:
4:58 = 298 seconds
5:02 = 302 seconds
5:13 = 313 seconds
-> mean 304, stddev 7, sigma confidence 296 - 312

I was holding down Q constantly, and dashing every 4 second (since I had ample spirit), reserving FitL for physical element.

TC / BF / BW / BoH, attuned to TC, low stacking:
5:48 = 358 seconds
5:42 = 342 seconds
-> mean 350, stddev 11, sigma confidence 338 - 361

I was holding down TC constantly, and only witched to BW and BF during their elements. I used BoH whenever I was having spirit problems. Note: In theory, the best way to play this spec would be to weave in BW and BF to provide buffs. But since I had enough on my plate trying to maintain spirit and hitting the right rotations while avoiding overdashing etc, I decided to forgo the buffs for now, in order to avoid messing up the tests due to constant misplay. I also barely have any crit chance due to the nature of the test, so I can't really reliably proc BF either way.

TC / BF / BW / BoH, attuned to TC, aggressive stacking:
7:38 = 458 seconds

Until the boss was at 75% life, I was using BW only - no element rotation at all. In theory, this should stack stricken faster, in exchange for doing less damage early. Totally not worth it, since it took ages to even get it down to 75%.

TC / BF / BW / BoH, attuned to TC, medium stacking:
5:38 = 334 seconds
6:27 = 387 seconds
-> mean 360, stddev 37, sigma confidence 323 - 397

Until the boss was at 50% life, I was using BW as my filler gen, but still switching to TC/BF for their elements. The high second score was because for almost the entire fight my templar would just refuse to taunt away one annoying add that kept following me. I was doing halved TC damage for basically the entire fight as a result. Highlights one of the major problems of the spec.

So far I much prefer the standard Q / WFF / SG spec, either Q-attuned or BFF-attuned. In groups, BoH probably wins out over FitL because of additive diminishing returns, and because of the lack of the templar's 10%.

Any other crazy ideas you want me to test?
What's interesting to me is how powerful Quickening is as the main/only generator. Should lead to very stable and easy runs. Do you even need BOH with it?
Adding some more data points to the more interesting specs.

03/01/2017 05:14 AMPosted by nand
Q / WFF / SG / FitL, attuned to Q:
4:58 = 298 seconds
5:02 = 302 seconds
5:13 = 313 seconds


5:18
5:31
5:38
I was using a 6 sec DS recast for these. After playing this spec for 10 ghom runs or so, I've figured out a reliable way of recasting on the exact 6 second mark: by aligning myself with the CoE cycles. I'll alternate cadence between recasting DS exactly on the CoE transition and recasting it on the halfway point. Took a bit of getting used to because of the element switching, but very doable after just a bit of practice.

5:43
5:20
5:22
I was using a 4 sec DS recast for these again. Although when FD was unprocced, I reverted back to 6s recast during cold/fire just to avoid messing them up with low spirit. (Switching back to the normal cadence after holy)

Join the Conversation

Return to Forum