• frmorrison

Sunder Armor/FF/and CoR are nearly impossible to model correctly, since the gain depends on the amount of armor already on the target, which varies on nearly every mob. Expose Weakness also varies on the Hunter's agility. The other buffs are relatively easy to model.


I would think adding the other FG modeling improvements you listed would be nice to see (just more boxes to click on/off on the pet section of the page), the idiot gets quite powerful once all those buffs up, when I was FG a few months ago he did 250 dps, sadly he died too much, I am waiting for Tier 5 to be buffed to fix that issue.

 

nalr

I am looking at the spreadsheet and cant figure out what corralates to Resist Peircing , Health/5 , and Mana/5

Can any help

Nalr

 

• frmorrison

Spell penetration has no worth in PvE, so there is no stat to show it. I can't think of any good health/5 items so it is ignored as well, and MP5 is part of the spreadsheet's inputs.

Since this spreadsheets seems past your intellect, try this one out:
Warlock DPS Calculator

 

nalr

wow nice response. still not helpful. trying to do a comparision of builds with my gear and such. looking for a defination of what is MP5? I usually run with a felhunter sacrificed, is MP5 and HP5 yoru base regen on health and mana?

 

Demi9OD

MP5 is how much mana you regen every 5 seconds while casting, since a warlock should hardly ever leave the 5sr, and we have shit spirit anyways. Our MP5 boils down to Fel Hunter Sac, Mana Pots, and Shadow Priest VT.

 

nalr

Thanks that is what i was looking for.

 

Brile

first i would like to say great spreadsheet.

i've been using this from the first day it was up. i just recently switched to the new version (1.14) and i noticed that devastation is no longer listed. am i correct in my assumption that the % crit is already reflected in my mouseover, hence it is not necessary to list on the sheet?

the first version of this sheet used to take into account gear sets you had on (such as frozen shadoweave, spellstrike) and also trinkets (mark of defiance for example). i would think the bonuses you gain from these would greatly affect your DPS output and should be calculated in. am i wrong here?

lastly, the stupid questions. just to be sure: mp5 is just the mouse over from my character panel, and is wand dps even relevant at all? with dots and bolts/incinerates are we ever not casting?

 

Byram

I've been using this spreadsheet for a while now, and first of all I've gotta say it's great

The only thing I'm curious about is that everywhere I can find on this site said hit capping is the first thing a warlock should do to bolster their dps (I'm currently 43/0/18, so I can see the relevance of +hit).
However, using this spreadsheet, getting only around 7% hit (without Supression) seems to be necessary before +dmg is a better choice for overall dps. Granted, both stats are always with .1 dps points of each other, but it makes me curious as to why hit capping is indeed so important.

 

Melkor

As far as I know devastation is still not included in the character tooltip.

 

PSGarak

Hey guys, I actually just joined these forums in order to talk with the author of the awesome spreadsheet I found about how to make it more awesome.
Forgive me if some of this comes across as a bit egotistical. I'm a mathematician, I have experience modeling, I'm assuming I know what I'm doing unless proved otherwise ;P. That said, rip away if you find a flaw.

One-hit roll system
(skip if you know how this works)Currently, nukes are modeled as a two-roll system, rolling once for hit and then rolling again to see if the hit crit. Current theory says WoW uses a one-roll system, where one number between 0 and 100 is generated with your miss percent being at the top and your crit percent being at the bottom.
Current spreadsheet model (incinerate)
(1-MAX(1,base_miss-hit)/100)*(1+crit/100*(0.5+ruin*0.5))
Proposed correction:
(MIN(.99, 1-(base_miss-hit)/100) + (crit/100 * (0.5+ruin*.05))
This change also reflects that the hit cap is 99%, not 100%, as is currently believed.

Lag
Instant-cast spells are actually not affected by lag, believe it or not. The global cooldown expires on your client exactly in time for your next spell to reach the server as its own global cooldown to expire. While the server is still casting this [lag] number of seconds after the command, the instant spell also went off [lag] number of seconds later, so they are occuring with a separation of the 1.5 second global cooldown. (this assumes reasonably consistent ping) I would therefore recomend the removal of lag corrections to instant-cast spells, or any spell that casts [lag] number of seconds faster than the 1.5 second global cooldown. Currently, corruption and hasted 1.5 casts are bottomed out at 1.5+lag; they should be bottomed out merely at 1.5
Most people think of lag happening at the beginning of the cast. It's easier to track if you think of it happening at the end of the cast, when your client is still casting but the server's still done. In this model, casted spells have it and DoTs don't.

Improved Shadowbolt
If you have enough warlocks running around (my guild has 5 =P) that you can assume that ImpSB procs are always limited by the four charges, and not the twelve seconds, it's a really easy calculation to make. It's just 1-(chance that the last four shadowbolts haven't crit), or 1-(1-x)^4 where x is the crit chance.
To account for the fact that ImpSB is not affected by shadowbolts that got resisted, let x be the crit chance out of unresisted shadowbolts, that is crit/MIN(.99, 1-(base_hit-hit)/100).
Obviously in a single-person scenario, the twelve seconds is a huge concern and you factor in SB uptime, but as you get more and more warlocks, this becomes a more accurate model (and is always a cieling value). You can take an average value of x across your guild (weighted by shadowbolt dps time) to come up with your own custom value.
This becomes infuriatingly more compex if you let your shadowpriests cast mindblast or SW, so I recomend you don't =P.

Other Stuff
The programmer in me says make another column for the hit chance of each spell, expressed as a fraction. It makes your internal equations nicer.
On that note, when I was hacking it up I also made a seperate area for multiplicative effects on fire and shadow damage with totals at the end, rather than having a large string of multiplication at the end of every spell; this almost begs a column for damage type, which would also make modeling resists at some point in the future easier.

A deathcoil row is not hard to add. A second curse row for using CoA when CoD gets resisted is, but if you can handle DP/LT you can handle that =).

A threat-per-second column would not be difficult. You would just need talent rows for impDS, grim reach, imp MD, and a pulldown menu for BoS and TA totem (which I am assured do not stack), and then it's straight-up multiplication against the dps column (or the sum of the dps column and half the hps columns).

I threw in a DoT uptime column, which was just real duration/effective duration, just because I was curious. It can help if you're parsing a combat stats stuff to see if your dot gap is accurate.

Future Stuff
While iterative cycles are generally good for computing numerical answers, I like closed-form solutions, so I'll try and come up with a nice solved arithmetic formula for lifetap and shadowbolt weighting.
Wands are basically casted, so I suppose if you want to model it really good you would add a slot for wand speed as well as wand dps. However, I regard this as purely academic. Not only will I kill any warlock who uses a wand, the fact that they are not affected by school-specific agro reductions means that they actually lower your damage threshold in agro-sensitive situations. It's far better to dps very slowly with a subtletied school.

Other other stuff
I actually found that immolate is not worth me casting it, contrary to popular knowledge o O. My shadow/fire damage is pretty hefty, 1250 to 915(with DM:C up). Immolate is actually more damage per cast time than shadowbolt, as anticipated, but the extra mana cost means that the extra time lifetapping outweighs the damage gain. I was quite surprised.

Procs and stuff:
Recasting a DoT when a +dmg effect procs (like 2-piece T4) is generally not worth it. On averge you lose half a tick, around 250 damage or so, for the gain of the dot being recasted with 135. Even with a long dot and talents, the damage gained rarely exceeds 170 (this is before mob debuffs, which would affect both numbers equally). The correct usage is simply ignore it and go about your business (except maaaybe a CoA that's still in the first few low ticks). For modeling it, parse some uptime % and add that much of it to your static gear damage.

And yeah, devestation is not in the tooltip because it's destruction crit, which is not one of the categories available =P. I would recomend adding a devestation talent choice, but it's okay to leave out demonic tactics, since that already shows up.

Anywho, let the mathing begin.

 

• frmorrison

Last time I checked, the spreadsheet reflected the 1% miss rate, and no one knowledgeable here believes that there is a 100% chance to hit with a spell.


TPS adds extra complexity that do not have much value. BoS and TA stack, but multiplicatively, not additively.

 

PSGarak

Sorry for double-posting, but I decided to finally get around to closed-form non-iterative lifetap/filler balancing.
So, you have a bunch of other spells and their set-in-stone contributions to your mps and the chunk of your cast cycle they take, which we shall call M_T and %_T (mana total and time percent total... yes they're ugly variables. ask my friend about the time he defined the variable "3" in his number theory class). Your filler has a mana per second when spammed, M_f, and lifetap has a mana return when spammed, -M_L. You are attempting to find measures for the relative uptime of your filler and lifetap, %_f and %_L, such that sum(M) = 0 and sum(%) = 1.
Let's juggle variables! (for clarity, variables being solved for are bolded)
M_T + (M_f * %_f) = (M_L * %_L); %_T + %_f + %_L = 1.
Solve one variable in terms of the other:
%_f = (1 - %_T) - %_L
Substitute into other equation:
M_T + M_f * (1 - %_T - %_L) = M_L * %_L
Solve for variable:
%_L = [M_T + M_f * (1 - %_T)]/(M_L + M_f)
Solve for other variabe:
%_f = [M_L * (1 - %_T) - M_T]/(M_L + M_f)

I don't quite know how to enter this into the spreadsheet, since I'm not familiar with this vtable lookup thing where all your filler infoz seem to be stored. However, I can clearly define these variables:
M_T: (-?)sum(j17:22, j24:25)
M_f: casting cost of the filler divided by cast time (NOTE: cast time != effective cast time for channeled effects, recasts still cost mana)
M_L: (580+add_s*0.8)*(1+imp_lt/10)/1.5
%_T: sum(f17:22, f24:25)

Just as a check, make sure that sum(f17:26) is indeed 1. I sanity-checked my answers, but I chose math as a career path so my sanity is questionable.

EDIT since someone replied: yes, on second parse it does, it took a bit of inspecting to reveal it. My apologies. I thought the 1 inside the MAX function was a 1 (=100%) hit chance, when it fact it is a 1% miss chance.

 

tetracycloide

The problem with the 1-(1-x)^4 model of ISB is that there is a 12 sec time limit and 4 shadowbolts in a 12 sec window is not a given. Using the proper model, whithout cutting corners, ensure the ISB results are accurate in both situations and not just raids with 4 shadowbolts in 12 secs or less.

There's no reason at all to simplify the equation if the complex form is always right and the simple form is only right under a specific set of circumstances beacause it just limits the usefulness of the sheet.

This feature is already on the sheet, just set lifetap to 'needed.'

It is impossible to bottom out a GCD spell at 1.5 secs because the client lock is 1.5 secs long and it takes at least your lag time to get another cast off server side. The GCD does not start client side until after the message is back from the server that says casting finished succesfully, at least as far as I know.

My vanity is justified.

 

Sidewalk

I just downloaded the lastest spreadsheet and upon opening I get a Circular reference error that seems centered around I17 cell.

Looks like somebody else was also having a problem, but I didn't see them get answered.

 

tetracycloide

The circular reference is intended.

On the windows version of excel:

Allow or correct a circular reference
1. On the Tools menu, click Options, and then click the Calculation tab.
2. Select the Iteration check box.
3. To set the maximum number of times Microsoft Excel will recalculate, type the number of iterations in the Maximum iterations box. The higher the number of iterations, the more time Excel needs to calculate a worksheet.
4. To set the maximum amount of change you will accept between calculation results, type the amount in the Maximum change box. The smaller the number, the more accurate the result and the more time Excel needs to calculate a worksheet.

This is cut and pasted from the first entry that pops up on the help function within excel (open with F1) if your search string is 'circular reference.' I do not use the apple version of excel so I cannot cut and paste the requisite help entry for circular references on macs but I can only assume that it takes roughly the same amount of time (read: less than a second) to find with a minimum of effort.

My vanity is justified.

 

PSGarak

Point conceeded, for general use. It's probably more useful as a guidline for filling in "custom" ImpSB uptime when you have a large number of warlocks with similar crits.

I know the functionality is there, and it should provide the same answer up to a large number of decimal places. I was providing a solved, closed-form version of the solution that avoides the numerical approximations and occasional errors people get from circular references.

That's pretty much my point, in my experience it's not. My observation has always been that the global cooldown starts client-side when you activate the ability and is not affected by when the server returns a spell cast.
Easy way to check, how many times can you cast Detect Invisibility on yourself in the duration a single shadow ward? My result was a hair's breadth less than 19, not 18*ping less than 19.

 

tetracycloide

I'll run the same test and see if I come up with the same results.

I'll get back to you on the closed form equations it's the weekend and since I'm not at work I'm not bored enough to do math.

My vanity is justified.

 

• lebaxe

After running some numbers through the spreadsheet I have a few issues that I was hoping people would address.

From what I've gathered, the general consensus is for affliction warlocks is that spell hit scales better with any other stat until hitcap. However, after putting my gear/spec into the spreadsheet, I found in my current situation (hitcapping via suppression and having around 150 spell hit) that I'm getting an effective dps boost per point of spell hit of 0.39 compared to 0.7 per point of damage. Was I misled as to the general consensus regarding spell hit or is there something wrong with my calculations?

 

PSGarak

The idea about hitcapping assumes no supression. The idea is that you would want to hitcap your shadowbolts anyways, so the extra hit from supression becomes redundant at that point. It does depend on your spell rotation; the value of hit goes up if, for example, you're on CoS duty and dropped immolate, so you spend more proportional time shadowbolting.
Hit is the cheapest dps increase on shadowbolt for an affliction warlock. With your DoTs, it affects DoT recast time and uptime. Supression is probably worth it if you haven't capped yet (202 without dependable Totem of Wrath) but if you're at 150 spell hit you're definitely wasting a few points of supression if you have it maxed out.

 

• lebaxe

The question is why would you want to hitcap your shadowbolts if you're getting diminished returns from spell hit after you've hitcapped your dots via suppression? Honestly a raiding UA lock's spec loses nothing from putting 3-4 points into suppression, as unlocking the T4 affliction talents requires one to dump points that have relatively sketchy raid value anyways. Why not account for 4 points into suppression, get the much smaller amount of hit necessary to hitcap your dots, and then stack +damage? It seems like it pays better DPS returns.

 

Bahkauv

The answer is Soulshatter. And the around 40% of your damage that come from shadow bolts, even as an affliction lock. If you have 10% misses, you waste 4% of your potential damage.

If your soulshatter is resisted, you can start looking for that nice 150+ dps wand and forget all you ever heard about warlock dps.

I think that every warlock should try to get at 202 hit rating, or as close as possible.

 

• lebaxe

The point about less damage from shadowbolts is moot. This is something that has already taken into account in the spreadsheet modelling. The numbers clearly state that spell hit scales worse than damage once you hitcap your affliction spells with suppression. Shadowbolts comprise roughly 35% of a warlock's dps on fights where he is able to use all of his dots. But you neglect to mention that the other dots, all of which are affected by suppression (minus immo) and together comprise much more damage than shadowbolt alone, are not going to be affected by this reduction in hit. The difference is also more pronounced if you're allowed to use CoD instead of other curses in raids.

The only reason I see in weaning yourself off suppression is related to soulshatter resists. In a given fight I've found that I soulshatter at most twice, and with 4/5 suppression with hitcapped dots you stand to have a soulshatter resist 9% of the time. It seems like, until you get a large amount of the t5/t6 instance gear that already has loads of hit on it already, that there would be a point before completely hitcapping via gear where it wouldn't be entirely unreasonable to reap the better scaling that damage offers affliction warlocks while attempting to reach a balance with the potential of a soulshatter resist? I would think the point where this would be most efficient would be somewhere in between, although it's dependent on the nature of the fight, tank threat generation, etc.

 

tetracycloide

I'm at a fairly early point progression wise in what drop are available for me and I've only got two points left in suppression just by gearing up assuming hit capped DoTs. The real reason you don't just stack +dmg is because there really isn't any gear out there that lets you do that after you get through kara. I'm at 12% hit in my raid gear with tailored epics, kara drops, and T5 shoulders.

My vanity is justified.

 

• lebaxe

I believe the ramifications of decreased scaling value of spell hit versus damage can lead to a few basic assumptions about gear.

Runed living rubies will ALWAYS be superior on a dps value increase basis than veiled noble topazes if you run UA and have points into suppression while running an ideal dot rotation (including immolate)

Also, there are many gear choices where you have two alternatives, with one piece of gear containing spell hit and damage with the other having no hit but higher damage. Before I came to this conclusion I would always pick the one with spell hit, as I always saw myself working towards the gear hitcap. Now I am finding myself using the spreadsheet for each individual piece of gear and finding that a lot of my hit pieces are actually dps downgrades from the damage heavy pieces.

 

tetracycloide

I think you'll find that if you're gearing and specing apropriatly that many options with +hit, spellstrike, ruby drape, mantle of the corruptor + 2x veild noble topaz, handwraps of flowing thought, and band of crimson fury for example, are superior to other options that might have more spell damage. With hit capped DoTs +hit is worth less than before but that does not mean it's worthless.

Spell hit also scales better with your spell damage so if you are using your base stats in the spreadsheet to find optimal gearing ratios then spell hit is more than likely worth more than the sheet is giving it credit for.

My vanity is justified.