PVE Raiding Compendium - Page 110

Morwen · 04/29/08, 3:31 PM

Originally Posted by Elija

Hello, is there any qualified calculation about the new mgt heroic trinkets DPS: Timbal's Focusing Crystal ?

Is it better than Mags Eye or even Crescent? (depending if you cast 1, 2 or even more dots)

As mentioned there are some calculations on wowhead, although they tend to be mathematically questionable (most of them assume something like E[1/X] = 1/E[X]).

Here are some actual values from shadow priest testing (who should generally have a higher proc rate)
TBC Shadow Priest 101: How to Melt Faces Effectively

In general it is pretty close to the Icon, and due to luck can be better or worse (it is not quite as good for warlocks as for priests). Given that, I still prefer the Icon for a couple reasons: in a short fight you will get more than the average for a clickable trinket, its timing is controllable, and you know exactly what you're getting out of it every time you click.

Edit: added math.

Part I - Computing the average damage

Tooltip average is 380.

Multipliers:
CoS - 1.13
Weaving - 1.1
Misery - 1.05
Crit - 1.1 at 20% crit
ISB - 1.12 at 60% uptime
Hit rate - 0.99
Partial Resist - 0.94 at 6% partial resist rate

Average damage = 569.

Part II - Computing the dps based on the proc rate distribution

Define
k := number of DoT ticks per second
X := the period of time elapsed until a proc, when the cooldown is ready.

Let f be the probability density function of X, and F be its cumulative distribution.
In other words, F(x) is the probability that a proc happens in the first x seconds.

Then F(x) = 1 - 0.9^(kx)
= 1 - e^(ln(0.9^k)*x)

So X is an exponential distribution with lambda L = -k*ln(0.9),
and f(x) = L*e^(-Lx).

Expected dps of proc = Average damage per proc * E[1/(15+X)]

E[1/(15+X)] = Integral of (1/(15+x) * L*e^(-Lx) dx) from 0 to positive infinity

Sample values
k = 1/3 --> E[1/(15+X)] = 0.03134, dps 17.8
k = 2/3 --> E[1/(15+X)] = 0.04042, dps 23.0
k = 3/3 --> E[1/(15+X)] = 0.04545, dps 25.9

In terms of comparing this to Icon's activate effect, since the dps value 
of +dmg is gear-dependent you should just plug your gear/spec in the 
spreadsheet and check the dmg-to-dps conversion rate in a zero-lifetap 
situation.  For me it's something like 21 dps on average and scales with 
haste effects whereas Timbal's doesn't.

Lock · 04/29/08, 3:45 PM

I was woundering why the dps spreadsheet goes up when using dark runes, do they not have a global cause wouldnt lifetapping be more efficent?

Flamingcloud · 04/29/08, 3:56 PM

They do not trigger the global. As far as I know, no consumable instant item does.

Arelenda · 04/29/08, 4:11 PM

Originally Posted by Lacky

My question is the real world application of haste when you only have a limited amount of it. Do you really need a good % of haste before you see returns in the 'real world' applications? I know what the spreadsheet says but when you look at movement and slight lag during fights is it worth sacrificing crit for a little bit of spell haste. These are the two gear scenarios i am playing with.

spell hit caped in both
Gear Set A: Dmg-1267 Crit-20.91 Haste-2%(2.43 sec SB)
Gear Set B: Dmg-1277 Crit-22.18 Haste-0

*edit* btw these stats are all unbuffed

You won't get any noticeable difference, since it's too low to stand out. Shadowseer used to be able to browse a combat log and estimate what your gains would have been if you had 100 more spellpower, 100 hit rating, 100 crit rating and 100 haste rating.

Haste typically beats any other stat, except hit, on a one-for-one basis, when you have reasonable amounts of haste (<250).

Torq · 04/29/08, 8:38 PM

Originally Posted by Flamingcloud

I am acting under the assumption it is *1.3 not +0.3 Torq.. I tried to get a shaman to heroism me 4 times but apparently there is some sort of bug where heroism wasnt hitting me.. If it is +0.3 then it is even more clear not to stack them tho.

Torq, you should do two of those graphics you do Torq with total bolts casted vs passive haste 0-400 one for trinket during hero, one for trinket after hero. Curious where the inflection points are.

Edit: Fixed my original numbers for a mistake(wrote bolt count down off by 1second), would seem as tho stacking is still better, at 200.. and the gap changes so insignificantly that it is probably always better to stack them.

Ask and you shall receive!

Surprisingly, it's actually linear.

[E2] Turns out I wasn't doing the BL calc right. Go me for always using the base cast time (DOH!) Uploading a new ver.

Okay, fixed version. Still linear, but now stacked and separate are humping each other.

Tobolin · 04/30/08, 2:35 AM

i see more and more locks going fire. what am i missing?! to me, i see shadow being more benefitial(while i agree that both shadow/fire are very good dps options) because of spriest synergy, boosting their dps and their mana return to their group. my question is more so this: my shadow dps is already top in my guild, should i consider respeccing/reenchanting/regearing to accommodate fire?

Torq · 04/30/08, 3:19 AM

Originally Posted by Tobolin

i see more and more locks going fire. what am i missing?! to me, i see shadow being more benefitial(while i agree that both shadow/fire are very good dps options) because of spriest synergy, boosting their dps and their mana return to their group. my question is more so this: my shadow dps is already top in my guild, should i consider respeccing/reenchanting/regearing to accommodate fire?

Fire dps generally = more personal dps.

It's more a case of locks respec'ing because 1) there are plenty of shadow locks to go around (every lock past the first contributes less and less to ISB uptime) or 2) they just want big numbers.

/shrug

Tobolin · 04/30/08, 3:41 AM

Originally Posted by Torq

Fire dps generally = more personal dps.

It's more a case of locks respec'ing because 1) there are plenty of shadow locks to go around (every lock past the first contributes less and less to ISB uptime) or 2) they just want big numbers.

/shrug

i dont know about your guilds, but we raid with 3 locks, one puts CoS, one put CoR, i CoD. if i go fire, im obviously putting up CoE, and basically pigeon holing our mages as fire to buff my dmg(they currently go as arc/frost =\) i dont know, i do comparable dps to what every other lock thats farmed illidan since august does that ive seen(2200-2600) depending on the fights im not threat capped(GG tanks with A.D.D).

and about big numbers... how much higher am i going to see? im already destro with 1440 shadow dmg(soulfrost/trash cloak)

Carnate · 04/30/08, 6:23 AM

With three locks you should be using CoE instead of CoD anyways. As long as the raid has two or more mages CoE is better dps.

Fire can be better personal DPS but you bring nothing to the table and can hurt your raid by not adding to the ISB. Shadow also provides the side buffs of Nether Prot and Soul Leach.

Side note last I heard Fire was the better DPS spec for mages anyways. Ours all go fire cept on Illidan (parasites and flames damage).

Arelenda · 04/30/08, 7:02 AM

Originally Posted by Torq

Fire dps generally = more personal dps.

It's more a case of locks respec'ing because 1) there are plenty of shadow locks to go around (every lock past the first contributes less and less to ISB uptime) or 2) they just want big numbers.

/shrug

I _really_ wish people would stop propagating this "dps uptime" myth.

Additional locks do improve ISB uptime by less when you have more shadow locks, this is true. But you're also increasing dps uptime for more people.

ISB uptime comparisons are only valid if you have an equal number of shadow users, like when you're calculating what crit rate increases do for you. For everything else you should take a look at total damage boosted by ISB.

And you really do want to use CoE if you have a somewhat competent mage. You do about 10-12k damage per minute with CoD, so if there's 100k damage per minute (1666 dps) in fire/frost CoE is better. Not to mention CoD can be totally or partially resisted and needs to be recast every minute, and is more threat to the warlock.

Madlax · 04/30/08, 8:17 AM

Originally Posted by Arelenda

And you really do want to use CoE if you have a somewhat competent mage. You do about 10-12k damage per minute with CoD, so if there's 100k damage per minute (1666 dps) in fire/frost CoE is better. Not to mention CoD can be totally or partially resisted and needs to be recast every minute, and is more threat to the warlock.

I second that.
With current gear and boss fights, CoE should always be put up instead of CoD as 3rd curse once you have a properly dpsing mage. In a fight like Brutallus, every mage should reach 600k damage done on his own.

Alazin · 04/30/08, 8:59 AM

Does anyone have a link to the standard fire demo/sac build? I know everyone pushes shadow sac but I've always been the fire lock who provides CoE for our runs.

bambatsa · 04/30/08, 9:28 AM

Forgive me if i am off the subject here.
I am a destro lock 0/20/41_hybrid with almost full t5 set and i am wondering if there is an update on my Sextant of Unstable Currents trinket. i am using it with the Icon of the Silver Crescent. I have spell hit caped and my crit is arround 26,5 fully buffed. Have seen your suggestions on trinkets and didn't see this one amongst them...so thought i should i ask. Should i aim at a better one?
thanx

Anexus · 04/30/08, 9:34 AM

Has anybody done any math on [Fel Mana Potion] vs [Super Mana Potion]. With and without the alchemist stone. I am assuming it works for both potions. And I am prettry sure the debuff from the fel potion stacks so some kind of decreasing spelldmg over time model is needed. Not sure how to calculate the effects.

Flamingcloud · 04/30/08, 10:04 AM

Well If I take my numbers in the spreadsheet and give myself an extra 50 mp5 and remove 25 spell damage (ignoring any kind of stacking debuff) super mana pots come out way way ahead. Only time I could see fel mana pots being better is at the very end if a fight, and only once(so the effective -damage is more like -6).

Torq · 04/30/08, 11:17 AM

Originally Posted by Arelenda

I _really_ wish people would stop propagating this "dps uptime" myth.

Additional locks do improve ISB uptime by less when you have more shadow locks, this is true. But you're also increasing dps uptime for more people.

ISB uptime comparisons are only valid if you have an equal number of shadow users, like when you're calculating what crit rate increases do for you. For everything else you should take a look at total damage boosted by ISB.

I'm not sure what exactly this myth is that you mention. As far as you've stated, I was correct in pointing out that each lock added to the raid adds less to the ISB uptime than each lock before him. If I implied something else that was inaccurate, let me know and I'll correct myself, and also learn, because I honestly have no clue what you're talking about, being relatively new to this whole theorycrafting thing...

Though, it does bring up a nice modelling problem, one that I've seen been tossed around and touted as "very difficult" given the scale and complexity. I'll probably end up taking a swing at it and reaching the same conclusion myself, but it might be a fun exercise

Arelenda · 04/30/08, 12:21 PM

Originally Posted by Torq

I'm not sure what exactly this myth is that youmention. As far as you've stated, I was correct in pointing out that each lock added to the raid adds less to the ISB uptime than each lock before him. If I implied something else that was inaccurate, let me know and I'll correct myself, and also learn, because I honestly have no clue what you're talking about, being relatively new to this whole theorycrafting thing...

Though, it does bring up a nice modelling problem, one that I've seen been tossed around and touted as "very difficult" given the scale and complexity. I'll probably end up taking a swing at it and reaching the same conclusion myself, but it might be a fun exercise

I'll try and illustrate.

Compare these scenarios:

A. a raid with one lock who does 400k damage in a fight (without ISB)
B. a raid with 3 locks who do 400k damage each in the same fight. (without ISB)

Assume ISB uptime in both is 20%. This makes:

A. total damage 400k, uptime 20%, total damage by ISB alone = 400k * 0.2 * 0.2 = 16000 damage
B. total damage 1200k, uptime 20%, total damage by ISB alone = 1200k * 0.2 * 0.2 = 48000 damage

So even they have the same uptime, it results in 32000 more damage in scenario B. ISB uptime returns three times as much there.

The general formula for this is
Total damage added by ISB = 20% * (total shadow damage done without ISB) * (uptime)

Both factors matter for the net result. Looking at just one of them when the other is not constant will distort the whole.

Krazen · 04/30/08, 12:38 PM

Originally Posted by Arelenda

I _really_ wish people would stop propagating this "dps uptime" myth.

Additional locks do improve ISB uptime by less when you have more shadow locks, this is true. But you're also increasing dps uptime for more people.

ISB uptime comparisons are only valid if you have an equal number of shadow users, like when you're calculating what crit rate increases do for you. For everything else you should take a look at total damage boosted by ISB.

And you really do want to use CoE if you have a somewhat competent mage. You do about 10-12k damage per minute with CoD, so if there's 100k damage per minute (1666 dps) in fire/frost CoE is better. Not to mention CoD can be totally or partially resisted and needs to be recast every minute, and is more threat to the warlock.

Yeah, but a disproportionate amount of that damage is backloaded in the last 20%. There is no reason you cannot CoD the first 4-5 minutes and swap your curse later.

Torq · 04/30/08, 12:52 PM

Originally Posted by Arelenda

I'll try and illustrate.

Compare these scenarios:

A. a raid with one lock who does 400k damage in a fight (without ISB)
B. a raid with 3 locks who do 400k damage each in the same fight. (without ISB)

Assume ISB uptime in both is 20%. This makes:

A. total damage 400k, uptime 20%, total damage by ISB alone = 400k * 0.2 * 0.2 = 16000 damage
B. total damage 1200k, uptime 20%, total damage by ISB alone = 1200k * 0.2 * 0.2 = 48000 damage

So even they have the same uptime, it results in 32000 more damage in scenario B. ISB uptime returns three times as much there.

The general formula for this is
Total damage added by ISB = 20% * (total shadow damage done without ISB) * (uptime)

Both factors matter for the net result. Looking at just one of them when the other is not constant will distort the whole.

[E] Nevermind these, they're bad numbers! Leaving them in, though.

Oh, okay, I took that into account, already. I was merely pointing out that each shadow lock adds less to the uptime, but not less to the raid damage (as they'd be profiting off of imp scorch, instead).

Also, I looked into the ISB uptime modelling thing, and came up with this:

Let t = N(shadow spells that consume charges) per sec.

Let c(i) = cast time for lock i
Let P(i) = crit chance for lock i

Each lock has an expected reapplication time T(i) = c(i)/P(i)

Overall expected application time T0 = sum(T(i):1-N)/N or the average expected reapplication time.

Calculating t and T0 for your particular raid gives expected uptime of each ISB to be 4t, and overall ISB uptime % of 4t/T0

To verify this, I ran a simple simulation of 4 locks as (cast, crit) of {(2.5, .25), (2.4, .20), (2.5, .25), (2.3, .20)} constantly casting over 7 days time. The expected dropoff time of the ISB stack was 4t = 1.65145 sec with an expected uptime of 60.74296% with the calculations above.

These were the sim results:

168:00:00.0000>> Uptime: 386692.99914557 seconds (63.93733451481%)
168:00:00.0000>> 0 >> Hits: 241919 (241920 Expected) Crits: 60367 (24.95339349121%)
168:00:00.0000>> 1 >> Hits: 251999 (252000) Expected Crits: 50424 (20.009603212711%)
168:00:00.0000>> 2 >> Hits: 241919 (241920 Expected) Crits: 60346 (24.944712899772%)
168:00:00.0000>> 3 >> Hits: 262956 (262956 Expected) Crits: 52299 (19.888878747775%)
168:00:00.0000>> Average time to fall: 1.7306656006443 seconds

So obviously my model slightly underestimates the uptime % and average uptime. I ran the sim 4-5 times over the same duration with pretty much the same results: average time to fall was always ~.1 sec higher than expected, with uptime % being 2-3% higher than expected.

Thoughts?

Trickykid · 04/30/08, 1:22 PM

Originally Posted by Arelenda

Both factors matter for the net result. Looking at just one of them when the other is not constant will distort the whole.

It's pretty easy to figure out how much ISB uptime you lose by going fire and then multiplying that by the shadow damage in the raid to see how much more fire would have to do to be worth switching. The difference between the two relies pretty heavily on raid composition.

wind · 04/30/08, 1:53 PM

I'm currently debating the benefits of fire vs shadow with the other warlocks in my guild. This is the latest thing I came up with. I don't see where my calculations could be wrong, but I'm sure you guys will find the glaring holes in my logic(if they exist) :

Let's start with the premise that increasing raid dps is the main objective. I think we've all agreed that, with similar gear, going from shadow to fire provides a personal dps increase for the warlock. Based on 2 Brutallus kills on wws(which might or might not be considered not enough to be conclusive), I'm judging the dps gained from speccing fire to be around 10%(and I'm being quite conservative, the figure has been closer to 15% from my personal observations).

Let's assume a "normal" raid setup: 3 warlocks, 2 spriests:

- to justify the benefit of having fire locks, their 10% extra dps would have to be greater than the dps loss the 2 spriests take (due to the absence of ISB); assuming a 70% ISB uptime would mean that overall, the shadowpriests get 14% extra damage from ISB.

In conclusion, if 10%(lock total damage) > 14%(spriest total damage), all warlocks should spec fire. This translates to a 1:0.71 damage ratio for locks:spriest. Therefore, it the shadowpriests total damage does not go over 71% of the warlock total damage, all warlocks respeccing fire would provide a greater benefit to the raid.
Normal lock dps = 2.2k , normal spriest dps = 1.5k (conservative figures).
Fire lock dps = 2.4k.
Therefore, 14% of 3k = 420 dps, should be greater than 660 dps, to justify keeping all locks as shadow.

tetracycloide · 04/30/08, 2:42 PM

It's not quite that simple. Shadow priest's damage is just one part of the whole, you also have to consider that a percentage of shadow priest damage is returned to their entire party as mana and that that mana, in turn, can again be converted into DPS (assuming they're in a DPS group).

In your example the shadow priest loss of 14% of 3k damage resulting in 420 DPS loss to the two shadow priests also results in a MPS loss of 5% of 420 DPS or 21 MPS (or 105 MP5) from two groups. In addtion, on many encounters the raid utility of their HPS from DPS is also a significant factor (brutalis comes readily to mind). Even if you are stictly only interested in comparin raid DPS before and after you must at least factor in DPS changes as a result of 2 groups loosing that much mana regen.

Torq · 04/30/08, 3:00 PM

Originally Posted by Torq

Let t = N(shadow spells that consume charges) per sec.

Calculating t ... gives expected uptime of each ISB to be 4t...

Something about this part bothers me. I can't quite nail it down as to the actual mechanics behind it, nor can I figure out how to model it properly. Just because what I'm using gives me close to the simulator's answer doesn't mean what I'm doing is right, and that's what's bothering me.

Also, when I attempted to add shadow priests to the model, the expected uptime rose, instead of decreasing, which is contrary to what actually happens (more users consuming charges with less users supplying should equate to less uptime).

Someone give me some help here?

Anexus · 04/30/08, 3:21 PM

Originally Posted by wind

Let's start with the premise that increasing raid dps is the main objective.

Your argument makes sense under the assumption of this statement. However, the main objective should be to kill the boss, which includes the need to consider your tanks and healers. Dont underestiamte the value of a shadowpriest. They can make the difference between taking an additional dps instead of a healer in some cases and that will definitly boost the overall raid dmg.

Trickykid · 04/30/08, 3:27 PM

Originally Posted by Torq

Something about this part bothers me. I can't quite nail it down as to the actual mechanics behind it, nor can I figure out how to model it properly. Just because what I'm using gives me close to the simulator's answer doesn't mean what I'm doing is right, and that's what's bothering me.

Also, when I attempted to add shadow priests to the model, the expected uptime rose, instead of decreasing, which is contrary to what actually happens (more users consuming charges with less users supplying should equate to less uptime).

Someone give me some help here?

Expected uptime isn't 4t since some of those charge-consuming spells can re-applicate. I checked through your math and don't think I saw this taken into account since you just consider time per ISB and time per crit.

I think the model is much simpler than you think and was incorporated into leulier's sheet already. Just find weighted crit of all shadow nukes (with zero for SP spells) and use 1-(1-c)^4. Doesn't take many assumptions to consider that as accurate as we can get.