In terms of Mutilate: as I'm not currently specced Mutilate and have no real plans to do so, I haven't investigated the options as thoroughly; however, in very general terms: most of the notes on tradeoffs made for combat apply here as well - 4/5 setups frequently edge 2/5 setups, but not necessarily by enough to justify the loss of flexibility. Engineering and Blacksmithing are strong, though the balance tips more towards Engineering due to not needing Cat's Swiftness. I also find the Mutilate gear list to be a bit more static - I haven't found as many reasonable alternatives, but that might just be lack of time spent looking.
Since I do spec Mutilate, I've been spending time playing with Mutilate BIS setups, but there's not a lot to add to the gear list you found.
As you say, the only real choice is 4/5 versus 2/5 with [Geistlord's Punishment Sack] and [Cultist's Bloodsoaked Spaulders]. 4/5 wins by about 15 dps with my professions (ench/LW) and race (troll). 15 dps is a small advantage and, as per your comments on the 4 piece set bonus, it could disappear (or go negative) on any boss fight which isn't pure tank-and-spank.
That 15 dps difference was with HWT and Abom as the trinkets. WFS is an EP upgrade over HWT, but before the recent realization that the crit cap might be 4.8% higher, it was very difficult to make it actually come out as a DPS upgrade in the 1.3.2 sheet since equipping it would put you way over the cap. Now, in the 1.3.3 sheet, it doesn't cap the 2/5 gear setup, and barely caps the 4/5 setup. 2/5 gains about 32 dps, while 4/5 gains about 24 dps. That leaves 4/5 just 7 dps ahead. (Alliance results will vary.)
It's also worth noting that before you have access to hard mode Arthas 25 loot, the best weapons are double [Lungbreaker]. If you can't get two of them, heroic [Heartpierce] is the best 1.8. Assuming equal ilvl, fast MH / slow OH appears to be better than slow/fast, but only by a tiny amount (3.2 dps for [Rib Spreader] / [Lungbreaker]).
Finally, while there's usually a small DPS advantage to gemming AP instead of agility at lower gear levels, the value of agility goes up to a hair under 2 EP in BIS gear. I already gem agi at 264 gear levels since I lose less than 20 dps, and that's not enough for me to give up the dodge. At BIS gear levels the loss is about 1 dps, which makes it a no brainer to gem agi in my eyes. (Unless, of course, doing so would put you over the crit cap, which is a concern in the WFS setups.)
Last edited by Greymist1 : 02/05/10 at 7:54 PM.
Reason: fixed typos, added stuff
I have a minor nit for the latest version of the sheet--the list of 3% damage buffs doesn't include arcane mages, and FI claims to be from ferals instead of BM hunters. I'm not sure how many people actually refer to those comments to find out whether their raid has a particular buff or not, but it seems like the list might as well be comprehensive and accurate.
The table looks great though, a nice visual improvement over the old style. Thanks nelalas!
So, allow me to say a few words about the crit cap situation, as it is - as you note - fairly significant. As a general rule, if you want to follow the dialog and testing, you should keep an eye on this thread, which is where the discussion has been occurring. That said, I don't know that that thread is as clear as it might be at explaining the issues and the background, so let me take a few minutes to review the situation. I'm going to go over parts of this in probably more detail than is strictly necessary, because I think it's important to understand the whole picture here. So if I'm reviewing stuff you already know... sorry, but there's probably people reading this who aren't up on the details. So bear with me.
It's been know for years that the crit cap exists - specifically, it dates back to the discovery of one-roll hit table mechanics, which, as I recall, originated sometime in mid-to-late vanilla. It was discovered that all attack outcomes are decided with a single roll, which leads to the obvious question of "what happens if you have too much stuff to fit in 100%?" To expand on that a bit: your hit table has a guaranteed 24% glancing blows on it. You also have some dodge, parry, block, and miss chance, which eats up some amount of additional space on the hit table. You then have crit which takes up some portion of the remainder, and regular white hits comprise "everything else". So it's pretty clear that as you gain more crit, you convert more and more of your regular white hits into crits; but what happens when you "run out" of regular hits? If you keep stacking crit, can you convert the other results into crits? The answer, it turns out, is no. Crit has the lowest priority of the remaining options. So if we have, say, 24% glance, 10% miss, and 5% dodge (for a total of 39%) there's only 61% left for "other stuff". So when our crit gets to 61%, we already have all the crit we can ever use - more crit can't increase our crit rate further, as there's nothing else we can remove from the hit table.
At some point - I forget exactly when - it was noted that crit rates against bosses seem lower than would be "expected" from tooltip crit rates. And investigation of this with target dummies around the time Wrath came out showed that it appeared to be a 4.8% reduction - that is, your observed crit rate would be 4.8% lower than predicted by your tooltip. This was called "crit reduction". And - to continue our above example - it was assumed that what this means is that since you can benefit from 61% crit in the hit table, the crit cap occurred when your "tooltip" crit rate was 4.8% higher than that. So when our "tooltip" crit rate reached 65.8%, our actual crit rate against bosses is 61%, and we are crit capped.
However - and this is about where the thread I linked takes up the story - testing against a target dummy indicated that even with ludicrously large amounts of crit, we were still seeing hits against the target dummy. Such testing was done by a number of different people, but I'm going to refer to my own testing here for a moment, not in an effort to devalue the testing of others but because I can speak rather specifically to the mechanics of the testing.
So, for instance, refer to post 23 of that thread. With 27% miss, 24% glance, 13.39% parry, and 5.89% dodge, there's only room in the crit table for 29.72% of other stuff; hence, under the standing theory at the time, I'd need 34.52% crit to be crit capped - this would result in an actual crit rate of 34.52 - 4.8 = 29.72%, the exact amount of room remaining. And since my tooltip crit rate was 43.77%, I shouldn't see any regular hits at all.
...but I did. Specifically, about 5% of my attacks were registering as regular hits. As a result of testing like this from myself and others, it was conjectured that there was some minimum amount of hits that must always remain on the hit table - that is, you couldn't remove hits entirely, and thus always had "about" 5% hits no matter what you did. And there seemed to be a certain elegance to equating the (roughly) 5% forced hits with the 4.8% crit reduction, which is where the "crit conversion" theory came from. Under this theory, the crit "reduction" is actually converting crits to hits after the crit cap is calculated - so, in this case, I only had room for 29.72% "other stuff", so the rest of my crit was pushed off the table; and then 4.8% of those crits were converted back to hits, resulting in an observed crit rate of 24.92% with 4.8% hits.
However, what was recently discovered, tested, and confirmed (see posts 49 onwards) is that, in effect, this theory was never correct to start with. The "roughly" 5% of forced hits weren't really hits at all, and were never coming from crit conversion or any other such mechanic. Rather, their origin was much simpler. In order to get to the crit cap without raid buffs for testing purposes, we were always attacking from in front of the target dummy so as to get parries. What we forgot is that doing so also means we get blocks. And it so happens that Recount reports attacks that are partially blocked as hits in it's summary. I've heard this referred to as a "bug" in recount, which is probably overstating the case; while I think it would be a little clearer if it reported them as blocks instead, I don't think the current behavior is unreasonable. Regardless: based on this functionality, we looked at our parses and saw "about" 5% hits and concluded that it was some crit conversion thing... when in reality, those were just blocks taking up extra space on the hit table. And if one repeats this test from behind, negating both blocks and parries, one can pretty easily accumulate as many swings as you might want to where every single swing is a crit, glance, dodge, or miss. That is, hit can be totally pushed off the hit table, and the whole "crit conversion" thing is basically nonsense.
Note, however, that crit reduction is still real - it occurs, and has been clearly tested. So what this means is that the accepted theory as of when that thread started - as summarized in the third paragraph of this post - is actually, as near as we can tell, the way it works. And probably always has been. And under this theory the crit cap is 4.8% higher than it is under the crit conversion theories, as we aren't stuck with 4.8% hits in our hit table. In short: the fact that the crit cap was lowered by 4.8% when I released the early sheets for 3.3 was a mistake based on an incorrect theory due to misinterpretation of testing. The reason why the crit cap is 4.8% higher in 1.3.3 and 1.5.1 than it was in 1.3.2 and 1.4.4 is that that change has been reverted in light of testing showing that that theory is - and most likely, always was - bunk.
Hence - while such mechanics are always a bit uncertain (I mean, we've just demonstrated how easy it is to make a mistake) - at this point there is no reason to believe that the crit cap works any differently than we long thought it did. Hence, the crit cap reported in the latest versions of the sheets is most likely correct - we have no evidence to indicate that it's anything else, or that it ever was anything different. So unless and until someone comes up with testing that shows otherwise, I'd assume that it's correct.
In case anyone's wondering, Aldriana's post just above is a response to a question I asked in my BIS gear post about how certain we were that the crit cap isn't reduced by 4.8% after all. After posting, I searched more thoroughly and found the thread outside this rogue forum (in Class Mechanics) where people were discussing it in more detail. Reading the most recent posts answered my own question, so I edited my post and removed the question, but apparently you had already begun your reply, Aldriana. Oops. Sorry for making your post seem to be answering nothing.
It's a great summary of the issue, and I'll be pointing people at it.
While cross-checking some results between the simsheet and your combat spreadsheet, I found an issue, which I think might be working as intended, but it has a material effect on dps. The reason I think it might be working as intended is because (a) it is a pretty obvious simplification compared to what it should be in reality and (b) because once I started to think of a proper fix without a simulation, the formula got complex pretty fast.
So the issue is in calculation of uptimes of proc trinkets. The formula that you use is the same as the one derived in the Proc Mechanics thread and it works approximately well for fights of infinite duration. For those wondering, the formula of the uptime of an ICD trinket for an infinite fight is U=D/[C+1/(vp)], where D is buff duration, C is ICD, v is number of "valid" attacks per second, and p is the proc chance. So for instance, for WFS heroic, the uptime in close-to-BiS gear is calculated as 0.325. However in a 300 second fight in BiS gear, due to the timing of procs relative to the fight length and the high proc rate, the proc goes up exactly 7 times, and it does not get clipped. 0.325 uptime assumes the last proc on average gets clipped by 7.5 seconds. So in reality for a fight of length 300, WFS will be up 35% of the time (7*15/300), and the reason it is different from the theoretical 0.325 is because of finite fight length. As you change the length of the fight, the uptime will start changing again, and it is not reflected in your uptime length formula. The fight duration is reflected in uptimes of other buffs, such as haste pot, hysteria, heroism etc, but not that of the trinkets, which seems like a considerable omission.
The effect of this discrepancy for WFS is 31 EP, which in close-to-BIS gear was 38 dps. The same probably applies to every proc trinket. For instance heroic DBW is calculated to have 9.46% uptime of each kind of proc. However, given the 300 sec bounds, there are exactly 3 procs in each 300 second period and therefore the uptime of each stat is 10%. Again this uptime would vary as you change the fight length. So for a 300 second fight, heroic DBW is missing about 30 dps compared to what is calculated in the sheet based on infinite-length-fight uptimes.
Unrelated to this issue, while digging through proc formulas, I noticed that you model WFS to proc from white attacks, yellow hits, SS procs, and TAJ procs. My research shows that it has the same proc behavior as DMC:G and DV in that it also procs from poisons, DP ticks and rupture ticks. I have not independently checked that it can proc from instant poison and DP applications, but I know it can proc from rupture and DP ticks as seen here: DP ticks Rupture ticks
That is not a big change, but it is necessary to make sure WFS 7th proc is impossible to clip in a 300 second fight.
I asked Aldriana about the proc calculations a long time ago and I presume he won't mind me quoting a PM from him on the subject.
Originally Posted by Aldriana
We don't care what the actual answer looks like for a fight that's always exactly 300 seconds. That's an unrealistic encounter - we never attack something for exactly 5 minutes and then stop. More typically, we attack until the mob dies, which it does after around 5 minutes - give or take. And that's the key point.
See, the thing is, when you impose an exact duration, you get some wierd artifacts in trinket valuation - a fight that's exactly 90 seconds long (and thus ends shortly before the 3rd proc goes off) is going to give very very different trinket valuations than one that lasts 105 seconds (and gets in that third proc). The thing is: that's not a meaningful distinction. People don't really want to see the highly discontinuous curve of actual proc uptime for a fight exactly n seconds in length; they want to see the proc uptime for a fight that lasts around that length - but perhaps 10% more or less. And it's an added bonus if there are no wierd discontinuities in terms of fight length - people's expectation is that a 300 second fight and a 305 second fight will generally exhibit similar properties.
Thus, I intentionally used a model that, while inaccurate for a fight exactly 300 seconds long, is reasonably accurate for the real-world situation of fights that are "about" that length, and has the added value of being relatively continuous and well behaved.
Edit: There are definitely times when this particular issue frustrates me--fights like hodir or mimiron hm really did have set timers--but it's been less relevant in ICC. There are things that spreadsheets do well, and things that simulations do well, and this is an example of why people should do both If all you really want to do is compare your two sheets, just set the fight length to something equal to a multiple of the ICD, the assumptions won't be so different in practice there.
This is more or less what I was referring to when I said:
Originally Posted by Aldriana
For instance, on tank and spanks WFS will almost always outperform the spreadsheet
It's actually at some level a conscious design decision - the point is that it's rare to know that a fight will be *exactly* 300 seconds. In practice, what we tend to know is that a fight is "about" that length - one week it might be 4:40, and the next 5:15, but it's usually in the neighborhood of 5 minutes. And because there is some uncertainty in length, we may or may not get the last proc in. Hence, rather than tracking the exact number of procs for the exact fight duration specified, I use the average uptime to account for the fact that there is this uncertainty.
Moreover, one is rarely optimizing for one specific fight in an instance. It's true, if you want to optimize for, say, Festergut, you might be able to say that the fight is reliably between 3:45 and 4:00 for your guild... but that's just one fight. More typically one wants to optimize for all fights in a tier. So knowing what's best in a fight that's *exactly* 3:50 long - even assuming you know it's exactly 3:50 long - is not as generally useful as knowing what's best in a fight "about" 4 minutes long.
Like, the problem with the approach of counting 7 procs for a 300 second fight is that it's going to radically change your gear ordering based on exactly what fight length you enter - particularly for shorter fights. For instance, having a 130 second fight strongly favor Herkuml and Tiny Abom over WFS while a 150 second fight demonstrates the reverse... is wholly accurate in terms of optimizing a fight of exactly that length, but isn't really useful at all in terms of figuring out what pieces of gear are, in general, good - which is, after all, the point of a spreadsheet.
That said, I have been toying with the notion of implementing a mode where it would do this sort of exact fight-length calculations so as to allow specific optimization for a given fight... but in terms of day-to-day spreadsheet use, I think the current behavior is actually more useful.
Basically, I've never - nor will I ever - claimed that this (or any) spreadsheet will be an exact model of any given fight - simply because that was never the intent, as I'd argue that from a practical perspective that analysis isn't as useful. What I'm attempting to do is assess what items are generally good and bad for the sorts of fights that one generally winds up doing, and generally provide information about the sorts of tradeoffs one makes by making certain gearing decisions. It's up to the user to decide if the general spreadsheet assumptions apply to their exact circumstance... or not.
Basically, when one is a new spreadsheet user, it's easy to sort of assume that the spreadsheet is all-knowing and one should always simply do what generates the biggest number. And really, for a new spreadsheet user, that's not even really a bad thing - slavishly following the spreadsheet will probably yield better results than trying to guess for yourself what stats and items are good and bad. So I by no means want to discourage this practice in general. But I will say that a more sophisticated user can (and should) look beyond the spreadsheet and figure out for themselves when certain specs, items, enchants, and so forth will over- or under-perform the spreadsheet, and thus make more informed decisions than they could by strictly following the spreadsheet. The goal of the spreadsheet is not to say that X is always better because it generates a slightly larger number; the goal of the spreadsheet is to allow you to assess how large the difference between X and Y is, so you can make an informed decision about whether the practical realities of the fight are sufficient to let Y pull ahead or not.
So, in the case of trinkets: no, the spreadsheet is not going to accurately model your 300-seconds-exactly Festergut kill. What it does allow you to do is to say "well, because the fight is exactly 300 seconds, I know I'll get an extra 2/3 of a proc beyond what the spreadsheet indicates, which will make WFS a little better in this specific circumstances; and since it was only 7 DPS behind Herkuml to start with, that means it's going to be a better option in this circumstance". But note that if I took the opposite approach and had the spreadsheet model *exactly* the 300 second fight and let people try to figure out for themselves what would happen in a fight that was "about" that length - it would be much harder.
So, yes, the spreadsheet could be more accurate at modeling a fight with an exact duration (or interruption pattern, or whatever else). I just think doing so would make it less useful overall, which is among the many reasons I haven't done it.
The thing that confused me was that the fight length was definitely used in calculation of other uptimes. Say heroism for instance, its effect gets larger, the shorter the fight is. The formula properly calculates the ratio of 40 to fight length and uses that to determine the uptime of heroism and therefore it's average haste multiplier. That in turn has an effect on the EP of stats and definitely influences the choice of gear. Your optimal gear list is likely to be different based on the spreadsheet if you plug in 60 second fight versus a 300 second fight. Hysteria works the same way, and I am sure the uptime of hysteria will change the balance between physical attacks and poisons, which will be reflected in gear values.
From philosophical point of view, neither approach to fight length gives you full information. Not modeling it in the trinket procs gives you information about gear that does not represent a single fight in the game, so that limits the usefulness of such information. Modeling the fight length perfectly, gives people accurate information about one specific fight length, that was selected in the spreadsheet, and if their fight length changes, the recommendation of the spreadsheet is off. I like the second approach personally because a) there are fights of fixed length, b) there are now trinkets with longer cooldowns, such as DBW, and you can be pretty certain that if the fight is within a minute of selected time, the trinket's uptime will be calculated properly, c) I can look up how long it took the raid to finish a fight like Saurfang and set that as a reasonable goal. Due to the nature of the modeling, in many cases changing the fight length by 10-20 seconds will not change the uptime of a trinket, so it is not as sensitive as it seems to be.
Regardless, for my own purposes, I know when people come to me and say this and that proc trinket value is different in the simsheet compared to Aldriana's spreadsheet, this could be one of the reasons.
The thing about Heroism is that it behaves continuously with fight duration - that is, Heroism in a fight that's 300 seconds long does "about" the same thing as heroism in a 310 second fight. If you look at the uptime of heroism vs fight length, it's a smooth, continuously decreasing curve. The same is not true of trinkets, for which uptime will jump up and down depending on exactly what the duration of the fight is. Yes, you'll get different gear recommendations for a 120 second fight than for a 300 second fight - but you'll get fairly similar recommendations for 120 second fight and a 130 second fight, and that in turn will give similar results to a 140 second fight. In short: small changes in fight duration will generate small changes in DPS and stat weights; this is preferable to small changes in fight duration generating *large* differences in fight duration.
Again: this approach does mean that if you know your Festergut kill is exactly 3:53, you're not going to be able to optimize as well for that exact situation. But in practice, you don't. And even if you *do* know Festergut is *exactly* 3:53... Saurfang isn't. To say nothing of, say, Blood-Queen.
Also note that this approach allows estimation of interrupted fight segments, where your first procs needn't occur at the exact same time you start attacking the boss. It's not perfect by any means - particularly on very short segments - but it's a lot better than it would be if you forced all segements, no matter how short, to have a trinket proc.
Again: there are definitely limitations to this approach. And I've never really claimed otherwise. But I really do think having a smooth model that behaves in an intuitive fashion is worth quite a bit when assessing gear to use across a range of fights - which is the reality of what we optimize for.
There is still one thing that I'm not fully understand about crit cap in the spreadsheet.
In the spreadsheets the crit cap is calculated with 100-D-M-G. This formular is showing our effective crit cap right? Because if we fight against boss mobs our crit is reduced by 4,8%. So if we really want to equip/gem up until the crit cap against boss mobs, we have to get 4,8% more crit to achieve that right? If we just equip/gem until 100-D-M-G in tooltip we would still get 4,8% normal white hits against boss mobs in the hit table?
If thats the case, then the real crit cap on our equip should read 100-D-M-G+4,8 in the char screen tooltip? Or am I missing something?
In other words, we can equip/gem 4,8% more crit than the "MH White Crit Cap" is showing right now right?
So isn't it more intuitive for the spreadsheet to show the crit cap what we really need in tooltip to be really crit capped when fighting against boss mobs which is 4,8% higher than normal?
The key point is that the "MH Crit Rate" number reported on the front page is the amount you actually crit against a boss - that is, crit reduction is already applied to it. Hence, in order to be a useful comparison, the crit cap needs to report the cap on what your actual crit rate can be, rather than then the "tooltip" crit rate that will yield the appropriate amount of crit.
Thanks for the spreadsheet development Aldriana but one thing I'm getting is this error when changing gear.
I'm on a Macbook Pro, using OS X, when attempting to change the gear so it calculates different values it in fact will not and pops up a window saying this:
Microsoft Excel cannot calculate a formula.
The is a circular reference in an open workbook, but the references that caused it cannot be listed for you. Try editing the last formula you entered or removing it with the Undo commant (Edit Menu).
Dunno what to do about this :\
You will need to use OpenOffice to open this spreadsheet on any version of OS X. This is due to MS Excel for Mac not being able to resource a number of Windows-specific (e.g. proprietary) files. The download is free, and the software is stable, so save your money on a pack of office.
Sorry for my bad english and if I'm too stupid to find a post concerning this error or whatever.
I have a problem concerning the current Mutilate-Sheet. When I remove all Buffs (setting all modifiers to "no") I miss 5% crit compared to the value ingame. I don't know if I just miss something or if there is an error in one/some of my Items. I couldn't check further because there is no option to wear no item on different slots.
First I thougt this would be the 5% forced Hit but unbuffed I'm way below the Critcap. Any Ideas what's happening?
Sheet includes the 4.8% crit reduction that you experience against boss-level mobs, hence should report 4.8% lower than your tooltip crit with all buffs off.
An interesting thing I've found is as mutilate, with non heroic BiS gear. It seems going with 4 set bonus and speccing in to full relentless and opportunity is an increase over the LR build. Even more so with Heartpierce,
The EP values are close like they always have been, but I don't understand why they go in such opposite directions with the spec change. Heartpierce is valued higher with 51/13/7, but even without that change the DPS difference is still there.
Last edited by Seditions : 02/08/10 at 2:10 PM.
Reason: clarification
An interesting thing I've found is as mutilate, with non heroic BiS gear. It seems going with 4 set bonus and speccing in to full relentless and opportunity is an increase over the LR build. Even more so with Heartpierce,
I noticed the same thing. 4p t10 = RS > LR for me in the App
In the spreadsheet LR is just slightly ahead of RS, but not enough to warrant the worse "Putricide-Fight" damage for me.
You will need to use OpenOffice to open this spreadsheet on any version of OS X. This is due to MS Excel for Mac not being able to resource a number of Windows-specific (e.g. proprietary) files. The download is free, and the software is stable, so save your money on a pack of office.
This isn't 100% completely true at least in regards to Aldriana's spreadsheets. I've been using them for quite some time under Mac OSX (10.6.2) with Excel 2008 for Mac 12.2.3 (091001), and have not experienced any problems. Though I do agree, Open Office is a great alternative for people having troubles in general.
A quick question about the spreadsheets and my apologies if this has been answered before. I have my current gear etc plugged in and it says under 'Ranged' in suggested upgrades that Shrapnel Star is 30.86. When I actually plug Shrapnel Star in place of Crimson Star my estimated DPS goes up 98 points. I thought the 30.86 in suggested upgrades was the DPS increase. Can someone let me know what it is I am missing please.
The values are in EP - that is, the units are not DPS, it's the amount of damage gained by 1 AP. In current content, however, 1 AP usually gives you "about" 1 DPS, so they'll run pretty close.
What you're seeing is the simple fact that your current ranged weapon is putting you way over cap on Expertise, so a bunch of it's itemization is being wasted. Hence, swapping to another ranged weapon that doesn't have expertise is going to be a significantly larger upgrade than the spreadsheet would predict.
I've tried searching the thread but can't seem to find a reply. For Hyperspeed Accelerators, what would be the optimal way to use it? I remember it being mentioned that Engineering tinkers/bombs should be used on CD.
Would it be best then to bind Hyperspeed to Sinister Strike? Or time it with KS (1.25m CD) whilst Hyperspeed is on 1m? I'm torn whether stacking Hyperspeed with other procs/trinkets or everytime I refresh SND would be worth the downtime, rather than just using it everytime it's up regardless of procs/CDs.
02/05/10, 7:29 PM
If all you really want to do is compare your two sheets, just set the fight length to something equal to a multiple of the ICD, the assumptions won't be so different in practice there.
