AngelM

Hi!

I missed some Fist-Weapons(Spread 2.4.0.5.a):

1. Fury
2. Merciless Gladiator's Right Ripper
3. Merciless Gladiator's Left Ripper

 

Professor Hurt

For another example where mutilate results in greater dps than combat swords, you can import my profile (mainly T5 gear, some T6 level stuff), switch to Mutilate, and Drop in equivalent daggers (Fang of Vashj and Tracker's Blade). I saw an increase of 35 dps unbuffed, 12 dps buffed. This made me very happy to see, even if it isn't modeled exactly. It gives me hope to top a meter or two as Mutilate, which I plan to spec soon after 2.4 hits.

 

Left

I've just spent some time going over the Mutilate cycle calculations in the DPS spreadsheet. They strike me as really... bizarre. The big problem is that they fractionalize Mutilates, combo point usage, and finishers, as well as assuming a continuous damage model for FW.

Honestly, it is quite difficult to follow all the calculations, but the algorithm that seems to be followed is this:

1. Use SnD with some set number of combo points
2. As long as SnD remains up, spam another finisher at some set number of combo points
3. Don't refresh SnD until it drops, but...
4. ...Do refresh the other finisher as soon as it becomes available again

What this means is that 5S/3R, for example, means "build to 5 CP and use SnD, then build repeatedly to 3 CP and keep using Rupture until SnD runs out", which is NOT the same thing as building to 5 CP, SnD, building to 3 CP, Rupture, rinse, repeat.

The total DPS is then calculated as SnD DPS over the "cycle" time plus Rupture (or Eviscerate) DPS prorated by the number of them that you can get off in the time you have remaining. I spent a long time looking at the prorating scheme for Rupture, and I don't think I understand it properly. The gist of it, though, appears to be that it first uses an energy model to predict how many seconds it will take to execute a given finisher (Evis or Rupture) - including the time needed for the instants (Mutilates) to build the necessary combo points - and then it calculates how many fractional Ruptures or Eviscerates you get per the time you have remaining after using time to build SnD. It has some mumbo-jumbo with "full" vs. "partial" Ruptures that I am not completely understanding, but with Eviscerates it just says you get "1.56" or whatever 4-point Eviscerates in your alloted time. That, of course, makes no sense. You can't get 1.56 4-point Eviscerates... you can get a 4-point and maybe a 2-point, but the damage calcs change a LOT when you do that versus just prorating the 4-pointer at "1.56". The prorating scheme in place now looks pretty fishy to me.

In other words, I just don't think the approach taken now accurately reflects the way you would actually approach playing a Mutilate rogue, and is therefore pretty inaccurate. It is kind of a nifty approach in concept, I just don't think it works.

It seems like it would make much more sense to use a state model with specific rules. IE, you start at 0 combo points, then get 2 or 3 depending on crits. You then take a different path depending on strategy, and come up with a "cycle" based on your path. The next time through, your path (and therefore "cycle") might be different. You can then compute the theoretical DPS of each path (some of which will be better than others!) and the percentage likelihood that you follow any particular path. Putting the paths into a table with the likelihood that you took each one should allow you to compute an average DPS for using a given ruleset for finishers. The coding for this would be completely different, but maybe I can come up with a way to do it. (If I have time! It took me all of my lunch break just to START to look into the modeling currently in place.)

(Of course, a simulator makes a LOT more sense here, but a simulator would also have to include all the functionality of all the other parts of the spreadsheet that I don't want to code in. Bleh.)

 

Kreebog

Using the 2.4.0.5a DPS sheet I'm seeing Mutilate out-dps Combat Swords by about 30dps using my T4ish gear

For the Combat Swords model I left everything the same and substituted in Gladiator Slicer MH and Akilizon's Talonblade OH, both with mongoose. I actually have the Talonblade in my bank and I've been working on honor for the slicer, but after seeing this I might just spend that honor on the gladiator chest and badge up to the new daggers as soon as I can.

Dunno... I'm really torn here. I love mutilate but my guild needs all the DPS it can get as we move up into SSC/TK. I have to decide if I'm badging up to the new mutilate daggers or switching over to swords before I blow the 18k honor and (even worse) 20 EOS tokens on the slicer lol.

Has anybody been able to do some actual testing on PTR?

 

saedo

Well reading the last page of posts, or even the one just before you posted (24 mins before you did) would give you some answers. Gist of it, it's not that accurately modelled.

 

Left

Although I'd like to point out that just because it isn't accurately modeled doesn't mean it's automatically going to be worse DPS than currently predicted. The current model seems to be following a model where Rupture can overwrite itself (which is a really inefficient way of doing things), as well as using/modeling SnD in such a way that Improved SnD causes a DPS decrease (which shouldn't ever happen). If/when a better model goes into place, you might even see Mutilate move up a notch. Unfortunately, it's almost impossible to tell without some very detailed analysis.

 

Kreebog

Yep, I didn't see Left's post until after I posted... the rupture damage thing does sound like a problem.

Anyway, I'm still hoping to hear from a Mutilate rogue who's been on PTR. Wish I had time to try it out myself


Edit: no 2.4 today - looks like I have at least one more week to think about it

 

Katoneo

I have one problem. When i'm trying to change talent build or load my armory profile i'm getting this msg


I'm using Microsoft office word 07.

 

Dontmindme

First the error...looks like you were sent to the debugger. What I don't see is why. Did you get an error before it sent you there, and what was it?

-------------------------------------

Ok, Mutilate. Truth be told, I have not extensively looked at the Mutilate cycles. Looked at it passively. Looked for obvious issues. Have a general understanding but didn't trace it that far. The DPS supposedly (with the Seal Fate fix) was coming out near expected but given it was not a top tier spec at the time, I felt my time was better spent elsewhere on the spreadsheet.

That said, from my reasonably extensive review of the normal cycles, the fractionalized Mutilates makes sense. What it is doing is calculating the average number of Mutilates it will take to hit a minimum of X combo points. Because of Seal Fate this can be a fractional number because of the extra combo points. Fractional Combo Points may make sense as well, if it is calculating how many times a 3s/5r is actually a 4s/5r because of an extra combo point.

Now, the fractional stacked Finishers, assuming that's what is happening, doesn't make sense at all, especially with Rupture. I may have to allocate some time to look at that more closely now that Mutilate stands a chance to be competitive.

 

Cyn

I'm seeing the same thing with my gear (T6 best of slot except MH glaive) vs mutilate with just changing weapons to shard/boundless. Mutilate does a few dps better than combat sword. The same with max Sunwell gear. It seems mutilate is beating swords at most gear levels now unless you have double warglaive.

In comparing to 2404 it appears to actually be that combat sword dps dropped by 150 or so, and mutilate dps was almost the same, i don't understand why, the changelog seems to suggest if anything it should have gone up.

 

Lastembrace

I'm fairly sure this is because you don't have the MSXML installed, download it here and then give it a try. Make sure to close excel while you're installing it.

http://rogue.raidcal.com/msxml6.msi

hope this solves your problems.

 

Katoneo

Thank you very much. It helped me.

 

Left

I'm currently doing the ground work for a new method of Mutilate modelling: using an X+ cycle pair (eg 3+s/4+r) or trio (eg 4+s/4+r/3+e). (X+ means using a finisher when you have at least X number of combo points.) The basic idea is instead of fractionalizing mutilates needed, modeling a fixed cycle time, and fractionalizing finishers used, you have a variable cycle time based on the probabilities of hitting various "states" or scenarios during your cycle. The average damage dealt by each outcome is analyzed, and then the probabilities of each outcome are used to produce a weighted average of all outcomes.

It's going to take a lot of work to put it together, and I'm nowhere near implementation yet. What I do have, though, is a decent state/path model for Mutilate cycles 1+, 2+, 3+, 4+, and 5. There are a finite number of states that can occur for each cycle, and each has a finite number of paths to get there. Each end state is defined in terms of (a) the number of combo points it yielded, and (b) the number of Mutilate attacks used to get to it. Between these two things and talent selection, we can find the average time of a cycle that arrived at that state.

To find the probability of being at any given end state, you can add up the probabilities of getting to that state via every path that leads there. At the end, you can check yourself by ensuring that the probabilities of being at your states add up to a total probability of one. I have done exactly that for each of the options (1+ to 5) described above, creating both a state/path diagram and a spreadsheet which calculates the probabilities based on the state/path diagram. I'd like feedback on what I've done so far before I go into the next step.

First, here are my state/path diagrams:






Second, I put the probability calculator sheet into Google Documents: Mutilate Cycle Probability Calculator

All of what I have done so far could be easily adapted into the current spreadsheet.

The next steps would be to start computing damage contribution of finishers used for the various end states and then averaging. The next step after that is to figure out a way to adequately crop time-based finishers that could be clipped, such as SnD and Rupture. And after that, the next thing is combining several of these state analyses back-to-back to produce a complete cycle model.

However, for the moment, does what I've done so far look useful?

EDIT: I should mention that everything I have done so far assumes 5/5 Seal Fate. For lesser ranks of Seal Fate, the probability calculations get more complicated, but it would still be doable. I guess at some point I'll have to take SF into account... bleh.

 

Donoma

How would you account for the times at which one has to bank energy during Mutilate cycles?

 

• Aldriana

At the risk on causing impressive bloat in complexity: don't forget to include T4 4/5 in your state/path diagrams.

The Seal Fate/crit thing is actually pretty manageable; simply create + store a parameter of "the chance the SF goes off" which has value of .2 * SF Ranks * (1-(1-MutilateCritRate)^2); then Mutilates give 3 CPs with this probability, and 2 CPs otherwise; it doesn't actually change the state diagrams at all (and if it sounds like this is a problem I've thought about before - guess what, it is. I started implementing something along these lines a while back, and basically gave it up once the amount of casework involved became appallingly high).

 

Left

@Donoma, that part comes later.

Let me elaborate: the only thing banking really does for you in a Mutilate cycle is determine how many attacks you can get into the Find Weakness window. Eventually, the model should account for that automatically by assuming a certain amount of banking (and thus calculating which of your attacks benefit from FW). At this point, though, I am mostly worried about predicting the various outcomes of your "cycle" and averaging them properly.

@Ald.

Ewww. Although you raise a good point; I'd have to account for that if they were using 4pT4.

As for the state diagram w/ SF thing, yeah, it doesn't change. However, the probability math gets more complicated, as you pointed out. One thing could change it, though: On a double-crit Mutilate, SF may have the chance to go off twice. Does WoW calculate SF chance one time per attack, or once per separate hand hit? I would guess once per attack, because you can only get 1 extra combo point. IE, either hand Mutilate crit, therefore we move on to check for SF. However, I don't know for sure... possibly because nobody runs around with anything other than 0 or 5 points in SF.

 

Donoma

Banking allows you to avoid certain outcomes on a future mini-cycle, especially the "oh crap" outcome of no Relentless Strikes or Ruthlessness proc + a non-crit Mutilate + not enough time to get another Mutilate in before FW and/or EW drop. Given the enormity of the project, it would probably be wise to consider banking from the beginning so you don't have to deal with avoidable outcomes. I don't envy you; I'm just trying to be helpful.

[EW = Exploit Weakness (Ashtongue Talisman of Lethality buff)]

 

Dorvan

You might also want a loop transition for each state to represent the case where Mutilate is dodged (Those transitions could also be to states with the same # of combo points but one more instant, depending on the purpose of the num instants stat). I suppose that depends on exactly how much you want energy integrated into your state model.

 

Dontmindme

It looks potentially workable, the devil will be in the details. The question is how one handles the cycles vs. the SnD time and how many branches one might have to consider off of that. Each decision branch effectively doubles the paths, but it might be possible to put rules in place to limit the options. At some point, if the PTR's come back up, I want to do some testing just to make sure I'm familiar with all the mechanics of Mutilate.

 

Left

Yes, I agree with you. (Although, EW... what is that acronym for?) However, basically what I was saying is that modeling FW comes at a later stage of the game than where I am now.

I thought about this, but in the end a dodged Mutilate can be effectively reapplied moments later. Instead of adding more paths to the diagram (which could quickly get unmanageable with loopbacks), I chose to use the strategy of the person who made the original Mutilate cycles and just increase Mutilate's effective energy usage to account for dodges. This puts Mutilates, on average, a bit further apart and causes the cycles to lengthen, on average, very slightly. Take a look at the linked spreadsheet if you want the math for that...

Indeed, the details are the daunting part. It's not hard to come up with an average damage contribution from Rupture or an average uptime from SnD. The issue is, I think, going to be properly clipping SnD when it goes over and properly accounting for the times SnD drops, as well as accounting for Rupture clipping if I have to do that.

Today at lunch I'm going to try to modify the state/path diagrams and the spreadsheet I have thus far for 4pT4. If I still have time I may start in on the next bit.

 

Myrx

EW = Exploit Weakness I believe. The Ashtongue trinket proc.

 

Zaazel

Expose Weakness is a Talent in the Assassination Tree, boosting Specials for some seconds after using a finisher.

 

Myrx

No, that's Find Weakness which was abbreviated as FW. EW is Exploit Weakness, and was not understood by Left what EW was referring to.

 

Left

No, that would be Find Weakness. Expose Weakness is a Hunter talent.

Apparently Expose Weakness is also the name of the Ashtongue proc, which I didn't know. EDIT: Actually, it's Exploit Weakness. Thanks to Myrx for pointing that out.

EDIT: Eh, Myrx beat me to it.

EDIT2: Er, Apparently I read wrong. Fixed.

 

Myrx

Meh, I feel dumb now. Exploit Weakness is the Ashtongue proc (still EW) not Expose Weakness, which as Left mentioned, is a Survival Hunter talent. EW was still referring to the trinket buff though!