Guild Wars 2

I recently wrote a paper (still not complete but you can take a look) concerning stat, dps and optimization using a guardian build as example using a continuous approach. For now it only has one example, I will be updating it in the near future. All credits of the build and skill coefficients goes to Obal and DEKeyz. Also thanks to Nike’s for his wonderful damage calculator which took me quite a while to work out the difference between his and mine.

Warning: math intensive

Here it is:

https://drive.google.com/file/d/0BzLXxOTLiwe3QVFTTjEyV1hkVE0/edit?usp=sharing

Edit: May 29 – uploaded an appendix with full proof why berserker is best.

Stars.2179:

I recently wrote a paper (still not complete but you can take a look) concerning stat, dps and optimization using a guardian build as example using a continuous approach. For now it only has one example, I will be updating it in the near future. All credits of the build and skill coefficients goes to Obal and DEKeyz. Also thanks to Nike’s for his wonderful damage calculator which took me quite a while to work out the difference between his and mine.

Warning: math intensive

Here it is:

https://drive.google.com/file/d/0BzLXxOTLiwe3eGNGQ1Jrb3RxNUk/edit?usp=sharing

Holy crap! you weren’t kidding when you said this was math intensive _

Still good info. I’ll be waiting on the expanded armor and build exposition. (I personally run the hammer deeps when I go dungeoning. I’m sure people would be interested in hammer vs gs/sw vs sc builds)

I am still working on the 4/5/0/0/5 build with split stats into vitality today. Probably will update it on Tuesday, then I’ll move to the 4/6/2/0/2 build next with more complicated traits such as Retributive armor.

Ghotistyx.6942:

Still good info. I’ll be waiting on the expanded armor and build exposition. (I personally run the hammer deeps when I go dungeoning. I’m sure people would be interested in hammer vs gs/sw vs sc builds)

The hammer vs gs/s stuff has already been done by fade some months ago. I think his table is still linked somewhere in obal’s sticky, but the TLDR is: hammer meta is approx. 5 % weaker than gs meta and scepter can overtake both of them – if all conditions necessary for smite are fulfilled, which is practically never the case.

Yeah, I’ve read through anyone who’s posted calcs. However, our friend Stars is going fairly in depth, and I’m interested in a) confirming everyone else’s math, b) possibly learning some hidden way to better optimize in some situation, and c) increase the data’s visibility.

I’m really glad that I can finally point the “armor has diminishing returns” crowd to someone else’s thread so they no longer have to take my word for it, now.

I recommend that you start exploring condition damage ASAP, as you aren’t really breaking new ground until you start to do so.

I enjoyed how through your paper is and the combination of all that information in one place is really good. Having someone go through the math and theory is alot better than just taking someone’s word for it. Thanks for all that. I wouldn’t bypass anything but the faster you can get to condition damage, the better. I’ve got my own little pocket of work on that matter but I want to see if other thinking aligns with it.

Sorry,. going to bump this.

Can the OP comment on what method was used to solve the constrained optimization problem you get for the simplified Offensive equation?

Without making a math text out of this paper, it would be valuable to mention how you did it. I’ve limited knowledge but it doesn’t appear you can use Lagrange Multipliers or Linear Programming because the equation you are solving for has a quadratic term in it; when you expand out the Simplified Offensive equation, you have a precision*ferocity term.

How do I put it! It is rather difficult. I’m trying to type up an appendix explaining it. It will take a bit. I will still need to discuss it with some of my friends to make sure I didn’t make any error. Maybe I’ll be able to append it this weekend. Kinda busy with other math problems.

Yeah I hear you. These aren’t easy problems … they are almost easier to solve with brute force; pick many points within the constrained space, solve the simplified equation for each point, find the vector that results in the highest number for equation … Repeat, picking vectors ‘correctly’ around the initial max vector and see if you can do better.

What I’m REALLY interested in is being able to maximize the following:

Effectiveness Rating (E) = aO x bD

It’s similar to your equation on page 8 but I added a and b as parameters. You have already shown the case for a = 1, b=0 (the full offensive case). I don’t think it’s a stretch to see that the a = 0, b = 1 is probably based on VIT/tou/pow gear.

What has real value to me is a case where a = b = 0.5; a situation where I tend to swap depending on the capability of the team. I suspect that’s probably a complex mixture of different types of gear. I’ve got some estimates for what I think is close but I would like to see how close I am to the theory. Yours is the most methodical approach I have seen so far.

aO x bD only works when a = 0, b = 1 or a = 1, b=0, which is the same as looking at offense and defense separately. I thought about putting a weight on each but quickly realized that since multiplication is commutative you get:

aO x bD = (ab) (O x D) which, if a, b are constants, then so is ab, then we will still just maximizing OD anyway, which then defeats the purpose of putting them there to begin with.

To make it works, one would have something like

aO + bD, in which case a and b will actually matters, but we have to ask the question why would we add the two values? would it imbue a meaning to the rating? As I explained in the paper, the multiplication will actually be more meaning full.

Note that maximizing O+D is not the same as maximizing OD, so we have a difficulty translating the meaning.

A few months ago an in game friend, EvaristeGalois (I forgot how he spelled mr. Galois’s name :P ) gave me the idea that maybe putting the weight in the exponent would do the job, i.e.,

O^a x D^b

but then after consideration I discarded the idea because it just amplified the difficulty of the problem by a thousand times (literally) (chain rule during that partial derivatives and then solving multiple equations (about 21 equations) with 9-10 variables…)

But I’m still up for taking suggestion.

Simplex optimization?

It’s been a while. With recent festival events, I’ve been enjoying the blitz rather than working on this lol. However, I’ve finally written an appendix as requested to see the full calculation / proof for berserker in 4/5/0/0/5 build.

I also have the data for split vitality stat for this same build as well as the 4/6/2/0/2 build and how to maximize if one are not too comfortable to run in full berserker and prefer to opt out a bit of stat toward vitality instead.

Will upload the next part in a week or two (I have a life and enjoy the game too :P)

Now THAT is a proof that Zerkers is highest damage output for this profession … not some slapping around golems in EOTM.

Thanks for your work.

While the condi build is interesting, I don’t think anything this sophisticated is necessary in the Guardian case. Simple change your variables for your gear/stats, see how much damage you lose for how much condition damage you gain.

What would be interesting is to see if there is a case with a high condition application profession (Necro I bet) where the move from Zerker to Rampager (or even Carrion) would be better and under what conditions. i.e. how many conditions do I need before a Rampager damage output > Zerker. I’m pretty certain that point exists because of the Ferocity damage contribution.

And how many other professions can equal or best their zerker setups with conditions …