DKP-Borgar

WoW-Europe.com Foren -> DPS Addon .Alpha

The Simulator is on it's way. It looks promising when all the buttons work and the customisable phase begins.

 

Moogul

I've been contemplating writing a simulator in C++, it seems more and more like it's the only way to get accurate values, due to the difficulties in modelling flurry and rage generation (amongst other things).

The only issue is where am I going to find the time to write it...

 

• landsoul

On note of combinatorics, it is easy to find the total number of combinations. I also believe someone else put up a mention of it somewhere.

My commenton this:

This is idea deals with combinations, not permutations. Permutations is where order matters, and i'm sure it doesnt make a difference if you have mongoose, executioner, and dragonspine proc, versus fragonspine proc, mongoose, and exeuctioner. Therefore we would use combinations. The numbeer of possible combinations is explained by the following series formula.

n = total number of possible procs
k = number of possible procs up at a time

combinations = n! / (k1! * (n - k1)!) + n! / (k2! * (n - k2)!) + n! / (k3! * (n - k3)!)....

for example if you had 8 possible buffs, the total number of combinations would be 1+8+28+56+70+56+28+1 = 256 possible combinations which would be a real pain in the ass and take an unreasonable amount of time to code to get minutely more accurate results. Even then your margin for error in making something so complicated would get really high. It's something a computer really would need to do unless you are really F'in smart, patient, and have a lot of time.


However for doing flurry and misses within the 1.5 GCD cycle it might be feasable, yet doubtful. Let's take a look at all the possible scenarios. You can crit, hit, glance, or miss a white attack in either MH or OH while having or not having flurry, while having or not having heroism, recklessness, or haste potion. We also need to know how many swings per each weapon we are going to get every 1.5 seconds. The problem is we want to know if we are going to get the 15/10/7 rage needed to pop execute, how much rage we will have, and convert that to how much damage that will do. Rethinking it, getting an accurate equation setup to cover all of these possibilities 100% accurately would be the hardest of all problems we can ever face. There could be hundreds of possible scenarios.

I think we best just stick to something more simple.

 

BronxMontana

while finally getting the warrior spread sheet to work properly, was very surprised at one of it's finding's, which makes me question it's validity...



example: using my stats to the tee, triple checked, i came up with a 1218.3 dps figure for my toon....yet when i changed my oh from The Talon of Ashharra to Fury (fist weapon from ZA) the figure jumps to 1230.8... i can't figure out why it would rate the fist weapon so much higher... while i do gain some ap (6), i lose .38crit, and 5 expertise....my only conclusion is that Fury is .1 faster then the Talon...but using it also would make both weapons speed inbalanced, which ive read will affect flurry negatively....


any conclusions here i might be missing? thnx for the input.....

 

Grim13

The current sheets combine cycle phase and execute phase together for an average. A likely reason the unbalanced weapons fare better is likely in the execute and flurry models. First, as to flurry, we have not modeled this theoretical "4th hit" gained by matched speed weapons. It has never been conclusively proven, so it's not included. Second, assuming you are using my sheet, slow weapons with matched speed result in horrendous execute DPS. My execute model attempts to calculate an estimated execute "cooldown" based upon the speed of your two weapons, the difference in speed of your two weapons, and your crit rate. I am working on a much more comprehensive model right now, but that will be some time coming. I am also working on decoupling execute and cycle DPS, and allowing specific execute weapons. The release with that should be out in the next week at the most, and will make situations such is this much easier to understand. Assuming all the stuff I said and assumed about your case is true.

 

Brissa

I have had the very same thoughts (albeit in Java) and I even started sketching on some class diagrams. But yes its quite a big undertaking and finding the time for it seems unlikely.

 

bomzix

Instead of trying to model a lot of factors with unknown variables isn't it easier and more accurate to simulate a combat?

For the most part yu guys seem to know all the random rolls involved and how they interact, the tricky part is doing weighting them together and over-time. A simulation instead of a model would solve this, imho.

 

Moogul

It would, but would take a lot of time to implement (which is why I haven't done it yet).

 

Barmbul

Grim, Landsoul have you guys had time to look at each others sheets and hypothesize what the reason is for the discrepancy between the value of Str and Crit on the 2 sheets? Landsoul once you dropped the ppm on MOTB it made Tsunami much closer to MOTB but still inferior by about 10dps. I question this only because if you look at endgame itemization, Crit is rather low after stacking ARP. Grim is it possible for you to get Rod of the Sun King working in your sheet? It hasn't for a long time now and for me at least would help compare the two sheets. I'm currently trying to decide on using ROTSK or Rising Tide as my OH. Longsouls sheet shows ROTSK as better than all but the Blade of infamy and S3 off hand. Thanks for all the work guys!!!!!

Armory profile

 

• landsoul

The thing about running a simulator is this, its random. What you would have to do also, is do a very large number of runs of data for each differing gearset and plot them and then make an average of best fit for that data. The accuracy of this method is directly proportional to how much data you can collect for the set you are running your data on. If it is possible to run a program that can run these tests a hundred times or so, collect the data, form the average, etc. It could work but it would be a monster to code, debug, and get it all right without pulling hair out.

Until then, we have spreadsheets.

 

Deathwing

Hey all, good to see the warrior dps spreadsheets are continuing along fine. The spreadsheets look really good, Grim and Landsoul, especially the prentation(not an engineer's strong point).

Just throwing this out there, but is a simulator really needed? As I delved further into developing my original spreadsheet, I found from an initial prospective that a simulator wouldn't be worth the amount of time. As landsoul points out, it would be a lot of work, for maybe not much benefit. How often are you going to sitting there contemplating a weapon upgrade that's really only worth a few dps?

edit: presentation*...apparently neither is spelling.

 

• landsoul

The reason tsunami is close to MOTB is the following:
Tsunami's proc is a lot of raw attack power. And of course it beats the ArP proc from MOTB by a mile regardless of how much ArP you have. The amazing passive stats that MOTB has on it is what makes it beat tsunami and its godly proc


Oh and BTW, DE Rising tide plz unless you are going to use it for RP...

I have also hear mixed proc rates on ROTSK. I dont have one personally but ive read 4.5% and ive read 6 PPM so IDK. 6 PPM is what I have in the sheet currently cause I ponder it might work like unbridled wrath.

 

Moogul

...Or you're an Orc (which some of us are, you know).

Judging by the length of this thread (and others on EJ), quite a lot ;-)

 

Grim13

I'm going to try and get RotSK modeled, along with separate execute weapons, today, and get a release pushed out the door =maybe 4 or 5 hours from now. As to the simulator discussion, dr_AllCOM3 is working on an in-game .lua based sim that was looking rather promising last we saw.


***edit***
It also just occured to me that a web-based solution could be written as a sim rather than as a spreadsheet, though I am not certain that would be a good idea.... I have a feeling that the CPU demands of such a thing could escalate rather quickly, if there was anything resembling a decent amount of traffic. Assuming that is the case, then I don't see any way to have a "sim" solution without involving downloadable content. Not that that is a bad thing, but right now, the niche I am looking at is cross-platform and out-of-game. Basically, I am working on something to keep people's minds occupied while they're at work. The funny part, I suppose, is that after reading a few html, javascript, php, xml and AJAX tutorials yesterday and today, I have a solid idea how to code the guts of the thing, but still need to learn how to make it presentable (not ugly) as a web page. What's really sad is that I haven't played WoW in over a month, but I still spend hours and hours on these projects.


Doesn't look like this update is gonna be ready in time to get out the door before this weekend. I'll get back to work on it Monday. To recap, right now we have separate execute weapons, dual wield WF tweak, and RotSK to be added. Is there anything I am forgetting? Please remind me if so, as I would like to have all currently outstanding issues resolved in the next update.

 

Mardraum

I know for myself the execute weapons part is what i'm looking towards.

Thanks for all the hard work on the spreadsheet.

 

Grayson Carlyle

Flash.

It would be far faster than javascript (math in Flash is very fast, only graphics processing isn't, since it relies on the CPU, not the GPU) or in-game (where WoW takes up enormous amounts of CPU). Extremely portable, and the size would be pretty small (the XML files for the items would probably be a larger download). I started a sim in Flash, but then my job became busier and didn't have time to work on it.

 

• landsoul

Deleted: moved to a different topic

 

Shha

Ugh, just deleted a looong post, because of some math errors I found. Im gonna redo the math, but some major points.

In both Niuta/Grim spreadsheets, Mongoose is modelled wrong, and it can lead to a big difference in final stats.
Both spreadsheets model Mongoose as a chance to gain 120 agility. In fact though, mongoose adds 2% haste too. Its also haste thats not based by ratings - just a straight 2% speed boost.

Modelling a proc that gives 4% crit and 2% haste, as just 4% crit seemed wrong, so i started to investigate. In short - there are gear/buff combos that are realistic (aka happens to me) when dual mongoose outperforms executioner/mongoose. In short with minimum possible crit buffs, and a boss with decent armor (7700 without recklessness), Mongoose is better. Even more in case of resist fight like Shahraz. Im currently both testing in game and in spreadsheets the mongoose WITH the haste included. Seems in sunwell Mongoose will catch up even more (im not sure about bosses armor, but the gear in general will LOSE some ArP affecting it negatively, and will gain haste , increasing white/heroic damage ratio to instants,and making the haste from mongoose worth more). Same way for a lot of less-then-BT geared warriors, Mongoose might prove better because of their low crit rate.

Also just rational reasoning. Mongoose for me (DST, haste pots and dragontrike giving me 15% haste on average) is worth 2.3% haste (or 36 rating). 36 haste rating using my SEP values is equal to 82%*7*36=206 ArP. That "leaves" executioner with 634 ArP vs Mongoose 132 (kings) agility. Mongoose =4% crit basically or 88 crit rating.

Again using SEP values 68%*88*7= 420 ArP. Executioner wins by 214 ArP as expected which is basically meaning its "33% better then mongoose".

However if i take away my LOTP/ret paladin/agi totem and try a boss at 7700 armor (previously was 6200) WITHOUT CoR, it seems to change.

Mongoose ArP equivalent becomes

Haste Portion - 36x7x82.25%=
Crit Portion - 88x7x81,7%=

Total is 710 to cut the Math. So Mongoose caught up a bit.

With Sunwell gear and the "mongoose friendly boss setup

Haste Portion 36x7x110.2%= 278
Crit Portion 88x7x89.2%= 549

827 ArP. Thats close enough to imagine that with some setups/gears (i might drop lower in ArP "on the way" to my perfect sunwell gear), Mongoose might come ahead.

Anyway point of the post is simply to remodel Mongoose by adding the haste component (even just straight % of 32 haste rating will be better , although its 2% not rating based). I think it wont be too hard, and in some cases might prove benefitial.

Also the math Ive done pretty much concludes any debates about double executioner for me. If you count the 2% haste, i dont see many setups that would show for dual executioner.


Additional change , although cosmetic one, would be to switch elixirs of major agility for mongooses in buffs section. Mongooses are cheaper, and give better buff, so why would anyone use agility (well save for the time when im too lazy to get a flask and i leech off fellow rogues!). 10 agility is less then 8 crit albeit very close .

 

• landsoul

Deleted: Moved to a different topic

 

WernerVonBraun

It should be possible without makros. You have to make another page, that copies everything from the calculation one and adds 1 to the desired stats. I takes up a lot of space .

 

• landsoul

The problem with doing this is when you have to go in and make ANY modifications to any part of the model infrastructure, you gotta go in and do it 8 more times in the SEP infrastructure. This includes all listed procs and their uptimes... Once I updated my sheet one step further, I would lose interest in keeping up with the SEP crap.

 

Lambach

When you're arp starts closing in on 1350 duel ex wins, definatly in 6200 bosses and is about medium on 7700 bosses. Most guilds should be able to use COR, thus making your double mongoose math mostly wrong except rare circumstances.

 

Shha

My math wont be "wrong", just different. My point wasnt to prove that mongoose is better then executioner. Point was to prove that under maybe not very common, but still realistic circumstances, mongoose MIGHT be better. That + the fact mongoose is modelled at around 2/3 of its effectiveness. I just suggested it would be worth to fix the model mongoose is using, not only for the sake of it, but possible actual applications. Some guilds in SSC dont use recklessness and their warriors dont have any or low ArP.

As for DUAL executioner , it doesnt stack. My uptime meter shows around 45% uptime for MH 30% for OH. Combined they should be at 1-0.55x0.7 = 61.5%. Therefore the extra executioner adds around 16.5% more uptime, or in other words, "works at half efficiency". The math clearly shows its not really the case of dual Executioner winning then as mongoose is at 0.75% effectiveness of executioner, and not 0.5% or around.

Using spreadsheet values of uptime at 33% for MH and 23% for OH or around that, dual executioner is at 1-0.67x0.77= 48% or 15% increase. 15% of Executioner uptime, is still lower then 23% uptime of mongoose although by a small margin.

 

Graul

Until every boss fight, and to a lesser extent trash has almost zero movement, I won't be sold on Mongoose ever being better than Potency in the off hand. For any fight where you can stand there for 100% of the fight without moving, then sure, the math supports that, but what the math doesn't show is all of the "downtime" you have where Mongoose may have just procced and you can't hit anything in any situation where movement is involved or where you are incapacitated in some way. It's also the same reason I can't see Ex/Ex ever working out either. The main reason any of the proc enchants are fine for the main hand is simply because they have a greater chance of activating while you are actually hitting something, PPM statistics or not.

I guess it just boils down to what you want to be doing with your character. Consistently good damage through the whole raid, or simply excelling on a handful of bosses while being just a little less effective on everything else.

 

Shha

Well graul way i see it potency=22str with kings.

With uptime i calculated (30%) mongoose is 40 agi with kings and around 12haste for me, for OH that is.

I honestly doubt i have THAT much downtime. IN straight numbers, 40 agi = 34 str. 12 haste = 18 str.

Mongoose provides 52 str equivalent on a straight tank and spank. Are you suggesting my effective downtime is 60% of dps time? I doubt it. Granted with the numbers from spreadsheet (22% uptime). Mongoose is only around 39 st equivalent. But even then, 20 str vs 39 seems way more then downtime would support.