Warrior DPS Calculation Spreadsheet - Page 19

landsoul · 03/25/09, 7:21 PM

Originally Posted by Machinator

Would it be right to assume the sheet will not be 'officially' updated until after 3.1 hits? Is there any progress on execute models? I was thinking of trying to add a basic one to see how Imp. Execute will be in 3.1.

It will not be officially updated until the PTR's stop getting updates.

DarkS · 03/26/09, 5:37 AM

I modified on the sheet almost all current PTR changes but still miss the -10% damage Aura from TG. Someone could give me some tips about how should I modify it? My current concern is how affects Rage Generation from White Damage...

Thanks

landsoul · 03/26/09, 6:45 AM

*(.9) at the end of DIM cell

Rallik · 03/26/09, 12:32 PM

Originally Posted by landsoul

That's great and all, but this is just repeating what others have said without any real backing besides just saying it or "I tried it out in a sim." How close were switching 4 and 5, at what times would using SD be better than MS. At what times would hitting MS be better than SD? According to the scenarios I used and comparing their opportunity costs, the results were not clear. Is there any way you can show some results that you did see Rallik?

With a gear setup I had including deep wounds chance and crits, a 30 rage execute did 6802, and a mortal strength had weight 5517.

If swing time was 2.58 seconds, then inside of a GCD, there would be a 1-((1-0.9*1.5/2.58)*(1-.09))=13.76% chance to overwrite a SD. Opportunity cost = chance * weight = 936. The opportunity intercept time of the mortal strike was then .848 seconds. So if you waited 1.5-.848 = .652 seconds until MS was about to come up and hit execute instead, you would be breaking even. Would the be an acceptable concept to use?

Also, what about bladestorm? should it be glyphed? should it be even used? In practice (I can calculate it NP) does it do about the same DPS as a regular rotation or less, more? Is it only used when everything is used and low on rage?

Just swapping SD before MS was only a 6 DPS loss. I could make the GCD decisions arbitrarily complex, like calculating the damage of MS/SD, determining the probability of SD proccing in the next x seconds, and deciding which ability should be prioritized based on that and MS's cooldown time, but it seems overkill when no human can make decisions like that in the necessary time frame(or at least I can't), so I just keep it simple. If an ability is usable and it has a higher priority, it gets used. At only a 6 DPS difference, it could shift either way depending on gear very easily, so for spreadsheeting purposes, I'd imagine you'd have to use something like the method you described to determine a clear cut winner for a closed formula.

Disabling BS was a 119 DPS drop, and yeah it showed the highest DPS being used as a replacement for slam when: Rend is up and MS/OP/SD are inactive. Attempting to add a rage limitation increased its DPS very slightly(less than 5) with rage <= ~75-85. Any lower reduced its DPS by increasing the delay between uses too much. In practice it's likely best to over-HS a tad just before using it, which isn't simulated.

BS glyph was only a 10 DPS gain, pretty terrible. In comparison rend glyph was 117, MS glyph was 63, and execution was 54. Surprisingly, HS glyph was ~58. All of these numbers are with my gear and the rotation I mentioned, so their values could easily shuffle around a bit. I'd say Rend is the only absolutely necessary glyph, with any 2 of MS/execution/HS as acceptable backups. I'd likely go with MS/Execution still just because the simulator does HS much more intelligently and reliably than I do and tends to show higher usage rates than I see in-game.

Shha · 03/26/09, 1:39 PM

Originally Posted by landsoul

That's great and all, but this is just repeating what others have said without any real backing besides just saying it or "I tried it out in a sim." How close were switching 4 and 5, at what times would using SD be better than MS. At what times would hitting MS be better than SD? According to the scenarios I used and comparing their opportunity costs, the results were not clear. Is there any way you can show some results that you did see Rallik?

With a gear setup I had including deep wounds chance and crits, a 30 rage execute did 6802, and a mortal strength had weight 5517.

If swing time was 2.58 seconds, then inside of a GCD, there would be a 1-((1-0.9*1.5/2.58)*(1-.09))=13.76% chance to overwrite a SD. Opportunity cost = chance * weight = 936. The opportunity intercept time of the mortal strike was then .848 seconds. So if you waited 1.5-.848 = .652 seconds until MS was about to come up and hit execute instead, you would be breaking even. Would the be an acceptable concept to use?

Also, what about bladestorm? should it be glyphed? should it be even used? In practice (I can calculate it NP) does it do about the same DPS as a regular rotation or less, more? Is it only used when everything is used and low on rage?

As far as SD vs MS. There is plenty of time when your GCD comes up with MS "fully there". Thats when you prioritize MS over SD, and basically only there (obviously save for "waiting 0.01 sec for MS" and other theoretical stuff).

Your model is a bit wrong in this Landsoul - because arms doesnt operate in vacuum so to say. If you wait for a skill, you waste GCD. That GCD could be used for something.

Proper model wouldnt use "execute strength" and "MS strenght". Proper modelling would require:

1) Model "Slam strength" - as the "always up, but worst" "filler".
2) Model "Modified Slam Strength" = "Slam Strenght" - white dps loss from slamming (0.5*white dps)
3) Use "Modified execute str" = "Execute Strength"- "Modified Slam Strength" - to model the gained dps from executing over filling the gcd with Slam
4) Same for MS

Now the opportunity cost of using MS over SD = 13.76% * MODIFIED execute str = 13.76%*Execute Str-13.76%*Modified Slam STR

Cost of WAITING .652 sec for MS is .652/1.5*Modified Slam STR - which is way higher already. Plus obvious SD overlapping procs.

Machinator · 03/26/09, 1:39 PM

How would one go about updating the sheet for the 3.1 armor debuff changes? Debuffs should be 0 now, but I am not sure where the new % reductions would go, Base Pen or Average? Are they additive or multiplicative?

Edit: Also an unedited 2.305 keeps giving 0 SEP for haste with my gear, any idea why?
Edit 2: Search didnt show the iteration fix, not an issue anymore.

Madamedub · 03/26/09, 2:03 PM

Might want to move your files. As of March 30, FileFront is closing it's site down.

landsoul · 03/27/09, 1:07 AM

Anyone know a free hosting site?

Also, I figured something was messed up with that arms model because no matter what I did using slam all the time was always better. I'll try to add in slam opportunity cost to MS and Exec now and see how it works. Also, I was always implying that arms isnt in a vacuum, I'm trying to arrive at a point where waiting for a particular GCD would be better than just taking it with whatever like slam.

Khazes · 03/27/09, 1:21 AM

I'm working on the armory import code. I give absolutely no guarantees as to when I'll be finished, though.

landsoul · 03/27/09, 1:28 AM

Just give me something that I can easily work with, the talents and names have to match perfectly references with lookups or the whole sheet will #N/A or #DIV/0 out

@ machinator, PTR ArP data is still a little whacky right now so I can't really say 100%.

Khazes · 03/27/09, 2:17 AM

It should be fairly easy to integrate my changes. I'm basing everything off of the currently hosted spreadsheet and adding references as I go; I'll try to make a full changelist when I get the chance but it's not going to be small.

LittleHamster · 03/27/09, 7:37 AM

Originally Posted by landsoul

Anyone know a free hosting site?

I don't know what license you use for the spreadsheet. But if it's "free", I believe you can host it on sourceforge.net or googlecode. These are very popular for open source software. These spreadsheets are arguably software, given all the macro/VB code are embedded.

hellord · 03/27/09, 10:25 AM

I'm working on the armory import code. I give absolutely no guarantees as to when I'll be finished, though.

I'm working on a armory importer too (for another project anyway) and if you need some hints feel free to PM me.

What I can say is that the xml structure is quite simple to parse.
The item list is in

<page>
<characterInfo>
<characterTab>
<items>
<item>

The relevant parameters are slot, id, gem0Id, gem1Id, gem2Id, permanentenchant.
The only problem so far is that if you parse item-info.xml you can't retrieve the stats, but you can anyway get the name of the item if you referenced it through id or name.

<page>
<itemInfo>
<item_attr>
(
[icon] => inv_helmet_58
[id] => 40528
[level] => 213
[name] => Valorous Dreadnaught Helmet
[quality] => 4
[type] => Plate
)

The rest of the xml only provides info about who sells the item and what it does give when disenchanted.

EDIT: I just found out that to retrieve item data you have to query item-tooltip.xml and not item-info.xml

landsoul · 03/27/09, 4:42 PM

Yeah but I'm clueless on how to "query" xml or whatever using the VB language. someone would have to write some code with comments in the code saying what each script does that would help a lot.

hellord · 03/27/09, 6:31 PM

I'm not good at vba but to just get the itemlist formatted like

slot:name

I can post a php and somebody can translate it.
I'm not sure I can figure out how to do it in vb but I guess it shouldn't be much different.

Khazes · 03/27/09, 10:23 PM

Thanks for the tips. Currently I'm using the code that ebs2002 posted on page 10 of this thread and it's going fairly well; the only inconvenience is I need to create a mapping for the XML -> spreadsheet conversion. I haven't run into any issues yet, though. It just takes time

I hope to be done either tonight or tomorrow.

Rallik · 03/28/09, 2:49 PM

I've been going over a bunch of mechanics I wasn't 100% sure on on the PTR the last few days:

A) made a post here about arpen stacking: Official 3.1 changes and discussion

It was 100% accurate for every test I ran with varying combinations of arpen/bstance/mace spec/sunders/FF.

B) I did some testing of 2h wep spec and BF to see which abilities they each were able to affect.

Ability Unmodified After2hSpec AfterBF
Rend     1253       1253         1303
DW        419        445          444
MS        309        328          340
OP         88         93           96
Exe      2862       3035         3155
WW         85         91           93
Slam      229        243          252

Both 2hspec and BF affect MS/OP/Execute/WW/Slam which is to be expected. But, 2hSpec does not affect Rend, and BF does not affect DW.

C) I also double checked the commonly used rage generation formula and found it to be very inaccurate. After about 40 test hits with one weapon, the linear regression equation for the data came out to be r = .00848184d + 5.709228. If you solve (15/4c) = .00848184 for c per the rage formula you get c = 442.12, waaaay higher than 320.62. I did not have flurry, but my hasted swing speed was 3.18, which plugging into (fs/2) since I was only using normal hits gives an expected (3.5*3.18)/2 = 5.565, or pretty close to what it should be considering the small sample size.

Machinator · 03/28/09, 5:29 PM

Doesn't the rage formula use base, not hasted swing speed? Otherwise haste wouldn't give more rage/time. Been a while since I tested that though so I could be wrong.

Rallik · 03/28/09, 5:35 PM

Originally Posted by Machinator

Doesn't the rage formula use base, not hasted swing speed? Otherwise haste wouldn't give more rage/time. Been a while since I tested that though so I could be wrong.

I assumed so too, but that would put the expected y-intercept much higher(at 6.3), which is way too far off to be true. I intend to do some further testing with different weapon speeds/haste/no haste.

Khazes · 03/28/09, 7:34 PM

Here's a very rough version of the current spreadsheet w/ armory import enabled:

Free File Hosting Made Simple - MediaFire

I'm sure it's got some bugs, so please let me know if anything doesn't work for you.

Note: Currently the only enchants/items supported are what is currently available in the spreadsheet. That means that any resilience/pvp gear will probably say that it couldn't be found.

landsoul · 03/28/09, 10:28 PM

Wow it works beautiful thanks! I'll have a lot of fun putting this into the 3.1 sheet thanks a lot!

Khazes · 03/29/09, 2:51 AM

Just glad I could help. The base code was what ebs2002 posted a while back; I just modified it to match the sheet. It does a talent and gear look up (I'm not sure if you can look up glyphs?) at the moment, although the talent import will probably have to be updated in 3.1. So far the only issues I've found is when someone is using gems or enchants that aren't currently supported in the spreadsheet. I didn't want to change the sheet too much (like adding more gems or enchants), but if those other gems/enchants are ever added then it should be pretty easy to keep up to date. It took longer to add the named ranges than it took to update the code, actually.

I don't intend to make any further updates (at the moment), so please let me know if you have any questions about what I did.

Pogues · 03/29/09, 4:49 PM

Khazes amazing job! That is awesome and will really help when Landsoul puts out new versions, thank you. I did find one problem, when importing my gear it said that The Runeblade of Demonstrable Power was on a weapon choice, but it is the drop down menu. Not sure why it couldn't find it.

Khazes · 03/29/09, 8:39 PM

Originally Posted by Pogues

Khazes amazing job! That is awesome and will really help when Landsoul puts out new versions, thank you. I did find one problem, when importing my gear it said that The Runeblade of Demonstrable Power was on a weapon choice, but it is the drop down menu. Not sure why it couldn't find it.

There's a typo in the spreadsheet that causes this; in the dropdown it's listed as "Runeblade of Demostrable Power" (it's missing the 'n' in Demonstrable). I didn't spend the time looking through the gear lists for typos :P

DarthGreg · 03/30/09, 2:43 PM

Could someone whose updated their sheet to reflect at least some of the ArP changes enlighten us on how they did so?