[Shaman] Itemizing Enhancement at 70 - Page 16

Disquette · 05/05/07, 11:05 AM

Ok, I have to run, but the "big test" results are in. Parameters are found on the .xls sheet, and the results are show at the bottom of the jpg.

Quick result... for an even item budget increase (100 AP = 2.3%crit = 3.16% hit (someone please verify this!))

AP = 3.7% dps increase
Hit = 2.0% dps increase
Crit = 2.8% dps increase

Also, to make it easy for the pawn strings....

AP steps were done by 100, 2.3% crit and 3.16% hit are each 50 rating steps. So, equivalencies for damage purposes are:

3.7/100 = .037
2.0 / 50 = .04
2.8 / 50 = .056

so to normalize to Attack power, we get:
1AP = 1.08 HR = 1.51 CR.
**** VERY IMPORTANT EDIT****
I'm leaving the line above because I wrote it, but it is NOT an equivalency formula. I'm a moron for writing it that way (thanks to Azaranth for pointing this out). It should be, as other posters have written:
1AP = 1 point
1 HR = 1.08 points
1 CR = 1.51 points
**** end edit ****

This matches up pretty well to the first raiding test we did. I guess the test where I got a parameter wrong was the one for oglock, so i'll have a look over that one to see what happened.

mek · 05/05/07, 2:03 PM

Originally Posted by Disquette

Quick result... for an even item budget increase (100 AP = 2.3%crit = 3.16% hit (someone please verify this!))

2.3% crit = 50.83 crit rating
3.16% hit = 49.928 hit rating

Seems pretty close.

so to normalize to Attack power, we get:
1AP = 1.08 HR = 1.51 CR.

Did you mix something up here? You mention that 50 hit is 2% dps, and 50 crit is 2.8%dps, but here you have HR being more valuable? Or are these the "dps values" of each stat, rather than the equivalency points? Arrrrg what do these numbers mean, let me figure it out.

1% dps = 27ap = 25 hit rating = 17.9 crit rating.

Well, AP is certainly the winner, since it costs only 13.5 rating. Calculating back to item budget equivalencies...

2ap = 100% efficient
1HR = 54% efficient
1CR = 75% efficient

Aha, so what you have above are those item budget efficiency values, gotcha.
From there we can get an "ideal" enhancement itemization style, with an 8:3:2 (i think) ap/crit/hit ratio. Midnight Legguards a decent example of that, but a little light on AP still.

Now let's look at Cataclysm and see how efficiently budgeted it is (laying aside the Agi for a second...)

Cataclysm Chestplate
Binds when picked up
Chest Mail
846 Armor
+36 Strength
+28 Agility
+40 Stamina
+24 Intellect
Red Socket
Blue Socket
Yellow Socket
Socket Bonus: +4 Crit Rating
Durability 140 / 140
Classes: Shaman
Requires Level 70
Equip: Improves hit rating by 15.

72 ap (80 with kings), (28/25 x 22.1)+4 = ~29 crit rating (31.5 w/kings), 15 hit rating.
Our ideal ratio would be 77:29:19, so, pretty close!

mek · 05/05/07, 2:50 PM

I just quickly edited a version of HunterAEP (great little mod, ace2, not my work) to pick up Strength, and changed all the default values to Shaman-appropriate ones based on the above. (it counts Intellect as a dps stat still, you can get rid of that if you are a "rogue in mail") It will spit out tooltip numbers which give you a very good idea of the DPS contribution of any given item, works nicely with sockets and enchants as well.

http://members.shaw.ca/mekrith/Hunte...7-shamaned.zip

Malan · 05/05/07, 3:35 PM

How exactly do I read the AEP tooltips? Seems straightforward when it just says that one item is 450 vs 473, but some of my gear shows an entry like
T4 Helm - AEP:626 (+336)(+308)
T4 Gloves - AEP:578 (+240)
What do the other values mean?

mek · 05/05/07, 3:52 PM

First is the base item, +green is enchants, +white is gems/socketbonus, and then on the right you'll get a Total.

/haep is the slash command which shows options and all the base values for various stats.

Morelis · 05/05/07, 4:06 PM

If I'm interpreting the results correctly it appears as though crit scales at a better rate than AP. Is there a way to determine exactly at what AP level crit becomes a better use of item budget?

Disquette · 05/05/07, 4:31 PM

Hey Morelis,

I haven't looked over intra-run results very much, so I'd be interested in the data you're using which leads you to believe that crit scales better. I'd assume that crit scales worse compared to AP, especially since your flurry uptime and UR uptime is getting very little additional benefit after you get past 30% crit or so.

Also, since hit is capped at 24% (and 6% or so for yellow damage), that doesn't scale so great after a certain point as well.

All of the above leads me to believe that the only stat which will give great returns "forever" is AP.

Mek - that's a very interesting way of looking at item budgetting, as ratios, since stacking any single stat in item budget has exponentially increasing costs (as I remember it). Very cool work/intuition there :-)

Morelis · 05/05/07, 4:45 PM

Originally Posted by Disquette

Hey Morelis,

I haven't looked over intra-run results very much, so I'd be interested in the data you're using which leads you to believe that crit scales better. I'd assume that crit scales worse compared to AP, especially since your flurry uptime and UR uptime is getting very little additional benefit after you get past 30% crit or so.

Also, since hit is capped at 24% (and 6% or so for yellow damage), that doesn't scale so great after a certain point as well.

All of the above leads me to believe that the only stat which will give great returns "forever" is AP.

Mek - that's a very interesting way of looking at item budgetting, as ratios, since stacking any single stat in item budget has exponentially increasing costs (as I remember it). Very cool work/intuition there :-)

I was comparing the critdelt and apdelt colums on your most recent sheet. It appears as though as base AP increases the relative value of 100ap worth of item budget decreases, whereas the value of the equivalent amount of crit budget keeps giving good returns. Crit should continue being good right up to the crit-cap which with the amount of +hit available to Shaman should be pretty high. I'm assuming at some point spending your budget on +crit % will be a better return than a static amount of +ap.

Disquette · 05/05/07, 8:18 PM

Ah, I see. That could be. It'd be interesting to find out whether or not the numbers bear those out for sure.

mek · 05/07/07, 11:13 AM

Originally Posted by Morelis

I was comparing the critdelt and apdelt colums on your most recent sheet. It appears as though as base AP increases the relative value of 100ap worth of item budget decreases, whereas the value of the equivalent amount of crit budget keeps giving good returns. Crit should continue being good right up to the crit-cap which with the amount of +hit available to Shaman should be pretty high. I'm assuming at some point spending your budget on +crit % will be a better return than a static amount of +ap.

The above is all definitely true, but the point at which crit passes AP in efficiency is probably somewhere a bit past the 3k range. This is achievable in brief spurts at the moment, with full consumables including flask and a melee group buffed to the gills, but consistently?

Another point in crit's favour is the glancing change, which has pushed the crit cap so high it'll probably be forgotten about. But in reality, I don't think we're going to get to the AP totals that make crit the superior stat in this expansion. Though, in our brave new world of combat ratings, any increase to the level cap would also push upward the point where crit passes AP in efficiency.

Rescorla · 05/07/07, 1:05 PM

Disquette,

Will 2.1 render your simulation obsolete/inaccurate until the new mechanics of WF are coded in or have you already accounted for how WF is working on the PTR?

Thanks for all the hard work you have put into it.

Disquette · 05/07/07, 1:18 PM

I don't have anything modeled within the current (live) system, except for the aspect where doubling up same-rank windfury enchants reflects the system seen on the test server. In short, I say:

- Every hit that produces a windfury locks out all chance of windfury from either hand for the next 3 seconds
- Every attack that lands (no dodge/miss/parry/block) outside of this rule has a chance to proc WF:
- 20% for an eligible attack when only weapon has WF applied
- 36% for an eligible attack when 2 weapons have WF applied, regardless of ranks
- Either hand of Storm Strike can proc WF if it is an eligible attack according to the 3 second rule.

Friedrich · 05/07/07, 1:31 PM

Originally Posted by Disquette

I don't have anything modeled within the current (live) system, except for the aspect where doubling up same-rank windfury enchants reflects the system seen on the test server. In short, I say:

- Every hit that produces a windfury locks out all chance of windfury from either hand for the next 3 seconds
- Every attack that lands (no dodge/miss/parry/block) outside of this rule has a chance to proc WF:
- 20% for an eligible attack when only weapon has WF applied
- 36% for an eligible attack when 2 weapons have WF applied, regardless of ranks
- Either hand of Storm Strike can proc WF if it is an eligible attack according to the 3 second rule.

I think you missed one thing - in your code, if the MH and OH swing at the same time, GetSwingResults() is always called first for the MH. This means that the MH always gets a chance to WF before the OH does. It's a bit of a subtle point but I think this is why you always see more MH WF hits than OH if your weapons are the same speed.

I'd pay good money to know the way Blizzard's melee damage system is really implemented. In fact, since I bet most of that stuff is client-side, why hasn't someone reverse-engineered it out yet? Would Warden stop you from doing something like that?

Disquette · 05/07/07, 1:46 PM

You are correct Friedrich. However, the only times this is guaranteed to happen are:

- initiating combat when you have auto-attack up more than 1/2 the OH weapon speed in advance of being in melee range. (I assume this to be the case for the sim, which is why both MH and OH attack at time 0.00)
- stormstriking

after you hit your first flurry, however, and weapon speeds go from nice even x.x00 speed to x.x_ _ where _ _ is a number other than 00, then times that you're going to have simultaneous attacks is going to be miniscule because of how large a common multiple it will be.

Stormstrikes are indeed getting creditted to the main hand before offhand in my model. I too would be very interested in knowing how blizzard runs the code :-)

Morelis · 05/07/07, 2:48 PM

Originally Posted by Disquette

after you hit your first flurry, however, and weapon speeds go from nice even x.x00 speed to x.x_ _ where _ _ is a number other than 00, then times that you're going to have simultaneous attacks is going to be miniscule because of how large a common multiple it will be.

I wonder if this is how Blizzard has done it though, it's been my experience that when using 2 weapons the same speed they seem to always stay in sync. I'd wager the flurry buff is removed after the final charge is consumed, which would mean you'd always get the 1 free charge and your weapons would always swing together. Could this be modeled by just checking to see if weapon speeds are the same and then changing your flurry to have 4 charges instead of 3 if they are?

Disquette · 05/07/07, 3:51 PM

Morelis, is that done by "feel" or how the combat log shows the two attacks? I'd be most interested in a combat log, especially since it shows flurry fading and applying.

Friedrich · 05/07/07, 3:55 PM

Same speed weapons and Flurry: fact or fiction?

slant · 05/07/07, 4:11 PM

Disquette, does the simulation take the various shaman bugs into account? The big ones:

- WF and stormstrike attacks don't get the 5% crit from thundering strikes Thus roughly 50% of enhance shaman damage has a 5% lower crit chance.
- If the mainhand attack of a stormstrike is missed/dodged/blocked/parried, the offhand never attempts to strike. So 15% of stormstrikes from the front (and 5% from the back) don't attack with the offhand, losing 1/3 damage and a chance to windfury, proc flurry, etc.

Friedrich · 05/07/07, 4:18 PM

From my work with the code,

Originally Posted by slant

Disquette, does the simulation take the various shaman bugs into account? The big ones:

- WF and stormstrike attacks don't get the 5% crit from thundering strikes Thus roughly 50% of enhance shaman damage has a 5% lower crit chance.

This is not accounted for in the code, although it could be fairly easily by setting up the yellow/white hit tables differently given a user flag. Can you provide a reference/evidence for this bug?

- If the mainhand attack of a stormstrike is missed/dodged/blocked/parried, the offhand never attempts to strike. So 15% of stormstrikes from the front (and 5% from the back) don't attack with the offhand, losing 1/3 damage and a chance to windfury, proc flurry, etc.

This is properly implemented.

slant · 05/07/07, 4:37 PM

I tested it in 2.0 myself a few months ago. It's very easily verifiable grinding a couple of hours and looking at recap.

BurntSushi · 05/07/07, 4:44 PM

I have a quick question. So is the reason that a 2H build is not as viable as DW is because you need a weapon slow enough to have a swing time greater than 3 seconds with the 30% haste from flurry in order to not eat windfurys?

Disquette · 05/07/07, 4:45 PM

Hey Slant

I was very seriously considering putting in the thundering strikes bug until I went through all my recaps of some blasted lands mob grinding and found no statistical difference in the crit rates.

Still, it's stayed in the back of my mind, because I know that you maintain the FAQ and probably did equally good testing as I did. It's extremely easy to change the code to account for this, but I guess I'd like some additional data to either support your testing or mine. The problem is, I'm not sure how to get it.

My testing was done on mobs 15 levels below me (blasted lands). That could work into the figures I got, though I don't see a *logical* reason it would. Are there other mobs where I could go to verify your results that don't do much damage, are level 70+, and are plentiful? I'm willing to do the testing, but I'd like a head start if you have a good idea of where to do so.

edit: for the flurry question. This all seems like theorycraft in the link provided. I'll do some testing to verify one way or the other. I completely agree (and have modeled accordingly) with the people that say "you are either in a hasted state or not". The way that the sim models it is as follows (lifted from a post I made in the official forums to someone that had a similar question):

 Lets say we have a 2.6 MH and a 2.2 OH.

at 0.00 the MH and OH swing
at 2.20 the OH hits (next swing at 4.40)
at 2.60 the MH crits (next swing at 2.6/1.3 = 2.00 + current time = 4.60)
    recalculate OH, so of the 1.8 seconds left on its swing, decrease that to 1.8 / 1.3 = 1.38
   this means it will hit at 2.6 + 1.38 = 3.98 instead of 4.40
at 3.98 the OH hits, a charge is consumed, and OH will hit again at (2.2/1.3 = 1.69 + curr time = 5.67)
at 4.60 the MH hits, a charge is consumed, and MH will hit again at 2.6/1.3 = 2.0 + curr time = 6.60)
at 5.67 the OH hits, the last charge is consumed, so OH will hit at normal speed = 2.2 at 7.87
   recalculate MH, so of the 0.93 seconds left on its swing, increase that to .93 * 1.3 = 1.21
   this means it will hit at 5.67 + 1.21 = 6.87 instead of 6.60 (if the flurry had been maintained by
   another crit in the interim.

So, my sim will adjust weapon speeds mid swing. What I need at the moment is a 2.6 speed vendor bought very low dps weapon so I can look through a combat log to answer definitively how the game handles it.

Disquette · 05/07/07, 4:57 PM

Originally Posted by BurntSushi

I have a quick question. So is the reason that a 2H build is not as viable as DW is because you need a weapon slow enough to have a swing time greater than 3 seconds with the 30% haste from flurry in order to not eat windfurys?

The main reasons I think of it are as follows...

1) You get 100% of your AP-to-damage-conversion added to your dps with a two hander. You get 150% of it as dual wield <-- this is the big one.

2) You get more value out of 2 enchants than a single 2H enchant (2x mongoose, for instance)

3) You get 9% hit "for free" via talents using a dual wield enhance/resto spec. You could be a better utility player as a 2H build without them, but as far as dps talents go, you can't really replace them with other 2H talents that would provide an equivalent dps boost.

4) UR will be up more often with dual-wield.

Azide · 05/07/07, 5:00 PM

Originally Posted by mek

1% dps = 27ap = 25 hit rating = 17.9 crit rating.

Well, AP is certainly the winner, since it costs only 13.5 rating. Calculating back to item budget equivalencies...

2ap = 100% efficient
1HR = 54% efficient
1CR = 75% efficient

Using these numbers, you can come up with the following based off 1AP (for Pawn):
ap = 1
hr = 1.08
cr = 1.51
agi = 0.884
str = 2

Obviously, this means Bold Living Ruby (+8 str) crushes Rigid Dawnstone (+8 hit.) This makes me kind of cringe. Inscribed Noble Topaz (+4 str +4 cr) comes out just behind Bold Living Ruby. Valuing one stam at 1.5AP makes Solid Star of Elune come out ahead of the BLR.

I guess based on the above gem comparison RedSocket should be valued at 16 and Blue at 18. (Obviously if you're satisfied with your stam level and/or don't care about socket bonus you could load it up with Reds anyway.) As Str completely slaughters the other stats the "mixed" gems without it fall short. I think I'm going to "value" all gem slots at 15, as thats roughly what they'll average out to.

Here is my current pawn string.

(Pawn: v1: "Enh": Ap=1, Strength=2, Agility=0.884, CritRating=1.51, HitRating=1.08, RedSocket=15, YellowSocket=15, BlueSocket=15, Intelect=1, Stamina=1.5, Mp5=.3)

It makes me uneasy to resocket out of +hit to +str but I guess thats the best option, according the math. Crit at ~20%, Hit at ~15%, Stam wherever you're comfortable, and load up on str seems to be my plan.

Thoughts?

EDIT: My agi math was wrong as pointed out by Rob later in the thread. The correct Agi value is 1.3384 (without kings.) If you use the pawn string, be sure to make the appropriate change.

Disquette · 05/07/07, 5:08 PM

Azide, that is indeed what the simulator suggests. Whereas you cringe at it, I feel nervous too. I have *tried* to account for everything in the sim that is possible to do, but I do feel nervous about people blowing large chunks of change based on a simulation. Short of new blasted lands mobs that are lvl 70+ to test on, I don't know how to test to make sure that actual game numbers match the sim.

I do know the sim matches the damage per hit correctly by using 0.9 dps weapons and checking the calculated values against the combat log, but whereas I hope/think I have the other aspects right as well, I can't tell you 100% that it is correct.