Ok, so I've been thinking of tweaking my spec a little. Right now I'm 2/2 into Guardian totems. I've seen some use in pvp but more importantly and the primary reason of my post: During Lady Vashj fight I switch myself in the MTs group to give him grounding totem effect to eat her Shock Burst on phase 1/3. I'm wondering if any other shaman that's been in charge has done this with or without this talent as sometimes I put it down and gets absorbed right away and I shit myself as I watch my CD tick away slowly. The timing is supposedly "random" was just wondering if anyone has ever experienced it more than once every fifteen seconds.

Hello. This is my first post, but ive been reading this forum for quite some time now,
especially this thread as i am playing a shaman myself. Thanks all you guys for
working out all these facts. I am particularly taken with the AEP calculations. Because
of that, i created a small app that allows me (and you) to compare items based on
those values without being in the game (pawn) or calculating.

I do not know if such a program already existed and if so, it will most likely be better
than mine. But well, give it a try if you are curious. Note that I created the program for
me mainly, i know that it works for me (shaman) on my pc, but i did not really test the
other template features (it allows you to create aep values for nearly every stat, not only
those relevant for us shamans.) and yea, the code is sloppy. if you want it to crash,
it will crash. but normally it should work. if not, tell me.

I currently do not update the program as it suits my needs. However, if some of you guys
(more than 3) like it, tell me what to change, tell me about bugs, suggestions and
i will see what i can do. does anyone need a good xml itemdatabase btw? thats the
main problem currently as the allakhazam database does not support haste (i think).



If you are curious, you can find further facts, a small howto and more screenshots here.

i can not create a new post because i do not have enough posts. if some of you guys
think that it is worth it, you can do so (but link to this post please).

oh yea and: USE AT YOUR OWN RISK! (the program does not do anything fancy and
it will not try to open or alter any file except the templates.xml file.)

I was under the impression the CD was 15 seconds, I've never done it myself since we have a resto shaman in the with tank for this, but I dont recall them ever complaining the CD was shorter than grounding.

Since I recently got a chance to grab [Icon of Unyielding Courage], I've been wondering if anyone has yet found an AEP value for -armor. (And yes, I know that +hit isn't the best stat I can get )

There's too many uknowns and its too dynamic since it depends entirely on what the enemy's armor is at the time that you decide to ignore some more. The best we can tell is that 1 point of - Armor seems to be worth 7 item budget points (this was stated 2-3 pages back) and that its AEP value would be something between 1 and 2.2. Not very conclusive, but it would seem to indicate that its a decent stat to have.

I'm new, so forgive me if this is obvious.

The only potential drawback to -armor is that on very low armor mobs, it's possible for it to be worthless since -armor has no effect once a mobs armor is reduced to zero. However, I don't think this is currently an issue for any significant raid bosses.

In fact, the recent testing information in the Hemo thread would seem to indicate that we (the theorycrafting community) have underestimated the value of -armor for quite some time.

Very cool, thanks!

On the topic of grounding totems on vashj, I was charged with the same responsibility. I have guardian totems in my build because I pvp occasionally and like them for that purpose. It just happened to be useful in raiding too. I never encountered an instance where grounding was needed and I wasn't able to place it in time.

The problem as described to me was that the calculations do not hold true between the various ranges. -800 armor when the mob only has 2000 armor will lead to a much higher damage increase than if the mob has 20,000 armor and you remove 800. Its not a linear curve.

I think I saw a post in the DPS warrior thread that investigated this specifically.

Yea, this post:
http://elitistjerks.com/442876-post49.html concluded that -armor stops giving any benefit at all once the boss is at zero armor.

Here is the prove, that shows that "average" AEP’s use is wrong for the majority of stat combos possible in game by a big margin.

I took data from empirical testing by Disquette and used following algorithm:
For each Hit/AP combo with given crit remember recorded dps.
Find combo that is the previous one with +100 ap and get new dps.
Find the difference and divide it by 100 to get dps increase per point of AP.
Do the same for +50 hit rating step (these are minimal steps presented in collected data) and find dps increase per point of HR. Divide dps per HR by dps per AP to find local AEP ratio for the initial Hit/AP combo.
Here are the results:




The first graph shows actual AEP numbers. Look at how big standard deviation (expected difference from the mean) is = 0.77. Mean = 1.1, it is not surprising that it is different from “generally agreed upon” 1.3 as the later can be easily obtained with another covered range of stat combos. It proves again that these general “average” AEP values provide nothing for specific stat combos that players come up with. Second graph shows probability distribution function for AEP. Each column is 0.2 range – that is a significant change for the weight value. 43% of combos fall into range where actual AEP is below mean by 10% and more, 41% are above mean by 10%, only 16% of combos are within 10% difference from mean. So, there is 84% chance over the whole range of covered combos, that average AEP will be different from your actual AEP by more than 10%. There is a solid 54.5% chance that actual AEP is different from the suggested by more than 30%. That is for one pair Hit-AP only. If same holds true for other pairs, one can speculate, that the probability of at least one weights being wrong by 30% or more is 1-0.455^(number of non-linear dependant attributes). There are 4 such major attributes (AP, Hit, Crit, Haste) so it becomes 96% chance that at least 1 of the weights will be more than 30% wrong.

Statistics to the rescue, nice work.

According to the theories mentioned above the effectiveness of Sunder Armor depends on the enemy's armor. Found something interesting on wowwiki (quoting Sunder Armor - WoWWiki, the Warcraft wiki) regarding those theories, though:

That article, if I'm not totally mistaken, suggests -armor is a very valuable stat against any enemy regardless of its armor. And, by the way, implicitly claims that negative armor does not exist, which confirms the result of a quick ingame test.

You can get TargerArmor to get an "idea" of how much AC your target has, but it doesn't seem to take into account armor penetration or it's entirely inaccurate. I just use the BT neck while a warrior uses all of the -armor jazz and we're showing nearly the same numbers for our targets ac value.

Yo,

Can you do some regression analysis on the variables (AP, crit, hit, etc) and derive a set of variable functions for us to use in our simulations? Barring that, we will have to stick to the linear models we have. You seem to have the mathematical prowess to accomplish this.

On another note. Even if you derive variable functions, they will still need to be applied from the gear backwards as in the lootzor or Tornhoof's model. The reason is that although you may discover the optimal combination of stats at specific values (maximums), the gear may not exist to accommodate that specific maximum. This is one of the reasons that I liked Tornhoof's closed simulation so much. It removed all of the uncontrolled variables and tested existing gear combinations to discover the optimal combinations of gear available to you at that time. I'm speculating, but I think a combination of variable functions and a closed model like Tornhoof's will be the closest we can come to our goal.

Look at the pretty line I get when I look at AP in steps of 400, against larger steps of the increased hit and crit rate



Out of the 36 datapoints, we had:
Mean: 0.23 Dps/AP
StDev: 0.02 (9% of the mean)

It looks like we have a really great model for AP, over a huge range of hit/crit values! Furthermore, since 30% is being accepted as the "close enough" barometer for values by Yo!, I'll use that to test the 36 datapoints I have. Exactly 0 of them fall outside 30% of the mean, so I guess I have a perfect data set.

It even works out like we'd expect, with increasing returns on AP as our hit and crit rate increase (shown by the upward trend in the graph).

Statistics to the rescue indeed.

Now, back to the big_test.xls sheet that i'd uploaded so that people could review it, did everyone just ignore the lines on the bottom of my data set where I included:



I specifically put in those calculations in the big results test so that people would know how widely varied the data points were.

So, let's look at those big hit rating swings. My biggest mistake, probably, was making the steps too small. If you use larger steps, you get nicer numbers to look at. They make pretty graphs like the one above. It has so much variation between runs (that's how shaman work), that sometimes you get flukey things. When you were going through the result data, did you really think that increasing hit rating should decrease dps?

I really hope not. Yet running the sim, there were instances of that. It's yet another thing I specifically pointed out in a separate column so that people would know what they're looking it. So, instead of wallowing through the minutae just so you can exclaim "it's all wrong and worthless", how about you take some time and figure out if, in a broader context, it's meaningful?

Here's an example. I took three of the AP levels (I chose 900, 1700, 2500, because they're evenly spaced and are WAY out of each other's range, so they should satisfy your complaint about gear levels resulting diff APE ratings). I then eliminated half of the hit rating steps (increasing them by 100 instead of 50). This should give me a better picture, and again, should address your notion that APE are not valid across gear ranges.

Now let's look at the pretty picture:


No, it's not nearly as nice as the AP graph, but it's still pretty consistent. Much more than the 70% variability you were discussing. In this case, the average dps/hitrating = .275. The stdev = 0.084 (for a stdev 31% of the mean). This time, using your "30% variation is ok" concept, 71% of data points fall into the acceptable range. If I look across all gear levels, and say that each hit rating is worth just a little bit more than one AP, I'd be pretty accurate, based on the two charts I have posted above, which cover almost the entire range of gearing choices.

Statistics can say what ever you want them to, pretty much. I could probably argue on Yo!'s side and say that all this theorycraft in here is utter garbage, at best meaningless to the shaman community, and at worst ill-advising people. I just need to find the right data that makes graphs that appear horrendous or beautiful.

That's why I'm glad that the theorycraft has been born out in actual game scenarios, with multiple people of varied gear-levels saying "i changed out my hit gems for AP or crit gems, started focusing more on crit and AP, and my dps went up significantly".

I don't know how shaman dps works in many instances. There's all sorts of things that are hard to account for. But just like looking for a coin at night, the first places you check is where it's already light, and you hope you find it there.

When looking at the armor range on bosses it really isn't that large a spread, just about every boss I've seen using targetarmor falls between 3500 and 5500 armor - 25% to 35% mitigation.

As for hitting 0 and -armor losing it's value the lowest I've seen is Essence of Suffering which seems to average about 1700 after sunder and FF, add in CoR and you'd still need -900 to hit 0. Since the most a shaman can equip right now is -301(-601 with MotB procced) it's really not an issue.

As for AEP values I was hoping to get some feedback on my speculation here: http://elitistjerks.com/447956-post1339.html It seems to me that we know how much -armor changes our dps since we know the armor mitigation formulas. The only variables are how much it costs in the item budget, which we don't know but can estimate and what are good values for boss armor which we have a pretty decent idea of. So if we assume 7 -armor for 1 item point and bosses around 4500 armor, I think it's safe to say that on a shaman with T6ish stats(I use my own) it's worth about 1.9(1.8 for 5500 and 2.04 for 3500, 2.34 on EoS btw).

That is the way I understood theory so far. The wowwiki article I mentioned earlier supports that, too. That'd make -armor valuable not only against targets with just enough armor left to reduce after Sunder Armor and Faerie Fire (which is pretty obvious), but also against high armor targets.

I remember a lengthy thread on the conquest forums a couple years back basically said the same thing, and at the time it was probably true, but it doesn't appear to be anymore.

The formula on Armor - WoWWiki, the Warcraft wiki show a non-linear curve on armor mitigation. You can check it plugging random armory profiles into that formula and you'll see the mitigation % matches it exactly.

As an example using that formula if you assume a 1000 hit and then a 1000 armor reduction we'll get the following results:
10000 armor = 513.56
9000 armor = 539.82
Difference = 26.26

2000 armor = 840.73
1000 armor = 913.48
Difference = 72.74

So unless I've missed something(which is very possible), the armor the mob starts with is a large factor in determining the value of -armor.

The thing you missed is that you based your example on a static 1000 damage hit.

Lowering the Armor makes the % of reduction go down, just like adding armor makes it go up. The value of -armor will go up as your base dmg go up.

For example:

If your target has 50% DR and you can push it to 48% with -armor then thats a 4% increase in dps.

So if you push at 800 dps, then you'd be at 832, for the duration of the -armor. So as your base DPS will go up, so will the value from -armor.

The numerical difference will increase with an increase in the base damage, but the percentage change in damage will always be the same, and will always be larger for the lower armor target. Right?

Since you are using x to value the increase/decrease for both cases, it will allways be the same.

I want to add some numbers to the armor penetration discussion to support Morelis claim.

Using the same formula from wowwiki

Levels 60+ DR% = Armor / (Armor+400+85*(Level+4.5*(Level-59)))

x / (x+400+85*(70+4.5*(70-59))) =
x / (x+10557.5) =

you: lvl 70, hitting for 1000 damage, attackspeed 1.0
enemy1: lvl 70, 30.000 armor = 73.97% reduction or 260.3 dmg/hit
enemy2: lvl 70, 20.000 armor = 65.45% reduction or 345.5 dmg/hit
enemy3: lvl 70, 10.000 armor = 48.64% reduction or 513.6 dmg/hit
enemy4: lvl 70, 05.000 armor = 32.13% reduction or 678.7 dmg/hit
enemy5: lvl 70, 02.000 armor = 15.93% reduction or 840.7 dmg/hit
enemy6: lvl 70, 01.500 armor = 12.44% reduction or 875.6 dmg/hit


Debuffing the mob with sunder armor, expose armor, fearie fire or wearing armor penetration gear for an amount of -1000 armor will net you those numbers:

enemy1: lvl 70, 29.000 armor = 73.11% reduction or 268.9 dmg/hit
enemy2: lvl 70, 19.000 armor = 64.28% reduction or 357.2 dmg/hit
enemy3: lvl 70, 09.000 armor = 46.02% reduction or 539.8 dmg/hit
enemy4: lvl 70, 04.000 armor = 27.47% reduction or 725.3 dmg/hit
enemy5: lvl 70, 01.000 armor = 08.65% reduction or 913.5 dmg/hit
enemy6: lvl 70, 00.500 armor = 04.52% reduction or 954.8 dmg/hit

enemy1: 3.3% dmg increase from 1000 armor penetration
enemy2: 3.4% dmg increase from 1000 armor penetration
enemy3: 5.1% dmg increase from 1000 armor penetration
enemy4: 6.9% dmg increase from 1000 armor penetration
enemy5: 8.6% dmg increase from 1000 armor penetration
enemy6: 9.0% dmg increase from 1000 armor penetration

Bossarmor is about 5.000 to 8.000 on average defore debuffs are applied, that's why a missing fearie fire will hurt your melee close as much as a missing pot or minor attackpower buff. Armor penetration could help here a lot and it's a decent pvp stat while fighting cloth wearer with 500 resilence. But there's no way you could say 1 AP = -12.5 armor pen on general or average.

Well we certainly could do an average and work an AEP from that, but it would be pretty damn inaccurate.