Combat Ratings at level 80 - Page 6

drumbum · 08/20/08, 9:57 AM

Well, it really depends on what Blizzard intends armor penetration to do. Currently it clearly is designed to affect low armor targets more than high armor targets. Surely we can assume Blizzard's staff has the competency to realize that armor penetration would work this way when they started adding it to gear, so they did so knowingly.

Sure, changing it to affect high armor more than low armor could work, but it completely changes the mechanic of the stat -- so it's a matter of whether doing that would be a good thing for the game or not.

Personally I'm all for it, except that if you use a pure percentage based configuration, it doesn't solve the nonlinear behavior of the stat. Instead of getting better the more you stack, armor penetration would actually get worse and worse the more you stacked. However, I'd be open to seeing a design where the stat remains linear while also affecting plate more than cloth. I'd have to think of a good way to accomplish that.

Erdluf · 08/20/08, 10:34 AM

Regarding Levva's ilvl table. assuming the formulas at Formulas:Item values - WoWWiki - Your guide to the World of Warcraft still hold, some rough equivalances:

t4(120) = 73-74 green < 70 (wotlk) blue
t5(133) = 72-73 blue > 80 green (about an 81 green)
t6(146) = 77-78 blue
t6.5(159) > 80 blue (about an 82 blue)

Formulas:

Item value
Purple = (ilvl-26)/1,2
Blue = (ilvl -.75)/1.8
Green = (ilvl-8)/2

So a green chestpiece of strength at 70 (ilvl 120) gives 56 strength. A purple (120) chestpiece of strength (if any existed) would give 78 strength.

Muphrid · 08/20/08, 9:12 PM

Well, drumbum, I could imagine that, if this is what one desired, we could force armor penetration to follow a model like...

$D/D_0 = 1 + p/(w-c)$

Where p is the penetration, c is an arbitrary constant (could well be 0), and w is the target's armor.

I suppose the key question is this: what formula should govern how much (if at all) high armor is advantaged over low armor when armor penetration is thrown into the mix?

And, related to that: what effects, if any, compensate for armor penetration pushing balance towards high-armor classes? Rather, if armor penetration is needed to push balance towards high-armor classes, it stands to reason there must be other effects that are pushing balance away from them, that armor penetration is necessary to keep balance in check. As long as we're playing the game of hypothetical developer, we have to consider those effects and try to keep armor penetration in line with counter-balancing those effects.

Edit: my mistake. I'm not sure what it is you're trying to ask for, drumbum. On the one hand, you seemed to ask for a stat that was linear, but then you also said something about having the value of the stat not diminish as you stacked it. I'm not sure what kind of model would accomplish this.

drumbum · 08/20/08, 10:32 PM

Your model seems to be accurate to me. When I say linear, I meant that $p$ should be linear with respect to $D/D_0$ , which your model achieves. This should mean the same thing as $p$ not diminishing as it is stacked, unless I am missing something. This is just a limitation I myself would impose since it is consistent with the way that most other DPS stats behave. Whether or not it is a reasonable limitation is up for debate.

I would probably generalize the model as:

$D/D_0 = 1 + f(w) \cdot p$

The only question is how to decide how much stronger armor penetration should become as armor value increases. In other words, what should $f(w)$ equal? I'd like to see $f(w)$ be an increasing function (meaning that armor penetration is more effective against higher armor targets), but this can be achieved any number of ways.

It was not my intention to suggest that $f(w)$ ought to be linear.

Muphrid · 08/21/08, 2:14 AM

I only meant that when you consider a linear increase in DPS, this affects a stat equivalence by shifting the equivalence away from the stat.

That is, we have this continuum:

<---- stacking shifts away from stat ---- no change ---- stacking shifts towards the stat ---->
 linear, quadratic, polynomial           exponential    anything that is faster-than-exponential

So when you said it shouldn't change the equivalence, I wasn't sure if you meant that in the strict sense (which would imply exponential) or in the looser sense of being on the same footing as other stats (which would be, for the most part, linear).

Anyway, I don't think we will come up with a definitive answer for what f(w) should be; ultimately, this is one of the reasons I prefer exponential curves for stats. As you can see by the continuum, the exponential is relatively unique; because it doesn't affect stat equivalences in any way, it's a natural starting point for a basis of comparison. When you start saying that a stat should get worse or get better as you stack it, you always end up with another question: by how much?

At any rate, I'm not sure I fully comprehend what you mean when you say that p being linear with respect to D/D_0 implies p does not diminish. Are we talking about the context of raw benefit, or the context of how the stat compares against other stats? Because if it's the latter, I do think we would want something more like...

$D/D_0 = e^{p \cdot f(w)}$

And that way, an increasing f(w) would increase the effectiveness of armor penetration on targets as their armor increases, but increasing armor penetration would not decrease its equivalent value in terms of other stats.

Althor · 08/21/08, 10:41 AM

Regarding spell hit in the beta, here are the results of my testing:
I'm level 77 and have 3/3 in Shadow Focus which gives me +3% to hit with Shadow spells.

Low Level Mob:
1056 casts of Mind Soothe on a level 5 mob with +0 hit rating and 3% hit from talents: 0 resists.

Equal Level Mob (Level 77, a priest mob in Zul'Drak):
137 casts of Mind Soothe with +0 hit rating, 3% from talents: 1 resist.
701 casts of Mind Soothe with a combined total of +5.04% hit: 0 resists.
585 casts of Mind Soothe with a combined total of +4.28% hit: 0 resists.

+1 Mob (Level 78, an elemental at The Maker's Overlook):
587 casts of Mind Vision with a combined total of +4.28% hit: 3 resists.
1268 casts of Mind Vision with a combined total of +5.04% hit: 0 resists.

+2 Mob (Level 79, air elemental near The Maker's Perch):
10 casts of Mind Vision with a combined total of +5.04% hit: 1 resist.
1004 casts of Mind Vision with a combined total of +6.04% hit: 0 resists.

+3 Mob (Level 80, water elemental near The Maker's Perch):
114 casts of Mind Vision with a combined total of 15.92% hit: 2 resists.

Alas I lack the +hit rating required on the character to reach 17%.

Conclusions?

Well it seems to me that the exact same miss rates that are in TBC are in WotLK beta, however that 1% unmitigatable resist is gone. Now if you hit the cap, you're going to hit. (Excluding +resist etc).

drumbum · 08/21/08, 11:56 AM

Originally Posted by Muphrid

At any rate, I'm not sure I fully comprehend what you mean when you say that p being linear with respect to D/D_0 implies p does not diminish. Are we talking about the context of raw benefit, or the context of how the stat compares against other stats? Because if it's the latter, I do think we would want something more like...

I meant the former. Specifically I meant that it ought to behave like other DPS stats do currently (i.e., that the absolute DPS gain from increasing stat $a$ by some $\Delta a$ does not depend on the current amount of stat $a$ that you have, but does depend on at least some other stats). I agree with Aldriana that an exponential relationship like you have proposed makes the game a little too uninteresting, and is probably incompatible with Blizzard's vision anyway (as you can see by all the previously unavailable stats being added to gems in WOTLK). They want more variety, not less.

Don't get me wrong; I do see the value of exponential stats, but I think that the game doesn't need to be perfectly fair with regard to how much each stat benefits the player. Instead, my goal is just to eliminate the current behavior of armor penetration at extreme values, specifically in PVP where armor levels can dip to zero easily. (See this graph for example, which demonstrates how much better armor penetration gets against low armor targets.)

Roywyn · 08/21/08, 12:17 PM

Originally Posted by Althor

Regarding spell hit in the beta, here are the results of my testing:

Awesome, thank you for the thorough and exhaustive testing!

So the model currently is 4/5/6/17% miss chance for spells against targets 0/1/2/3 levels above you.

Erdluf · 08/21/08, 1:22 PM

Originally Posted by drumbum

\my goal is just to eliminate the current behavior of armor penetration at extreme values

ArP is a stat with a cap (cap varys by target).

If ArP is OP, just change its cost (reducing ArP on existing gear). For a stat with a cap, the shape of its "value" curve is really less important than its magnitude.

Saying ArP is too difficult to model is like saying armor is too difficult to model. Should incoming DPS be linear in armor?

Jakuniku · 08/21/08, 1:47 PM

Originally Posted by Roywyn

Awesome, thank you for the thorough and exhaustive testing!

So the model currently is 4/5/6/17% miss chance for spells against targets 0/1/2/3 levels above you.

Is there a new rank of Mind Soothe in wotlk? There was some discussion on the elemental shaman topic that spell hit rates might be behaving differently for pre-wotlk ranks of spells versus the ranks you get post 70.

drumbum · 08/21/08, 3:08 PM

Originally Posted by Erdluf

ArP is a stat with a cap (cap varys by target).

If ArP is OP, just change its cost (reducing ArP on existing gear). For a stat with a cap, the shape of its "value" curve is really less important than its magnitude.

Saying ArP is too difficult to model is like saying armor is too difficult to model. Should incoming DPS be linear in armor?

Well, I also think armor penetration is very underpowered when attacking heavy armor targets, which is a little silly because physical classes need the most help against high armor targets, not low armor targets. I'd like to see armor penetration become stronger against high armor targets, but also weaker against low armor targets. This can't be done by just scaling the amount on gear up or down. It doesn't necessarily have to be better against low armor targets than high armor targets (that was just one possible idea -- see below), but I feel the gap should be much smaller than it is now. Basically I think Blizzard is in a situation where they can't balance armor penetration properly because of how wildly different it behaves relative to the target's armor value.

That said, to someone who knew nothing about the game, "armor penetration" as a term sounds like it ought to be superior against heavy armor. In fact, Blizzard actually refers to the stat as "ignores ... armor" everywhere in the game as far as I can recall, never as "armor penetration" -- that is, until the Wrath beta when they added "armor penetration rating" and armor penetration gems. I don't know if this is actually an intentional terminology change or if Blizzard just picked it up from what the playerbase has been calling it. Probably the latter, but it is possible.

Muphrid · 08/21/08, 3:25 PM

Originally Posted by Erdluf

ArP is a stat with a cap (cap varys by target).

If ArP is OP, just change its cost (reducing ArP on existing gear). For a stat with a cap, the shape of its "value" curve is really less important than its magnitude.

Saying ArP is too difficult to model is like saying armor is too difficult to model. Should incoming DPS be linear in armor?

I would disagree. The shape is very important; if we feel that armor penetration is ineffective in one situation and must be buffed, we must consider what this will do to the rest of the curve. Simply increasing the effectiveness of armor penetration by a multiplier, for instance, could very well make it overpowered at another point on the curve.

I don't think anyone is saying armor penetration is hard to come up with a model that matches its real behavior; rather it's hard to design a new model (we often use the two meanings interchangeably, but the distinction is important) to accomplish a given goal.

Finally, stats with caps have profound implications for item design. Having a stat be limited by a cap puts an extra burden on itemization, as the same templates that work at a low gear level cannot be scaled up to work at a higher gear level. With Blizzard's item budgeting function, the diminishing proportion of item points spent on a capped stat effectively diminishes the maximum potential that an item could offer. I can only think that, to ensure all classes are fairly affected by items, they would have to somehow ensure that all classes are equally affected by caps.

Erdluf · 08/21/08, 4:18 PM

Originally Posted by drumbum

Well, I also think armor penetration is very underpowered when attacking heavy armor targets, which is a little silly because physical classes need the most help against high armor targets, not low armor targets. I'd like to see armor penetration become stronger against high armor targets, but also weaker against low armor targets. This can't be done by just scaling the amount on gear up or down. It doesn't necessarily have to be better against low armor targets than high armor targets (that was just one possible idea -- see below), but I feel the gap should be much smaller than it is now. Basically I think Blizzard is in a situation where they can't balance armor penetration properly because of how wildly different it behaves relative to the target's armor value.

I disagree with the goal. Armor should be a "rock" to physical DPS's "scissors." A rogue needs more help against a Bear than he does against a Warlock. Perhaps Cloak of Shadows be changed to dispell bleeds instead of magical DoTs?

However the goal does seem easy to achieve. Current (rounded) numbers at 70 are

1k armor = 10% mitigation
16k armor = 60% mitigation

1k ArP could reduce mitigation by 10%. In other words, it removes the "first" 1k of the target's armor, rather than the "last" 1k.

Your damage against that priest goes from 90% to 100%. A boost of 11%.
Your damage against that warrior goes from 40% to 50%. A boost of 25%.

In both cases your DPS increases by the same absolute amount (at least before other forms of mitigation/avoidance). ArP would suffer the same kind of DR that armor suffers (DR in regards to DPS, linear in regards to TTL).

Toskk · 08/21/08, 6:24 PM

The one important stat conversion at level 80 that I haven't seen so far is the Agility-to-Dodge-Chance value for each class. Has anyone been able to derive these values? Druids recently discovered a similar nerf at 70 to that ratio that happened to our crit chance at 70.. this was partially concealed because like how druids were given a higher base crit chance to compensate we were given a higher base dodge chance to compensate for worse agility-to-dodge scaling. It would be very useful to determine precisely the agility-to-dodge chance in WotLK at 70 and at 80 for all classes.

Zade · 08/23/08, 4:16 PM

Any word on what base mana will be at level 80?

• frmorrison · 08/23/08, 6:36 PM

Base mana depends on Class and Race. So it would be a lot of data points to collect, but I am sure people will start hitting 80 soon.

Zade · 08/23/08, 6:53 PM

Originally Posted by frmorrison

Base mana depends on Class and Race. So it would be a lot of data points to collect, but I am sure people will start hitting 80 soon.

Base mana just depends on class. It's only 10 data points!

Xentropy · 08/23/08, 7:23 PM

Base mana for level 80 mages seems to be 3268, based on data from a lvl 80 mage posting in the mage theorycrafting thread.

Althor · 08/23/08, 8:43 PM

I'm not 80 yet so I can't give you an exact figure for Priests, but there's a fairly consistent trend in the increases so it's probably going to be around 3862 base mana.

Zurgat · 08/26/08, 2:26 AM

Another set of info which would be very useful, would be the dodge gained per agility.
Attributes - WoWWiki - Your guide to the World of Warcraft lists rogues to need 14.5 at level 60 and 20 at level 70 for 1% dodge.
Suposedly the dodge value is getting some change for WotLK though.

Whitetooth · 09/11/08, 5:12 AM

Bugged or intended? Parry, Dodge formula changed.

Beta build: 8926
Class: Death Knight
Race: Draenei
Level: 55
Base defense: 270

TotalParry	ParryRating	Parry%FromRating	+TotalParry	+ParryRating	+Parry%FromRating
6.794743061	27	1.991489329
7.005979061	30	2.212765922	0.211236000	3	0.221276592
7.215241432	33	2.434042514	0.209262371	3	0.221276592
7.422558308	36	2.655319106	0.207316875	3	0.221276592
7.627955914	39	2.876595698	0.205397606	3	0.221276592
7.966093063	44	3.245390018	0.338137150	5	0.368794320
8.033100128	45	3.319148882	0.067007065	1	0.073758864
8.299093246	49	3.614184338	0.265993118	4	0.295035456
8.496476173	52	3.835460931	0.197382927	3	0.221276592
8.692075729	55	4.056737523	0.195599556	3	0.221276592
8.821496010	57	4.204255251	0.129420280	2	0.147517728
9.014177322	60	4.425531843	0.192681313	3	0.221276592
9.205139160	63	4.646808435	0.190961838	3	0.221276592
9.644155502	70	5.163120484	0.439016342	7	0.516312048

Results show that although the rating forumla remains linear, the resulting total parry chance is not.

I'm not sure if its the more +parry you have the less parry each point gives, or if its that your base parry chance changes with the amount of +parry you have.

Tests on dodge chance have also shown a similar effect, that the resulting dodge chance is lower then what it should be if it were linear.

Block chance on the other hand does not have this effect.

At first I thought this may be a way to put a cap on avoidance, until a friend(Monolight) on beta helped me do some testing with a level 80 paladin.

The test was to see if parry from defense has this effect also, we know that each point of defense gives 0.04% parry, results showed it gave more then 0.04%, not less like the death knight.

These results are so weird I don't know if its a bug or an inteaded change.

It would be very helpful if anyone in beta could help collect more data points for analysis.

Commands:
/dump GetParryChance().." "..GetCombatRating(CR_PARRY).." "..GetCombatRatingBonus(CR_PARRY).." "..select(1, UnitDefense("player"))+select(2, UnitDefense("player"))
/dump GetDodgeChance().." "..GetCombatRating(CR_DODGE).." "..GetCombatRatingBonus(CR_DODGE).." "..select(1, UnitDefense("player"))+select(2, UnitDefense("player")).." "..select(1, UnitStat("player", 2))

/dump command is provided by AceConsole-2.0 or DevTools, if you have any ace2 addons you can use it.
Another helpful addon is Chatter, it gives you a button at the bottom right corner of your chat frame click on it to copy out your test results.
In addition to the results, please provide your Beta build, Class, Race, Level and Base defense. Its also recommended to do the testing with 0 talent points.

Thank you.

drumbum · 09/11/08, 6:06 AM

Based on the data you gave, the relationship between TotalParry (y) and ParryRating (x) fits a 2nd order polynomial regression with $R^2 = 0.9999998854$ :

$y = -0.0001007295x^2 + 0.0760130295x + 4.8163150115$

Roywyn · 09/11/08, 6:12 AM

Originally Posted by Zurgat

Another set of info which would be very useful, would be the dodge gained per agility.
Attributes - WoWWiki - Your guide to the World of Warcraft lists rogues to need 14.5 at level 60 and 20 at level 70 for 1% dodge.
Suposedly the dodge value is getting some change for WotLK though.

The values scales from 70 to 80 exactly like ratings. So at 80 for 1% dodge, you need (131/63) times the agility of 1% dodge at level 70.

Druid values are ~19.5 agility for 1% dodge at 70, and thus ~40.5 agi at 80. Not sure whether rogues changed.

Whitetooth · 09/12/08, 9:07 AM

First post updated with Armor penetration rating:

All items had the "ignore armor" stat converted to ArP rating with a conversion ratio of 7 ignore armor = 1 ArP rating.
At level 70, 7.4 ArP rating reduces enemy armor by up to 1%.

Rating	Level 60	Level 70	Level 80
Armor Penetration	4.69512177	7.40384579	15.39529991

Modal · 09/12/08, 10:35 AM

Originally Posted by Whitetooth

These results are so weird I don't know if its a bug or an inteaded change.

It would be very helpful if anyone in beta could help collect more data points for analysis.

I've been trying to get people to look into this with me over at the maintankadin boards. We have collected a little data and have a preliminary formula for dodge rating:dodge % ratio.

The formula we came up with is was: Ratio = 0.05465x(0.9994866869)^Rating

But honestly that sort of math is not my forte; you'd have to ask Arcand how he arrived at that from my data.

I won't reproduce all the data here (see the maintankadin thread if you want all the numbers), but what I found for my level 70 draenei paladin was

Dodge Rating     Dodge % per Rating
       38            .0542
      147            .0507
      277            .0471
      338            .0456
      442            .0433
      577            .0406
      742            .0377

Check out the thread here:

Maintankadin :: View topic - Defensive Stats in WotLK (diminishing returns)

I've asked about this a couple times on the beta forums, as well, but have generated virtually no interest and no blue response. Let me know if you make a post asking for confirmation and I'll gladly bump it for you.

Also, our testing shows that the contribution to Block % from both Block Rating and Defense Skill remains as in TBC, however, as you noticed, Defense Skill is contributing more than .04% per point to Dodge and Parry. The numbers are a little weird here, but it appears to average out to about .042% per point.