Remove Advertisements

Some preliminary conclusions concerning avoidance in 3.2

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Wardari » Tue Aug 18, 2009 2:15 am

I have some other concerns regarding avoidance, although not as a result of 3.2 this is more a result of 3.1 and the insane amounts of avoidance we're having to stack in order to tank bosses in Ulduar. Namely......

Holy Shield. Now I'm aware that HS isn't a large portion or a mainstay of our TPS these days, the table below from Theck's TPS breakdown post (viewtopic.php?p=474673#p474673) shows that that we're talking around 933 damage and the resulting TPS of 2586 under the effects of RF.


Ability Damage Glyphed Threat Glyphed
ShoR 3984 0 11041 0
HotR 2950 0 8175 0
Cons 3437 0 9525 0
JoV 2264 2490 6274 6901
AS 2562 5124 7100 14200
HoW 2525 0 6998 0
SoV 435 0 1206 0
VDoT 4810 0 13330 0
HS 933 0 2586 0
Melee 697 0 1017 0

Now with the huge amount of avoidance we're stacking, i'm personally pushing over 55% pure avoidance raid buffed, more likely nearer 60% if I properly calculated it all including misses etc. Now this is great for survivability as we all know if you're at 5% health block isn't going to save your life Vs a 25k blow, only avoiding or the boss missing you will prevent the "splat dead tank".

The way the attack table works, it resolves avoidance first and blocks, crits, crushes and hits etc are the later in the table, effectively meaning the higher our avoidance the more TPS we're losing from Holy Shield due to blows not landing and thus being blocked (i'm pretty sure most of us are still able to maintain unhittable or at least very near to when raid buffed).

Attack table included for those interested.

Miss
Dodge
Parry
Glancing Blow (doesn't effect bosses, included for completeness)
Block
Critical hit
Crushing Blow (no longer an issues for boss tanking, only mobs 4+ levels, again included for completeness)
Ordinary hit

Now don't get me wrong I know other classes like Druids and Warriors are punished for stacking higher avoidance through rage starvation, something we don't suffer too much from - although I have been noticing mana starvation issues these days as my avoidance ramps up and up, specially on Vezax. Druids don't really get a double hit from it, Warriors are mildly effected too as their damage shield is worse the more they avoid, we're also punished through the loss of Holy Shield TPS, even if only a small amount of threat in comparison to the likes of ShoR and HotR etc.

I'd personally like to see Blizzard build on their block revamp as the double block value has been incredible for heroics (i'm pushing 3.5k SBV with judgement up and Lavanthor's Talisman active - 40 second duration now). They keep stating they want the tanks to feel different, then why not make Paladins the primary blocking class - we've already got the tools through redoubt and Holy Shield they just need to build on this through talents and modifying existing abilities.

They could modify HS so that it increases block value significantly (300%/400%) to bump us upto a blockable amount of say 10/15k (i'm just pulling numbers out of my arse here, the values necessary to make us on par with other tanks in terms of mitigation i'll leave upto the number crunchers) but without increasing our SBV to the equivalent amount for the likes of ShoR - i can already see the PvP cries of a protadin running around shield slamming for 15k+ that just wouldn't be good.

I've always preferred block as a mitigation stat rather than avoidance, but it's never been high enough for the likes of raiding, just an icing on the cake stat rather than a stackable main damage reduction.

This would leave us wide open to drop our avoidance levels and increase our block value and rating further, if nothing else this would result in a far more stable tank - maybe too stable?

Thoughts?
Wardari
 
Posts: 45
Joined: Tue Apr 28, 2009 2:32 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Venoseth » Tue Aug 18, 2009 2:27 am

Soralin wrote:Also, dodge has always been superior to agility in terms of pure avoidance, however the difference is minimal (off the top of my head its about 0.02% difference) and agility offers effective health gains through armor and a minimal amount of crit for threat generation in addition to its avoidance. That's why people use it over pure dodge.


You're right about them being similar, and that some people (myself included) prefer agi/stam gems to the dodge/stam gems. That being said, Agility is only 85% as efficent as dodge. Unless I'm mistaken?
"Pleasure in the job puts perfection in the work."

"Education is an ornament in prosperity and refuge in adversity."
Image
Venoseth
 
Posts: 240
Joined: Thu Oct 02, 2008 11:05 pm

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Vedge » Tue Aug 18, 2009 4:11 am

Venoseth wrote:Agility is only 85% as efficent as dodge. Unless I'm mistaken?


I believe that is without Kings. Additionally you get some minor armor and crit
Imagehttp://eu.wowarmory.com/character-sheet.xml?r=Draenor&n=Vedge
Vedge
 
Posts: 29
Joined: Sun May 17, 2009 6:50 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Lutinja » Sun Aug 30, 2009 10:46 am

But wait, i see the different ratios between defense and parry, but what happened to defense rating?

How come no one talks about that anymore?

And is it still a 2.4 dodge: 1.8 defense rating: 1 parry ratio that is supposed to be hold?
Lutinja
 
Posts: 12
Joined: Fri Jul 17, 2009 6:19 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby arktik » Sun Aug 30, 2009 4:54 pm

I could use some help here. I was replying to a post at http://maintankadin.failsafedesign.com/viewtopic.php?f=3&t=25698 where a replier got me into reading this thread. Something didnt add up real well and furthermore maybe im just having a brain fart day but i also cant figure out how to get my proper ratio either. My current stats are 28.46% dodge (670 rating), 18.93% parry (218 rating), 543 defense value (706 rating), and this is all unbuffed.

Now according to my post on the other topic i took my main set and compared it with what ratings buster said 2 other peices of gear would give me(a dodge trinking and a parry trinket and i divided the % rating buster said with the rating that stat value was, furnace stone gave 105 dodge rating and rune of repulsion gave 95 parry rating, closest things i could find). I came out with the stats showing that dodge was still a better stat. But if i used the formula recieved here where u go dodge-10/parry-10 my ratio is higher meaning parry should be better. Which is wrong? I think its me and im not taking something into account. I think its information overload because i just went over this whole topic and had to keep changing my way of thinking after reading improvements and corrections. Also my wowarmory link is in my signature. thx for any input guys, this thread has been really enlightening, just plz help me chizel it out in my head lol.
User avatar
arktik
Maintankadonor
 
Posts: 161
Joined: Tue Apr 15, 2008 2:06 pm

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby theckhd » Sun Aug 30, 2009 8:10 pm

Vedge wrote:
Venoseth wrote:Agility is only 85% as efficent as dodge. Unless I'm mistaken?


I believe that is without Kings. Additionally you get some minor armor and crit

Actually that's counting Kings. You'll get about 85% as much dodge from 1 Agility as you will 1 Dodge Rating.

However, once you consider the armor, Agility is actually around 97% as effective as Dodge at reducing incoming damage.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Call to Arms 5.x, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7658
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Xenix » Wed Sep 16, 2009 12:00 pm

@Lutinja:

The main reason defense isn't discussed that much is that there normally isn't a choice between defense/dodge/parry - yellow slots get defense only and red get dodge/parry only, so the much more interesting question is which of dodge/parry to put in the red sockets.

The secondary reason is that the actual avoidance gain you get from defense rating depends on how much of it you waste in the defense rating -> defense skill conversion. This was a much bigger deal when gems came in step sizes of 8 rating and could push you up and down the defense lines. With epic gems, you stay on the same defense line for a long, long time and that would make it a much less random stat.

Also, in Theckd's analysis earlier in the thread where he counted the stepwise nature of defense rating, his step size on the analysis was small enough that his algorithm never added more defense rating because there was no benefit (It never pushed it up to the next point of defense skill). I'll bet if he re-ran it with an increment of 5 or 10, you'd see that when (dodge rating + parry rating) > ~(2/3 * defense rating) -and- you're close to the ideal ratio (e.g. from 1.9 to 1.85) that defense rating gives more avoidance than dodge or parry. (I'm not positive on the exact numbers, but I seem to remember seeing that general trend.)

I'll spend a bit of time running some iterations on it since I'm now curious, unless Theckd beats me to it with a post (should be much easier for him since he just needs to change his step size).

TLDR: Defense compared to dodge/parry was pretty much ignored halfway through the thread, and it focused on dodge vs. parry since the value of defense changed drastically depending on how much you were adding and your current amount due to the 8 rating step size forced by rare gems. Theckd's iterative analysis re-ran with a step size of 5 rating points would probably show a target ratio of defense:dodge:parry, and it would be feasible to aim for due to gems now coming in 10 rating increments.

@Arktik:

Checking your numbers and adding the trinkets, I get:

Base: 56.877% d/p of 2.02
+105 dodge: 58.211%, d/p of 2.23, 0.0141459%/pt., +1.49% Avoidance
+95 parry: 58.114%, d/p of 1.79, 0.0146151%/pt., +1.39% Avoidance

Given that your gear is what you posted, and that the trinket you're replacing with these has no avoidance stats on it, the parry trinket should be giving you 0.1% less extra avoidance than the dodge one, but will be giving ~0.0005% more avoidance per point.

This leads me to believe you're not showing these same avoidance numbers in Ratingbuster, which probably means one of three things (I checked my calculations with my own gear and the same trinkets to verify they were using the correct formulas):

1) Out-of-date Ratingbuster mod (e.g. pre-3.2)
2) Avoidance DR's not enabled in Ratingbuster
3) You're replacing an existing avoidance trinket, which changes the numbers entirely.

-------------------------------------------------------------

Edit:
All right - some changes to what I posted earlier in the thread based on new calculations (feel free to run the numbers yourself and check them, code is below). Running an iterative analysis starting at 689 defense rating, 0 dodge and 0 parry, and adding 5 points at a time, based on which of dodge/parry/defense provides the most avoidance yields the following (given that you have a function avoidance() to calculate avoidance based on defense/dodge/parry/agility):
Code: Select all
avoidanceArray = [689 0 0 avoidance(689,0,0,0)];
avd_temp = zeros(3,4);
step_size = 5;
for i = 1:1000
  avd_temp(1,:) = [avoidanceArray(i,1)+step_size avoidanceArray(i,2) avoidanceArray(i,3) avoidance(avoidanceArray(i,1)+step_size,avoidanceArray(i,2),avoidanceArray(i,3),0)];
  avd_temp(2,:) = [avoidanceArray(i,1) avoidanceArray(i,2)+step_size avoidanceArray(i,3) avoidance(avoidanceArray(i,1),avoidanceArray(i,2)+step_size,avoidanceArray(i,3),0)];
  avd_temp(3,:) = [avoidanceArray(i,1) avoidanceArray(i,2) avoidanceArray(i,3)+step_size avoidance(avoidanceArray(i,1),avoidanceArray(i,2),avoidanceArray(i,3)+step_size,0)];
  [max_avd stat] = max(avd_temp(:,4));
  avoidanceArray = [avoidanceArray ; avd_temp(stat,:)];
end

Image
Note that the plot seems linear in three segments: From the beginning until parry gets added, from that point until defense gets added, and from then on. I'll deal with the last two segments since the first is obvious. (Edit: fixed the plot - had couple numbers colored wrong).

Segment 2:
Starts at 689/225/0, ends at 689/485/140

In this region, don't add defense, but add dodge/parry in a 1.875:1 ratio once they have the same DR. (e.g. once your dodge/parry percentages are at 1.875:1). This will always be the case for any constant value of defense rating, and this is what the previous analyses noticed. The new part is next, though:

Segment 3:
Starts at 694/485/140, ends at infinity

In this segment, we are -still- keeping the dodge-parry percentages at a ratio 1.875:1. However, we are adding in defense rating, which adds dodge and parry in equal amounts. As such, you'll need to shift the ratio of dodge:parry ratings slightly in favor of dodge to balance out the extra parry you're getting from adding defense rating.

The ratio you add def:dod:par rating in this region (if you want to add all three stats concurrently) is 1.73:2.43:1, as calculated from the slope of the three lines in the last segment (the numbers shown on the plot above).

Keep in mind two things, however.
1) This is the ratio of stats you are adding to the base of 694 defense rating, 485 dodge rating and 140 parry rating. Subtract these numbers from your current values to compare ratios.
2) You are still keeping your dodge%:parry% at 1.875:1, just using all three of defense/dodge/parry instead of only dodge/parry.

TLDR:
  • Keep your dodge%:parry% at 1.875:1 still. However, both adding defense rating and adding parry rating will lower this ratio. (although I have no idea where you'd ever have a choice between equal amounts of the two and no other consideration).

  • If you choose to add all three of defense/dodge/parry rating instead of just dodge/parry, the ideal ratio is 1.73:2.43:1, from a baseline of 694/485/140.

  • As shown earlier, just keeping your defense rating constant and stacking only dodge/parry so that you have the ideal dodge%:parry% ratio of 1.875:1 will never put you more than a few tenths of a percent off of the maximum possible avoidance for your gear. This means that unless you're crazy enough to want to optimize your gear by tweaking all three avoidance ratings for barely any gain(and somehow have the gear choices to make that possible), you can just ignore this post. Ratingbuster will tell you what provides the most avoidance anyways.
Kimurellia - Holy/Protection Paladin - Scions of Destiny - Eredar-US
User avatar
Xenix
 
Posts: 244
Joined: Thu Jun 25, 2009 4:56 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby theckhd » Mon Sep 21, 2009 8:03 am

Updating my code gives me very similar results, with a few interesting notes.

First, here's the plot of the algorithm that chooses the best stat and adds 10 to it each iteration:
Image

A little harder to read than yours because it's a compressed y-axis, but the results are identical. Segments 1 we're getting to and maintaining the 1.87 ratio (I'm starting with 300 Agi for raid buffs, which gives you enough dodge to skip your first dodge-only segment), and segment 2 we're going with the optimal three-stat ratio. My algorithm generates a 2.42 : 1.71 : 1 Dodge:Defense:Parry ratio, which agrees well with what you found. The slight differences are likely just due to boundary/initial conditions due to the discrete nature of the calculation.

There's also another source that's related to what we see on the second plot. Note the spikes in the red Defense curve. These are the "efficiency plateaus" you're talking about, where 10 defense rating gives us a three-point Defense Skill jump. They come less frequently with epic gems (as you correctly predicted), and they're reflected on the top plot by short sections of plateau in the dodge and parry lines in the top plot, though it's a little hard to see because of the limited resolution. These will tweak the ratio slightly depending on where you choose the boundaries - I'm using the values at iterations 100 and 500 to get the 2.42:1.71:1 ratio, so it includes all 5 of the defense spikes.

Just to check, let's plot the conceptually (and computationally) easier ratio of dodge/parry percentage and see how it compares to this three-stat ratio:
Image
It's clear here that the three-stat ratio and the "Keep (dodge%-base)/(parry%-base) percentage ratio at 1.87" are equivalent, as they both converge quite nicely to the values we're finding.


So I'll echo your TLDR:
  • First of all, don't slavishly follow this advice. This is for maximizing avoidance if you have two options with equal amounts of rating (i.e a gem with 10 dodge or 10 parry). It doesn't help you decide whether that item with 30 dodge rating is better or worse than another item with 60 parry rating. Use your own good judgment, or an addon such as RatingBuster for items with greatly disparate stat allocations.

  • Keep (character_sheet_dodge_% - base_dodge_%)/(character_sheet_parry_% - base_parry_%) = 1.88 to maximize avoidance. Look up your base dodge and parry. Don't forget to do this while fully raid buffed (MotW, Kings, Libram procs, etc.).

  • If you really want to micromanage, the proper ratio is 2.4:1.7:1 Dodge:Defense:Parry rating after subtracting out the break points, which is 689 defense rating, ~260 dodge rating, and ~140 parry rating after raid buffs (according to my graph anyway). These values are not exact, but they should be reasonable ballpark estimates for a fully raid-buffed tank.

  • Alternatively, here's an easier solution to micromanaging:
    • Keep the dodge/parry percent ratio above at 1.88.
    • If you have more than twice as much dodge rating as you do (defense_rating-689), then you'll be better off using defense over dodge if you have the option. The optimum ratio of dodge/defense actually declines from about 2.5 at iteration 100 to 1.6 at iteration 500, but 2 should be a reasonable enough value for most tanks and will give you a nice way to eyeball whether you should beef up your defense a bit
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Call to Arms 5.x, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7658
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby chibibro » Tue Sep 22, 2009 3:42 am

Okay I have read this threat very carefully and m impressed by the amount of calculations being dun here, I myself am a hunter and fully busy with theorycrafting on that side, But since my GF is a paladin I started to work on paladins aswell now.

After reading this post, I am wondering a few thing, mainly to the aspect of the dodge:def:parry ratio, since this ration should be 2.4:1.8:1 according to your posts I started to calculate these things for her. She has
defence741 rating - 589 base rating = 52
dodge 588-96 = 492
parry 314 -64 = 250

this gives me very weird ratio's so I gues I am doing something wrong yet I dunno what, I am working on converting your Matlab sheet to a Mathematic (program by Wolfraam) notebook since that Is what I usually work with.

Basically the bottum line is if here dodge is to low or to high, Dodge%(-10.03%) / Parry%(-10%)= 1.57 so that would say dodge is to low. But this is unbuffed(armory stats)

bottum line is I am a bit lost, I dun get scared easily by advanced math, but I simply do not get which numbers to use :P


http://eu.wowarmory.com/character-sheet.xml?r=Steamwheedle+Cartel&n=Nitika

is the paladin in question, and next to the Parry/dodge gem question is the question of the Eithrich's Oath Vs. Black Heart but that is a totally different matter
chibibro
 
Posts: 1
Joined: Tue Sep 22, 2009 3:16 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby theckhd » Tue Sep 22, 2009 6:14 am

Honestly, in practice the only thing this will be useful for is gemming and enchanting. I think people are getting the impression that having the 1.88 ratio is vitally important or something - it's not. It's entirely possible to actually reduce your overall avoidance by trying too hard to keep a 1.88 ratio if you're swapping in weaker items to do so.

As an obvious example, I'd rather have 30% dodge and 20% parry (2:1 ratio) than 28.8% dodge and 20% parry (1.88:1 ratio).

Just wear your "best" tanking gear, and make note of your (dodge-10)/(parry-10) ratio when fully raid buffed. When you're choosing gems, use that value to decide whether you get more mileage out of dodge or parry. Or use RatingBuster, which replaces the need to do any of this math at all.

In the current environment, the choice is fairly esoteric anyway, since EH is so much stronger than Avoidance at the moment. For example, I'm well above the 1.88 ratio, but I still socket Agility/Stam when I need a red gem because the agility gives me a small EH benefit, and a Parry/Stam gem wouldn't.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Call to Arms 5.x, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7658
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Thalastor82 » Thu Nov 26, 2009 4:50 am

Since this post is mainly related to guiding us in the choice of gems and enchants, maybe this is the right place to put my question

I had a look on your armory sheet Theck, and I noticed that, in accord with the fact that EH is now the most important tank characteristic, you tend to gem blue gems on yellow and even red sockets.

Now, what are the criteria to choose whether or not renounce to the socket bonus?
For instance, I am pretty sure that I can renounce without regret to the +4dodge rating on the Endurance of the Infernal neck, and also the +6stam on the triumph handguards...

But what about the Honorbound head (+12 stam, yellow slot) and the triumph legguards (+9stam, red slot)?

In any case, a mismatch in socketing makes you loose itemisation points; it is the same when enchanting 18stam on the shield instead of 20def. What would be the criterion to choose whether or not make this sacrifice?

Would it useful to look at the ratio you obtained in order to choose which green or red gems replace?
Thalastor82
 
Posts: 20
Joined: Mon Jul 20, 2009 6:27 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby KysenMurrin » Thu Nov 26, 2009 5:04 am

How to go about swapping avoidance for stamina was the main subject of the "On Progression MT Gearing" thread, and you can see the answers to your question beginning around this post.
Essentially it depends on how highly you choose to value stamina, and once you decide on that the order to follow is:

Meloree wrote:Enchant health to chest
Skip avoidance socket bonuses picked up with a yellow gem
Skip avoidance bonuses picked up with a red gem
Skip 6 stam socket bonuses picked up a yellow gem, enchant 18 stam to shield
PVP Shoulder enchant
18 stam to glvoes over agility, Skip 6 stam socket bonuses picked up with a red gem
Skip 9 stam bonuses picked up with a yellow gem
Skip 9 stam bonuses picked up with a red gem.


As you go down the list you're giving up increasing amounts of itemisation points for gains in health.
I don't play WoW any more.
Donnan - Nangun - Kysen - Kysen - Mardun - Timkins

Mostly-Book Blog.
KysenMurrin
 
Posts: 6675
Joined: Thu Jun 26, 2008 6:37 am
Location: UK

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby Thalastor82 » Thu Nov 26, 2009 5:54 am

sorry, I searched a little in the forum threads but this site is so big that I usually get lost...

I any case, thanks for the good (as usual) advice
Thalastor82
 
Posts: 20
Joined: Mon Jul 20, 2009 6:27 am

Re: Some preliminary conclusions concerning avoidance in 3.2

Postby theckhd » Thu Nov 26, 2009 10:09 am

Short version: I pick up 12 stam bonuses that require only one non-blue gem. Pretty much anything else goes straight stamina. Obviously choose to match your meta gem requirements with the biggest stamina gains you can find (like the T9 helm's 12 stam bonus).

If it takes more than one off-color gem to get the set bonus, it's always a bad investment numerically.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Call to Arms 5.x, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7658
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Previous

Return to Advanced Theorycraft and Calculations

Who is online

Users browsing this forum: No registered users and 1 guest


Remove Advertisements

Who is online

In total there is 1 user online :: 0 registered, 0 hidden and 1 guest (based on users active over the past 5 minutes)
Most users ever online was 380 on Tue Oct 14, 2008 6:28 pm

Users browsing this forum: No registered users and 1 guest