But if the purpose of gearing for expertise is to increase threat generation (which people here seem to agree that it is, setting aside parry thrash discussions for now), then the area of interest would be the marginal increase to threat for each point of expertise, which is what decreases with higher expertise.
I don't think "failure area" is a particularly useful thing to define. "Failure area theory" would claim that going from a 100% miss rate to a 50% miss rate is equally as effective/important/whatever as going from a 50% miss rate to a 25% miss rate. It should be pretty plain to see that this is not true.
This entire discussion gets even more complex because dodging/parrying have in themselves diminishing returns. Every time you dodge or are dodged by your opponent, the percent chance to dodge/be dodged again decreases for the next swing.
Gemming is designed to fill in the gaps. If you find your threat gen sucking wind, but you aren't taking a lot of damage hit/expertise/crit/AP are the stats to bump in your gemming. If you're getting burned down too easily then you should gem for Avoidance/Stamina. There is no one stop shopping solution, the secret is in being able to identify what you are coming up short on.
This entire discussion gets even more complex because dodging/parrying have in themselves diminishing returns. Every time you dodge or are dodged by your opponent, the percent chance to dodge/be dodged again decreases for the next swing.
What? I've always been under the impression that attacks were independent events.
But if the purpose of gearing for expertise is to increase threat generation (which people here seem to agree that it is, setting aside parry thrash discussions for now), then the area of interest would be the marginal increase to threat for each point of expertise, which is what decreases with higher expertise.
I don't think "failure area" is a particularly useful thing to define. "Failure area theory" would claim that going from a 100% miss rate to a 50% miss rate is equally as effective/important/whatever as going from a 50% miss rate to a 25% miss rate. It should be pretty plain to see that this is not true.
Here's how it would be:
Going from base 0 to 50 is 50% more effective.
Going from base 50 to 75 is 50% more effective.
Going from base 0 to 75 is 75% more effective, and the last 25% would account for 25% of it.
The difference here is that you have to account for the static chance of failure which you start with. You don't start from 0 when you're gemming, since your gear has stuff on it already, you start from, say, 75%.
Now if you're trying to generate threat, then there are two possibilities, a success (miss) and a failure (hit). Keep in mind that math in WoW is not as it seems from a cursory glance, you only roll to attack once per hit. That means that if you have 25% dodge, 25% block, 25% parry, and they have a 90% accuracy, this is what the roll will look like:
roll 1d100!
So as we can see, it's not the same math as if each were rolled seperately. If that were the case, chance to hit would be .75*.75*.75*.9 = 37%
But with wow math the chance is 100 - 25 - 25 - 25 - 10 = 15%
That's why it's important to deal with failure area instead of success area, because the 1% chance to hit you are granted by expertise has to be taken from 1 of the hit percents.
I don't think so, it even says in the mouse over that you have;
X% chance to dodge (before diminishing returns)
This is why no matter how high a level you are a mob will always hit you after a few tries.
Even if your Defense combined with general avoidance gives you roughly 99.99% chance to avoid an attack, I assure you that it will take less than a thousand swings to hit. Much less. Like in the realm of 15-20.
I don't know about diminishing returns from already having hit or missed, but I am certain than on a hit-by-hit basis, the math I stated above is accurate. Diminishing returns, if it exists as you claim it does, wouldn't change that math at all, it would just add the stipulation that after each successive X type avoidance, the chance of it happening again in the next (X units of time or hits) is reduced by X percent.
Well consider for the moment any recount from a fight in which you were tanking;
You can see what your avoidance was for that fight. Assuming a fight with no debuffs such as Patchwerk, you will find almost without fail that your avoidance in each category actually matches very closely to your avoidance percentages. I can't recall a fight where that was way off track unless I faced major positioning issues (ie. Heigan) and was being frequently attacked from the back.
In a world ruled by RNG, if these avoidance stats had no DR, wouldn't you see fights where these avoidance values are off by a significant quantity? Take a 3 minute patchwerk, as MT if each roll was completely independent you would expect to occassionally see some streaky results for avoidance. Yet you do not, this indicates that there is a diminishing return on avoidance stats, to break streaks eventually.
Otherwise we'd have to take it on good faith that every single fight is tuned for the boss to roll a statistically normal batch of attacks every single time they're fought. Survey says: Unlikely.
This entire discussion gets even more complex because dodging/parrying have in themselves diminishing returns. Every time you dodge or are dodged by your opponent, the percent chance to dodge/be dodged again decreases for the next swing.
What? I've always been under the impression that attacks were independent events.
They are. Only the ratings on your gear experience diminishing returns in how much avoidance they provide as you stack more and more.
The phenomena he talks about with the low level mobs can be figured in, too. I'm pretty sure avoidance is calculated against the opponent's level, so the DRs probably kick in earlier against lower level enemies.
Well consider for the moment any recount from a fight in which you were tanking;
You can see what your avoidance was for that fight. Assuming a fight with no debuffs such as Patchwerk, you will find almost without fail that your avoidance in each category actually matches very closely to your avoidance percentages. I can't recall a fight where that was way off track unless I faced major positioning issues (ie. Heigan) and was being frequently attacked from the back.
The sample size of attacks by the boss is large enough to guarantee that you'll come close to your avoidance percentage in the recount report. Basic statistics tells you that the more samples you take, the closer to the real value you're going to get.
Edit: And it flies in the face of your explanation of diminishing returns that I dismissed above.
Oh god, I was about to parse an entire battle, no thanks. Here's some ridiculously conservative estimates.
Again, I have never tanked a raid boss, only healed them. I'm just crunching the numbers, because I enjoy crunching numbers. However, for this sort of probability equation, results begin to average out rather quickly. If you have a TI 83 or better, it comes with a dice rolling program. If you push 12 or 20 times, you'll see some irregularities. But go to 50 or 100 and it will almost always balance itself out (and those chances are 1/6 so each 6 rolls are essentially one round).
At one attack per 5 seconds that's 36 times he will attack you.
So I parsed 8 of 20 fights, and only 2 of them went significantly off par with average avoidance. And that's after only 36 hits. At several hundred hits on the MT it's almost a mathematical certainty you'll hit your average avoidance percent on the nose.
[EDIT] The 20 rolls of 36d10 is a bit easier to look at if you just look at the totals. About 200-210 is average. Higher or lower than that means a bit more or a bit less hits. 176 was the lowest if I remember correctly and that was 22%, where 260 was the highest by about 40 points at 55%.
But if the purpose of gearing for expertise is to increase threat generation (which people here seem to agree that it is, setting aside parry thrash discussions for now), then the area of interest would be the marginal increase to threat for each point of expertise, which is what decreases with higher expertise.
I don't think "failure area" is a particularly useful thing to define. "Failure area theory" would claim that going from a 100% miss rate to a 50% miss rate is equally as effective/important/whatever as going from a 50% miss rate to a 25% miss rate. It should be pretty plain to see that this is not true.
Here's how it would be:
Going from base 0 to 50 is 50% more effective.
Going from base 50 to 75 is 50% more effective. [note that this is most definitely not as effective as going from 0% hit to 50% hit (i.e., 100% failure to 50% failure)]
Going from base 0 to 75 is 75% more effective, and the last 25% would account for 25% of it.
The difference here is that you have to account for the static chance of failure which you start with. You don't start from 0 when you're gemming, since your gear has stuff on it already, you start from, say, 75%.
Your starting point is arbitrary, though, and will vary from player to player. The example using edge case numbers just highlights the problem more clearly.
Let's say you start with an 80% hit rate from your ungemmed, unenchanted gear. This is a realistic WoW number. If you gem for enough expertise to gain 2% hit rate (up to 82%), you're now hitting 2.5% more often and generating 2.5% more threat. Your explanation would be that there is a 10% increase in effectiveness (from 20% miss rate to 18% miss rate, a reduction of 10%).
Let's say you enchant and gem your way to 90% hit rate, and then you have another socket and are able to gem for 1% hit rate in that socket (up to 91%). You're hitting about 1.1% more often and generating 1.1% more threat. Your explanation would be that there is a 10% increase in effectiveness (from 10% miss rate to 9% miss rate, a reduction of 10%).
Your theory claims that X expertise when you're at 80% hit rate is worth only as much as half of X expertise when you're at 90% hit rate, but this simply isn't the case.
Now if you're trying to generate threat, then there are two possibilities, a success (miss) and a failure (hit). Keep in mind that math in WoW is not as it seems from a cursory glance, you only roll to attack once per hit. That means that if you have 25% dodge, 25% block, 25% parry, and they have a 90% accuracy, this is what the roll will look like:
Uh, yeah, I'm obviously aware of that (and have been for years). My posts and calculations haven't indicated anything to the contrary.
So as we can see, it's not the same math as if each were rolled seperately. If that were the case, chance to hit would be .75*.75*.75*.9 = 37%
But with wow math the chance is 100 - 25 - 25 - 25 - 10 = 15%
That's why it's important to deal with failure area instead of success area, because the 1% chance to hit you are granted by expertise has to be taken from 1 of the hit percents.
Perhaps my edge case used because of its ease of making my point confused you as to my understanding, or maybe because I'm using the term "hit" to mean successfully landed attacks, when WoW has a more specific reference for the term "hit." Assume that in my example above (the 80%/90% scenario), your current gear has already pushed misses off the table and all that remains is 10% dodge and 10% parry (or 5%/5% in the 90% hit situation) and that we're ignoring blocks since they can't be avoided anyway and generally don't completely negate your attacks, so your single roll hit tables look like this:
80% hit scenario
1-80: hits and crits
81-90: dodge
91-100: parry
90% hit scenario
1-90: hits and crits
91-95: dodge
96-100: parry
With tables like that, expertise is the only relevant accuracy stat for removing misses/adding hits/increasing threat/etc.
Well consider for the moment any recount from a fight in which you were tanking;
You can see what your avoidance was for that fight. Assuming a fight with no debuffs such as Patchwerk, you will find almost without fail that your avoidance in each category actually matches very closely to your avoidance percentages. I can't recall a fight where that was way off track unless I faced major positioning issues (ie. Heigan) and was being frequently attacked from the back.
In a world ruled by RNG, if these avoidance stats had no DR, wouldn't you see fights where these avoidance values are off by a significant quantity? Take a 3 minute patchwerk, as MT if each roll was completely independent you would expect to occassionally see some streaky results for avoidance. Yet you do not, this indicates that there is a diminishing return on avoidance stats, to break streaks eventually.
Otherwise we'd have to take it on good faith that every single fight is tuned for the boss to roll a statistically normal batch of attacks every single time they're fought. Survey says: Unlikely.
Maybe I'm missing something because I'm not a tank expert, but on single Naxx WWS I see the percentage of hits parried by our warrior MT vary from 30% (Faerlina) to 16% (Gluth) with his base parry being 20%. Comparing two Patchwerk fights also gives about the same variance with 17% and 26%. I also believe your wrong about your definition of DR as it doesn't have anything to do with lowering your avoidance based on the previous attacks as each attack is a separate event. All DR means is that as you accumulate more dodge each further point of dodge rating is worth less.
Now if you're trying to generate threat, then there are two possibilities, a success (miss) and a failure (hit). Keep in mind that math in WoW is not as it seems from a cursory glance, you only roll to attack once per hit. That means that if you have 25% dodge, 25% block, 25% parry, and they have a 90% accuracy, this is what the roll will look like:
Uh, yeah, I'm obviously aware of that (and have been for years). My posts and calculations haven't indicated anything to the contrary.
Sorry! I don't mean to point out the obvious, I just have a habit of working from the most basic steps and showing my work all the way up, regardless of my assumptions about who I'm speaking to.
Moving on!
I understand your math and where you're coming from, but I would assert that the point you start from does indeed matter. If your gear gives you 70% hit, then adding 1% hit to your gear is definitely going to be less valuable to you than adding 1% hit to gear that grants you 90% hit, and more valuable than adding 1% to gear that grants you 50% hit.
We're not talking about starting from zero, obviously, we're talking about gemming gear. And so if your gear starts with 70% hit on it, then we can throw 70 of those hit cells out right now, they're not even part of the equation, because whether we gem for expertise or not doesn't change the fact that those 70 are going to come up a hit.
So lets use two examples, gear for 70% hit and 80% hit. Again we throw away the 70 or 80 % we hit already because this value is static and is not relevant to our improvement, only the area on which we can improve is relevant. In the first case, an extra % is worth 1/30th of the total area we are working on converting, and the other is worth 1/20th of the area we're working on converting. The main thing I think we're in contention on is whether the % of whatever statistic you gain from gear is relevant to the equation. I would say no, since we're talking about starting with the gear we've got, and regardless of how we improve it, those stats are going to be constant. So we only want to take into account the times where failure will be possible, and improve upon those areas.
Rend on
0
SmasherStarting to get dizzyRegistered Userregular
I don't think so, it even says in the mouse over that you have;
X% chance to dodge (before diminishing returns)
This is why no matter how high a level you are a mob will always hit you after a few tries.
Even if your Defense combined with general avoidance gives you roughly 99.99% chance to avoid an attack, I assure you that it will take less than a thousand swings to hit. Much less. Like in the realm of 15-20.
Sorry, but this is wrong.
The "before diminishing returns" for a given stat refers to the percentage displayed for your rating, though strangely not for your overall chance. For example, my tooltip says "Dodge Chance 23.17 Dodge Rating of 326 adds 8.29% Dodge (Before diminishing returns)". If I take off my keystone Great-Ring which has no defense and 49 dodge rating (1.25% at 80) it says "Dodge Chance 22.21 Dodge Rating of 326 adds 7.04% Dodge (Before diminishing returns)". Since 23.17-22.21=.96, it's clearly taking diminishing returns into account for that value. However, 8.29-7.04=1.25, so it's not taking it into account for that.
I just went to Elwynn Forest and aggroed a couple level 6 mobs and let them attack me for at least five minutes. They attacked once every two seconds or so, and there were two of them, so that's at least 300 attacks on me, and not one of them hit me.
We're not talking about starting from zero, obviously, we're talking about gemming gear. And so if your gear starts with 70% hit on it, then we can throw 70 of those hit cells out right now, they're not even part of the equation, because whether we gem for expertise or not doesn't change the fact that those 70 are going to come up a hit.
The 0 example is used only to highlight the endpoint in the curve that represents how marginally effective accuracy stats are at varying miss levels. The curve would look something like the first quadrant here, because of the infinite value of going from 0 to 1% hit (100% to 99% miss), and the trend from there on up should be clear. Regardless...
So lets use two examples, gear for 70% hit and 80% hit. Again we throw away the 70 or 80 % we hit already because this value is static and is not relevant to our improvement, only the area on which we can improve is relevant. In the first case, an extra % is worth 1/30th of the total area we are working on converting, and the other is worth 1/20th of the area we're working on converting. The main thing I think we're in contention on is whether the % of whatever statistic you gain from gear is relevant to the equation. I would say no, since we're talking about starting with the gear we've got, and regardless of how we improve it, those stats are going to be constant. So we only want to take into account the times where failure will be possible, and improve upon those areas.
You can't throw away the 70 or 80% hit that you already have, as that is what accounts for your current ability to put out X or Y TPS. Going from 70 to 71% hit takes you from X TPS to X*1.01429 TPS (because 71 out of 100 combat rolls will hit now instead of 70 out of 100 combat rolls, 71/70=1.01429). Going from 70 to 80% hit takes you from X TPS to X*1.1429 TPS (because 80 out of 100 combat rolls will hit now instead of 70 out of 100, 80/70 = 1.1429).
Assuming you do 1000 TPS at 70% hit:
70%: 1000
71%: 1014.29
80%: 1142.86
81%: 1157.14
This actually shows that each percent of accuracy stats past some arbitrary starting point increases your TPS by the same absolute amount (1/70*1000 = 14.29).
forty on
Officially the unluckiest CCG player ever.
0
Dhalphirdon't you open that trapdooryou're a fool if you dareRegistered Userregular
I'm hovering around the 70% mark now, but I'd argue that avoidance beyond that point is in fact less valuable as you accrue more of it. Once you hit a level where the raid's healers are comfortable, you don't need to stress those stats anymore.
On the other hand, being able to generate more threat is only going to become more important as damage scales up with new content and gear.
how are you at 70% avoidance? that would mean you have a solid 30% dodge and parry, which is impossible at this gear level.
Alright, I'll concede to that. With regards to something like TPS, it's not an issue of hit or miss anymore, if we assume that 1% hit = 14.29 TPS. I think that we are essentially arguing different points, or at least started by arguing different points, but regardless, are you asserting that the TPS benefit of 1% expertise rises, lowers, or stays the same in value relative to other percent points, as your percent gets higher? Are you saying it's a linear progression?
I can get behind that, I suppose. My original point was that if we are simply considering a success or failure scenario (which improving TPS is most definitely not) that my statements would stand, but it would indeed seem you are correct here about the improvement of TPS not following the same rules.
That would present an interesting decision on the topic of gemming. While avoidance does get more valuable as you put more percentage into it, expertise goes up linearly, as do HP (right?).
Does anyone know if diminishing returns are similar for expertise/avoidance ratings?
I'm hovering around the 70% mark now, but I'd argue that avoidance beyond that point is in fact less valuable as you accrue more of it. Once you hit a level where the raid's healers are comfortable, you don't need to stress those stats anymore.
On the other hand, being able to generate more threat is only going to become more important as damage scales up with new content and gear.
how are you at 70% avoidance? that would mean you have a solid 30% dodge and parry, which is impossible at this gear level.
Yeah sorry that's quoted from QuickArmory which does it wrong.
Iroh on
[SIGPIC][/SIGPIC]
0
Dhalphirdon't you open that trapdooryou're a fool if you dareRegistered Userregular
Alright, I'll concede to that. With regards to something like TPS, it's not an issue of hit or miss anymore, if we assume that 1% hit = 14.29 TPS. I think that we are essentially arguing different points, or at least started by arguing different points, but regardless, are you asserting that the TPS benefit of 1% expertise rises, lowers, or stays the same in value relative to other percent points, as your percent gets higher? Are you saying it's a linear progression?
I can get behind that, I suppose. My original point was that if we are simply considering a success or failure scenario (which improving TPS is most definitely not) that my statements would stand, but it would indeed seem you are correct here about the improvement of TPS not following the same rules.
That would present an interesting decision on the topic of gemming. While avoidance does get more valuable as you put more percentage into it, expertise goes up linearly, as do HP (right?).
Does anyone know if diminishing returns are similar for expertise/avoidance ratings?
there is no diminishing returns on expertise ratings.
Avoidance does become more valuable the more percentage of it you add, but because avoidance on gear isn't in the form of a percentage and is in fact in the form of ratings, which count for less and less as you get more, the value of avoidance actually stays relatively static as you increase it.
Alright, I'll concede to that. With regards to something like TPS, it's not an issue of hit or miss anymore, if we assume that 1% hit = 14.29 TPS. I think that we are essentially arguing different points, or at least started by arguing different points, but regardless, are you asserting that the TPS benefit of 1% expertise rises, lowers, or stays the same in value relative to other percent points, as your percent gets higher? Are you saying it's a linear progression?
Yeah, it appears that the progression is linear. Mathing it out like that kind of kicks off some intuitive agreement for me, as well.
Again, going back to the 1-100 combat roll system (and either ignoring the impact of crits or pretending they've been averaged into our average hit damage), if you cause X threat when one of your attacks hits, your average threat per swing/attack would be X*hit% (out of 100 swings, hit% will land and [100-hit%] will not). In this case, hit% refers to the percent chance for your attack to land successfully (or 100% less miss, dodge, and parry). Increasing your chance for a successfully landed attack by 1% (either though hit rating or expertise rating) would add X*1%, or .01X, to your average threat per swing/attack, since it is adding one more successful die roll outcome out of the 100 possible outcomes.
It's not really necessary for the math here to calculate this as TPS, since TPS could theoretically be derived from average threat per landed attack using attack speed, GCD, ability cooldowns etc., but it would add a lot of needless complexity and would not change the fundamental .01X threat per hit% factor.
Edit: Err, everything from here on down is basically exactly what Dhalphir posted.
That would present an interesting decision on the topic of gemming. While avoidance does get more valuable as you put more percentage into it, expertise goes up linearly, as do HP (right?).
Well, I think the diminishing returns on avoidance ratings work against the increased value of each percent of avoidance as you have more, such that avoidance stats have approximately linearly increasing value as well, much like how armor is a linear increase.
They're going to be a lot closer in DPS in 3.1, with Arms pulling ahead in a lot of tests I've seen. It's also slightly easier to gear for because you only need one 2H instead of two.
Gearwise, the hit cap is higher, but the expertise cap will be lower (if you take 2/2 WM). You prioritize crit slightly higher because you're in Battle instead of Berserker, but I'm not sure about the exact numbers to target.
Last night I was taking Hatefuls on Patchwerk. Dk was MT. I had to stop attacking for 30 seconds midway into the fight, then I stole aggro again, stopped attacking for 15 seconds, said fuck it and became new MT. Did this DK just suck, or is it his classes fault :P. I am oblivious to DK tanking abilities.
Warriors typically generate more threat than DKs do and you could have simply paid attention to Omen instead of potentially wiping the raid, though the DK is a better hateful soak then you are.
Warriors typically generate more threat than DKs do and you could have simply paid attention to Omen instead of potentially wiping the raid, though the DK is a better hateful soak then you are.
Yeah. That Patch story is more of a bad reflection on you than anything.
Once you establish a safe margin over DPS while watching Omen, making sure that you don't move up over the MT, just keep TC and Demo Shout up and you will pretty much hover at a good level without doing much more.
Last night I was taking Hatefuls on Patchwerk. Dk was MT. I had to stop attacking for 30 seconds midway into the fight, then I stole aggro again, stopped attacking for 15 seconds, said fuck it and became new MT. Did this DK just suck, or is it his classes fault :P. I am oblivious to DK tanking abilities.
Most people are shit at world of warcraft, and most DKs are worse than that, how much time did you give him to build threat?
I can safely say I needed about five seconds on that fight (as a paladin tank) to keep threat. Never had anyone pull from me, not even FFB mages.
As a OT I needed about, hrm, maybe slightly more time but not much.
Beck on
Lucas's Franklin Badge reflected the lightning back!
So I'm leveling a warrior. At 58 I respecced to fury in preparation for Titan's grip thinking "why the hell not" and started leveling in outlands.
It was painful, most fights were very slow due to a combination of using a single 2h in a built optimized for DW and the higher miss-chance from being low level. The spec was boring me and I figured if Titan's Grip wasn't a huge difference I was gonna go back to arms.
And holy shit what a difference it was. The spec went from boring and slow to a total blast to play. No wonder they're nerfing it next patch; it makes such a huge difference - perhaps less of one going from DWing 1hs to DWing 2hs, but still.
And holy shit what a difference it was. The spec went from boring and slow to a total blast to play. No wonder they're nerfing it next patch; it makes such a huge difference - perhaps less of one going from DWing 1hs to DWing 2hs, but still.
Maybe I'm misinterpreting your post, but are you implying that they're nerfing arms next patch? They're not. It's actually getting buffed a bit.
Who the hell still uses Omen? Blizzard built in threat is way more accurate. I just use ClassicThreatMeters to display the Blizzard threat in a graph representation. It's never been wrong.
Also, as I've gotten fucking tired of XPerl, I found a mod that turns your UI into the Fallout 3 UI. I'm gonna test it for the raid tonight. Should be interesting.
As a DPS, WoW's built-in threat doesn't help much on a fight with multiple tanks that need to stay above you on the threat meter (Patchwerk, Archavon, probably others) since it only tells you where you sit in relation to the highest threat player.
I was under the impression that Omen now just uses the built-in threat API to create its meters. Is this not the case?
Who the hell still uses Omen? Blizzard built in threat is way more accurate. I just use ClassicThreatMeters to display the Blizzard threat in a graph representation. It's never been wrong.
Also, as I've gotten fucking tired of XPerl, I found a mod that turns your UI into the Fallout 3 UI. I'm gonna test it for the raid tonight. Should be interesting.
Omen has been using Blizzards built in threat values since they were released.
khain on
0
Dhalphirdon't you open that trapdooryou're a fool if you dareRegistered Userregular
Who the hell still uses Omen? Blizzard built in threat is way more accurate. I just use ClassicThreatMeters to display the Blizzard threat in a graph representation. It's never been wrong.
Also, as I've gotten fucking tired of XPerl, I found a mod that turns your UI into the Fallout 3 UI. I'm gonna test it for the raid tonight. Should be interesting.
Omen is 100% accurate. More accurate, in fact, than the Blizzard one, as it shows threat values, not just percentage.
Who the hell still uses Omen? Blizzard built in threat is way more accurate. I just use ClassicThreatMeters to display the Blizzard threat in a graph representation. It's never been wrong.
Also, as I've gotten fucking tired of XPerl, I found a mod that turns your UI into the Fallout 3 UI. I'm gonna test it for the raid tonight. Should be interesting.
Omen has been using Blizzards built in threat values since they were released.
This I did not know. I thought it kept using it's own API.
And holy shit what a difference it was. The spec went from boring and slow to a total blast to play. No wonder they're nerfing it next patch; it makes such a huge difference - perhaps less of one going from DWing 1hs to DWing 2hs, but still.
Maybe I'm misinterpreting your post, but are you implying that they're nerfing arms next patch? They're not. It's actually getting buffed a bit.
The way I read it, he's talking about Titan's Grip being awesome.
Posts
I don't think "failure area" is a particularly useful thing to define. "Failure area theory" would claim that going from a 100% miss rate to a 50% miss rate is equally as effective/important/whatever as going from a 50% miss rate to a 25% miss rate. It should be pretty plain to see that this is not true.
Gemming is designed to fill in the gaps. If you find your threat gen sucking wind, but you aren't taking a lot of damage hit/expertise/crit/AP are the stats to bump in your gemming. If you're getting burned down too easily then you should gem for Avoidance/Stamina. There is no one stop shopping solution, the secret is in being able to identify what you are coming up short on.
What? I've always been under the impression that attacks were independent events.
Here's how it would be:
Going from base 0 to 50 is 50% more effective.
Going from base 50 to 75 is 50% more effective.
Going from base 0 to 75 is 75% more effective, and the last 25% would account for 25% of it.
The difference here is that you have to account for the static chance of failure which you start with. You don't start from 0 when you're gemming, since your gear has stuff on it already, you start from, say, 75%.
Now if you're trying to generate threat, then there are two possibilities, a success (miss) and a failure (hit). Keep in mind that math in WoW is not as it seems from a cursory glance, you only roll to attack once per hit. That means that if you have 25% dodge, 25% block, 25% parry, and they have a 90% accuracy, this is what the roll will look like:
So as we can see, it's not the same math as if each were rolled seperately. If that were the case, chance to hit would be .75*.75*.75*.9 = 37%
But with wow math the chance is 100 - 25 - 25 - 25 - 10 = 15%
That's why it's important to deal with failure area instead of success area, because the 1% chance to hit you are granted by expertise has to be taken from 1 of the hit percents.
X% chance to dodge (before diminishing returns)
This is why no matter how high a level you are a mob will always hit you after a few tries.
Even if your Defense combined with general avoidance gives you roughly 99.99% chance to avoid an attack, I assure you that it will take less than a thousand swings to hit. Much less. Like in the realm of 15-20.
You can see what your avoidance was for that fight. Assuming a fight with no debuffs such as Patchwerk, you will find almost without fail that your avoidance in each category actually matches very closely to your avoidance percentages. I can't recall a fight where that was way off track unless I faced major positioning issues (ie. Heigan) and was being frequently attacked from the back.
In a world ruled by RNG, if these avoidance stats had no DR, wouldn't you see fights where these avoidance values are off by a significant quantity? Take a 3 minute patchwerk, as MT if each roll was completely independent you would expect to occassionally see some streaky results for avoidance. Yet you do not, this indicates that there is a diminishing return on avoidance stats, to break streaks eventually.
Otherwise we'd have to take it on good faith that every single fight is tuned for the boss to roll a statistically normal batch of attacks every single time they're fought. Survey says: Unlikely.
The phenomena he talks about with the low level mobs can be figured in, too. I'm pretty sure avoidance is calculated against the opponent's level, so the DRs probably kick in earlier against lower level enemies.
Edit: And it flies in the face of your explanation of diminishing returns that I dismissed above.
Again, I have never tanked a raid boss, only healed them. I'm just crunching the numbers, because I enjoy crunching numbers. However, for this sort of probability equation, results begin to average out rather quickly. If you have a TI 83 or better, it comes with a dice rolling program. If you push 12 or 20 times, you'll see some irregularities. But go to 50 or 100 and it will almost always balance itself out (and those chances are 1/6 so each 6 rolls are essentially one round).
At one attack per 5 seconds that's 36 times he will attack you.
I rolled for him! http://invisiblecastle.com/roller/view/2003035/
let me parse some of that for you.
hhhmmmhhmmhmmmmhhmhhhhmhmhhmhhhmhhmm
hhhhhhhhhhhhhhhhhhhhmmmmmmmmmmmmmmmm
hit: 20/36 - 55% hit(outlier)
36d10 → [10,10,5,10,4,1,2,3,6,6,3,6,2,7,6,3,10,9,5,2,9,10,5,10,8,1,2,6,9,1,9,1,2,9,7,4] = (203)
hhmhmmmmmmmmmmmmhhmmhhmhhmmmhmhmmhmm
hhhhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmm
Hit: 12/36 - 33% hit
36d10 → [10,4,9,8,9,10,4,5,6,6,6,4,6,2,3,9,10,7,10,8,4,1,4,4,4,7,6,2,10,8,7,5,3,6,4,6] = (217)
hmhhhhmmmmmmmmmhhmhhmmmmmmmmhhmmmmmm
hhhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmmm
Hit: 11/36 - 30.5% hit
36d10 → [9,5,1,7,10,5,3,5,4,4,7,2,4,9,1,8,7,9,6,4,7,10,2,8,1,5,10,1,2,8,10,10,5,6,2,3] = (200)
hmmmhmmmmmmmmhmhmhmmmhmhmmhmmhhhmmmm
hhhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmmm
Hit: 11/36
36d10 → [8,2,8,9,2,2,2,2,4,9,5,4,8,4,4,5,7,10,10,8,3,7,4,4,10,4,3,7,1,3,9,6,7,1,3,2] = (187)
hmhhmmmmmhmmhmmmmhhhmmmmhmmmmmhmmmmm
hhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmmmm
Hit: 10/36 - 27% hit
36d10 → [1,10,10,2,6,2,10,10,6,10,3,2,9,8,9,3,2,6,8,6,10,3,7,4,6,4,4,2,4,9,9,3,1,4,2,5] = (200)
mhhmmmhhmhmmhhhmmmhmhmmmmmmmmhhmmmmm
hhhhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmm
hit: 12/36
36d10 → [8,1,7,3,9,9,3,10,9,2,8,4,8,3,2,7,1,5,4,8,8,3,1,4,9,4,3,5,8,1,1,9,3,10,1,7] = (188)
hmmmhhmhhmhmhmmmmmmhhmmmhmmmhmmhmhmm
hhhhhhhhhhhhhmmmmmmmmmmmmmmmmmmmmmmm
hit: 13/36 - 36% hit
36d10 → [9,1,6,3,3,9,4,7,7,9,3,2,1,10,1,3,6,7,1,8,2,2,1,4,6,9,9,5,4,5,6,2,7,5,8,1] = (176)
hmmmmhmmmhmmmhmmmmmhmmmmmhhmmmmmmmhm
hhhhhhhhmmmmmmmmmmmmmmmmmmmmmmmmmmmm
hit: 8/36 - 22% hit
So I parsed 8 of 20 fights, and only 2 of them went significantly off par with average avoidance. And that's after only 36 hits. At several hundred hits on the MT it's almost a mathematical certainty you'll hit your average avoidance percent on the nose.
[EDIT] The 20 rolls of 36d10 is a bit easier to look at if you just look at the totals. About 200-210 is average. Higher or lower than that means a bit more or a bit less hits. 176 was the lowest if I remember correctly and that was 22%, where 260 was the highest by about 40 points at 55%.
Let's say you start with an 80% hit rate from your ungemmed, unenchanted gear. This is a realistic WoW number. If you gem for enough expertise to gain 2% hit rate (up to 82%), you're now hitting 2.5% more often and generating 2.5% more threat. Your explanation would be that there is a 10% increase in effectiveness (from 20% miss rate to 18% miss rate, a reduction of 10%).
Let's say you enchant and gem your way to 90% hit rate, and then you have another socket and are able to gem for 1% hit rate in that socket (up to 91%). You're hitting about 1.1% more often and generating 1.1% more threat. Your explanation would be that there is a 10% increase in effectiveness (from 10% miss rate to 9% miss rate, a reduction of 10%).
Your theory claims that X expertise when you're at 80% hit rate is worth only as much as half of X expertise when you're at 90% hit rate, but this simply isn't the case.
Uh, yeah, I'm obviously aware of that (and have been for years). My posts and calculations haven't indicated anything to the contrary.
Perhaps my edge case used because of its ease of making my point confused you as to my understanding, or maybe because I'm using the term "hit" to mean successfully landed attacks, when WoW has a more specific reference for the term "hit." Assume that in my example above (the 80%/90% scenario), your current gear has already pushed misses off the table and all that remains is 10% dodge and 10% parry (or 5%/5% in the 90% hit situation) and that we're ignoring blocks since they can't be avoided anyway and generally don't completely negate your attacks, so your single roll hit tables look like this:
80% hit scenario
1-80: hits and crits
81-90: dodge
91-100: parry
90% hit scenario
1-90: hits and crits
91-95: dodge
96-100: parry
With tables like that, expertise is the only relevant accuracy stat for removing misses/adding hits/increasing threat/etc.
Maybe I'm missing something because I'm not a tank expert, but on single Naxx WWS I see the percentage of hits parried by our warrior MT vary from 30% (Faerlina) to 16% (Gluth) with his base parry being 20%. Comparing two Patchwerk fights also gives about the same variance with 17% and 26%. I also believe your wrong about your definition of DR as it doesn't have anything to do with lowering your avoidance based on the previous attacks as each attack is a separate event. All DR means is that as you accumulate more dodge each further point of dodge rating is worth less.
Sorry! I don't mean to point out the obvious, I just have a habit of working from the most basic steps and showing my work all the way up, regardless of my assumptions about who I'm speaking to.
Moving on!
I understand your math and where you're coming from, but I would assert that the point you start from does indeed matter. If your gear gives you 70% hit, then adding 1% hit to your gear is definitely going to be less valuable to you than adding 1% hit to gear that grants you 90% hit, and more valuable than adding 1% to gear that grants you 50% hit.
We're not talking about starting from zero, obviously, we're talking about gemming gear. And so if your gear starts with 70% hit on it, then we can throw 70 of those hit cells out right now, they're not even part of the equation, because whether we gem for expertise or not doesn't change the fact that those 70 are going to come up a hit.
So lets use two examples, gear for 70% hit and 80% hit. Again we throw away the 70 or 80 % we hit already because this value is static and is not relevant to our improvement, only the area on which we can improve is relevant. In the first case, an extra % is worth 1/30th of the total area we are working on converting, and the other is worth 1/20th of the area we're working on converting. The main thing I think we're in contention on is whether the % of whatever statistic you gain from gear is relevant to the equation. I would say no, since we're talking about starting with the gear we've got, and regardless of how we improve it, those stats are going to be constant. So we only want to take into account the times where failure will be possible, and improve upon those areas.
Sorry, but this is wrong.
The "before diminishing returns" for a given stat refers to the percentage displayed for your rating, though strangely not for your overall chance. For example, my tooltip says "Dodge Chance 23.17 Dodge Rating of 326 adds 8.29% Dodge (Before diminishing returns)". If I take off my keystone Great-Ring which has no defense and 49 dodge rating (1.25% at 80) it says "Dodge Chance 22.21 Dodge Rating of 326 adds 7.04% Dodge (Before diminishing returns)". Since 23.17-22.21=.96, it's clearly taking diminishing returns into account for that value. However, 8.29-7.04=1.25, so it's not taking it into account for that.
I just went to Elwynn Forest and aggroed a couple level 6 mobs and let them attack me for at least five minutes. They attacked once every two seconds or so, and there were two of them, so that's at least 300 attacks on me, and not one of them hit me.
You can't throw away the 70 or 80% hit that you already have, as that is what accounts for your current ability to put out X or Y TPS. Going from 70 to 71% hit takes you from X TPS to X*1.01429 TPS (because 71 out of 100 combat rolls will hit now instead of 70 out of 100 combat rolls, 71/70=1.01429). Going from 70 to 80% hit takes you from X TPS to X*1.1429 TPS (because 80 out of 100 combat rolls will hit now instead of 70 out of 100, 80/70 = 1.1429).
Assuming you do 1000 TPS at 70% hit:
70%: 1000
71%: 1014.29
80%: 1142.86
81%: 1157.14
This actually shows that each percent of accuracy stats past some arbitrary starting point increases your TPS by the same absolute amount (1/70*1000 = 14.29).
how are you at 70% avoidance? that would mean you have a solid 30% dodge and parry, which is impossible at this gear level.
I can get behind that, I suppose. My original point was that if we are simply considering a success or failure scenario (which improving TPS is most definitely not) that my statements would stand, but it would indeed seem you are correct here about the improvement of TPS not following the same rules.
That would present an interesting decision on the topic of gemming. While avoidance does get more valuable as you put more percentage into it, expertise goes up linearly, as do HP (right?).
Does anyone know if diminishing returns are similar for expertise/avoidance ratings?
there is no diminishing returns on expertise ratings.
Avoidance does become more valuable the more percentage of it you add, but because avoidance on gear isn't in the form of a percentage and is in fact in the form of ratings, which count for less and less as you get more, the value of avoidance actually stays relatively static as you increase it.
Again, going back to the 1-100 combat roll system (and either ignoring the impact of crits or pretending they've been averaged into our average hit damage), if you cause X threat when one of your attacks hits, your average threat per swing/attack would be X*hit% (out of 100 swings, hit% will land and [100-hit%] will not). In this case, hit% refers to the percent chance for your attack to land successfully (or 100% less miss, dodge, and parry). Increasing your chance for a successfully landed attack by 1% (either though hit rating or expertise rating) would add X*1%, or .01X, to your average threat per swing/attack, since it is adding one more successful die roll outcome out of the 100 possible outcomes.
It's not really necessary for the math here to calculate this as TPS, since TPS could theoretically be derived from average threat per landed attack using attack speed, GCD, ability cooldowns etc., but it would add a lot of needless complexity and would not change the fundamental .01X threat per hit% factor.
Edit: Err, everything from here on down is basically exactly what Dhalphir posted.
Well, I think the diminishing returns on avoidance ratings work against the increased value of each percent of avoidance as you have more, such that avoidance stats have approximately linearly increasing value as well, much like how armor is a linear increase.
There are no diminishing returns on accuracy stats, thankfully.
Is this a huge mistake on my part, and gear wise is there any significant change between them?
Gearwise, the hit cap is higher, but the expertise cap will be lower (if you take 2/2 WM). You prioritize crit slightly higher because you're in Battle instead of Berserker, but I'm not sure about the exact numbers to target.
Vene's been kinda dragging his feet on articles with valuable info, though.
Once you establish a safe margin over DPS while watching Omen, making sure that you don't move up over the MT, just keep TC and Demo Shout up and you will pretty much hover at a good level without doing much more.
Most people are shit at world of warcraft, and most DKs are worse than that, how much time did you give him to build threat?
I can safely say I needed about five seconds on that fight (as a paladin tank) to keep threat. Never had anyone pull from me, not even FFB mages.
As a OT I needed about, hrm, maybe slightly more time but not much.
It was painful, most fights were very slow due to a combination of using a single 2h in a built optimized for DW and the higher miss-chance from being low level. The spec was boring me and I figured if Titan's Grip wasn't a huge difference I was gonna go back to arms.
And holy shit what a difference it was. The spec went from boring and slow to a total blast to play. No wonder they're nerfing it next patch; it makes such a huge difference - perhaps less of one going from DWing 1hs to DWing 2hs, but still.
Also, as I've gotten fucking tired of XPerl, I found a mod that turns your UI into the Fallout 3 UI. I'm gonna test it for the raid tonight. Should be interesting.
I was under the impression that Omen now just uses the built-in threat API to create its meters. Is this not the case?
Omen has been using Blizzards built in threat values since they were released.
Omen is 100% accurate. More accurate, in fact, than the Blizzard one, as it shows threat values, not just percentage.
This I did not know. I thought it kept using it's own API.
Should I lrn2arms?