1. #5921
    Quote Originally Posted by Huntingbear_grimbatol View Post
    If you want to minmax dps and get ranks then go 70% crit and pray that your raid manages to kill it on one of your good attempts, crit has a higher potential dps than mastery does
    This is so wrong. How would you even write a nonsense like this?
    If you want to get ultra lucky, you need low crit rating and still crit a lot at important times on 1 great attempt. That way you get a lot of important crit while still having a lot of mastery for higher crits.
    Crit is a stat that doesn't increase your potential highest dps AT ALL. (For the exact same reason why it doesn't increase your worst case scenario.)


    Quote Originally Posted by Maegor View Post
    It absolutely blows my mind that we had a post that had probably the best explanation of what consistency means
    I take it you believe that worst case scenario is the best way to measure consistency then.
    So just to give you another example we should only use on-use trinkets. Since 25 % crit chance is so low we can't rely on it to do anything, I hope you agree that trinkets from Thok and Galakras are very weak as we can't hope to ever get them procced with only a 15 % chance.

    Say you have a 4set including legs and you get the 3-socket legs from Garosh non-warforged. Should you break your 4set to get that extra socket and higher socket bonus?
    In my opinion keeping the 4set results in higher consistent damage, because you will get at least 1 4set proc in a fight 99.999 % of the time resulting in higher damage.
    Yet the worst case scenario argument calls for replacing the legs because in the worst case scenario where you don't get a single 4set proc, having the 4set over one extra socket is wasted.

  2. #5922
    Deleted
    Quote Originally Posted by Huntingbear_grimbatol View Post
    If all you care about is "scumbagging" dps and getting that rank 1 on WoL to say "YES, I HAD RNG ON MY SIDE!" then that's your choice and I don't think that's wrong.
    Because that's what WoL ranking is isn't it? Doing the same thing over and over and just waiting for that 1 kill that makes everything line up perfectly.
    Not a fair way of representing ranking. Maximizing the amount of ranks, and how good they are, is about maximizing your average damage across many attempts of many bosses. I have not experienced my raid suffering as a result of my DPS swinging too much from side to side, and as such the consistency debate remains purely theoretical for me, that doesn't make it less interesting though.

  3. #5923
    Quote Originally Posted by Zardox View Post
    Not a fair way of representing ranking. Maximizing the amount of ranks, and how good they are, is about maximizing your average damage across many attempts of many bosses. I have not experienced my raid suffering as a result of my DPS swinging too much from side to side, and as such the consistency debate remains purely theoretical for me, that doesn't make it less interesting though.
    I know it's not fair, just making a point that you can play perfectly (or your entire raid plays perfectly) week after week but in the end the difference between a rank among top 10 and top 30 is most of the time down to your own RNG.

    - - - Updated - - -

    Quote Originally Posted by Neldarie View Post
    Kinda funny how people fail to understand that higher crit vs higher mastery is bigger consistancy in the long run for very obvious reasons. And that it is actually mastery stacking that has the highest dmg swing (from low to high). The ultimate low will always be lower for crit because mastery = free dmg from thin air (which is also exactly why Mastery stacking has the highest overall performance once in a bluemoon). But when talking about consistancy and the long run of things the probability of having higher crit %age will make your overall performance alot more stable (funny that my best siege belt performance was when I was reforged highest for crit).
    In the long run, crit will average out higher than mastery (don't think anyone is arguing that) but from 1 pull to another you can more reliably reproduce your numbers with mastery as opposed to crit. With mastery you increase the damage on most of your dps abilities all the time, on crit you increase the damage on all your damage sometimes. Thok bat spawns, one pull you get lucky and your 4set chainprocs like mad, the next pull it doesn't proc at all. I'd much rather have a set bonus that gave me the same increase at the same time every pull instead of one where you have to cross your fingers and pray/sacrifice goats for it to proc at the "right time" in a pull.

    I'm curious though, you say you did better crit on belt than mastery. Did you check each induvidual pull or even each induvidual time you went up on the belt and see if you produced the same number (or close enough) each and every time or did you just look at the end number after a night?

    - - - Updated - - -

    Quote Originally Posted by Neldarie View Post
    Bottomline for people who don't want to understand or read (those are mathematical definitions of them):
    Higher consistancy (smaller swing from low to high) = more crit.
    Higher potential (bigger e-peen? + making total fail rng scenarios not as painfull) = more mastery.
    The higher your % of crit is the smaller the "rng" becomes is true, but that costs mastery/strength or haste. Lowering any of those 3 and you're lowering your "worst case" dps, mastery (and strength) increases your "worst case scenario" damage. Crit doesn't do that, you can have 60% crit and still have a pull without a single TV or DS crit. And while it's true that you can somehow with godly rng (even though you didn't go for crit) manage to get crits at the right time produce insane numbers with a mastery build you don't slaughter your "worst case scenario" dps with a mastery build which you can end up doing with a crit build.
    9thorder.com | Recruiting exceptional players!

  4. #5924
    Deleted
    Quote Originally Posted by Huntingbear_grimbatol View Post
    I know it's not fair, just making a point that you can play perfectly (or your entire raid plays perfectly) week after week but in the end the difference between a rank among top 10 and top 30 is most of the time down to your own RNG.
    In my experience you don't need to rely on good RNG to be competitive in the overall rankings. I really like the revised PR point system, where going up from rank two to one gives five points instead of one, makes more sense.
    Last edited by mmoc451a784886; 2013-12-25 at 02:58 PM.

  5. #5925
    Quote Originally Posted by Meiffert View Post
    I take it you believe that worst case scenario is the best way to measure consistency then.
    So just to give you another example we should only use on-use trinkets. Since 25 % crit chance is so low we can't rely on it to do anything, I hope you agree that trinkets from Thok and Galakras are very weak as we can't hope to ever get them procced with only a 15 % chance.
    Yes... the reason we use thok and galakras trinkets is only because of the chance on hit effect... totally, wow.
    9thorder.com | Recruiting exceptional players!

  6. #5926
    Quote Originally Posted by Meiffert View Post
    Since 25 % crit chance is so low we can't rely on it to do anything, I hope you agree that trinkets from Thok and Galakras are very weak as we can't hope to ever get them procced with only a 15 % chance.
    We would be using those two trinkets even if they didn't have those strength procs.

  7. #5927
    Quote Originally Posted by Huntingbear_grimbatol View Post
    Yes... the reason we use thok and galakras trinkets is only because of the chance on hit effect... totally, wow.
    Yeah, I briefly considered not using this example because you could attack it in this way but I was naive enough to think that you might for once replay to the actual point rather than an offtopic formality.
    Just imagine that I talked about Ji-Kun's trinket instead. The hit was wasted because of being over the hit cap for many players yet it was still considered to be the strongest trinket in the tier just because of the random proc.


    Let me just sumarize the discussion from last 2 pages or so.
    Huntingbear_grimbatol made 2 claims in one of his posts, these being
    1. Consistency is very important, it is more important to have consistent damage than to have higher average damage.
    2. Adding more mastery rating makes your damage consistent, adding more crit rating makes it inconsistent.

    As far as 1. goes, I think there are situations where being consistent is better, there might be other extreme situations where it isn't so good. Overall though I agree that consistent damage is better way more often that not.
    That being said I would personally only sacriface small amout of average dps for a decent increase in consistency. I just don't value it as highly myself, but this is more of a personal preference.


    Which brings us to the 2. point, where the discussion went something like this:
    Huntingbear_grimbatol: Crit is inconsistent, mastery is consistent.
    Me: How much more consistent is it? Do you have any evidence for your claim? (As a side note I showed that mastery actually increases inconsistency, but maybe not as much as crit - I don't know.)
    etsumii92: Mastery is flat increase to some abilities and crit is random on all abilities.
    Me: You always have some crit bringing randomness to the picture, that doesn't mean having more crit makes it more inconsistent. (I proved that in some extreme cases even the opposite is true - getting more crit makes you more consistent if you are close to 100 % crit. I do understand that this is unrealistic example, I only showed it to disprove the false assumption that more crit always means more inconsistency.)
    etsumii92: Mastery is static and crit is inconsistent and that is a fact.
    Meiffert: Repeating something over and over doesn't make it a fact. Mastery is not consistent (more on that below).
    paonai: In extreme situation which never really happens (when you never crit), mastery is more consistent. While this is true and paonai did great job proving his claim, unfortunately it is just as irrelevant as my 100 % crit example, these extremes simply don't happen.
    Huntingbear_grimbatol: Having higher crit gives you higher maximum (higher potential dps).
    Meiffert: Now I'm sorry, but this is simply the exact opposite of how it works and posting this after we discussed these things in quite a few posts unfortunately shows that he has no idea what we are even talking about.
    Neldarie: Crit is more consistent than mastery. (Using some similar arguments to mine worded a bit differently, unfortunately not providing strong enough proof.)
    Huntingbear_grimbatol: Mastery gives you better worst case scenario. (That is true, Neldarie also said it himself as did paonai who came with this idea in the first place. As I already said though, worst case scenario isn't good enough metric for consistecy, because it simply doesn't happen.)


    I have to say that this discussion is getting rather tiresome since there are a few people repeating "mastery is consistent and more crit always means more incosistency" over and over again, although both of these claims have been already disproved by counterexamples; and calling me wrong.
    How can I even be wrong when I haven't even stated my opinion on the matter? If you read my posts, all I did was point out some mistakes in reasoning or false assumptions in other people's posts and asked you several times to provide proof or evidence for *your* claim.



    Let me say what I actually think about the consistency of mastery and crit (you can at least call me wrong after this post ).

    First of all, getting more mastery will increase your inconsistency. It will increase the difference between good and bad pull not only in absolute numbers (since both will be higher) but even in relative.
    Let me give you a very simple example, when on one pull we don't get any procs, on another we get one 4set proc and one DP proc. The ablities usage could look something like this:
    1. CS-J-Exo-CS-TV-J-CS-TV-empty-CS-J-Exo
    2. CS-J-Exo-CS-TV-J-CS-DS-TV-CS-J-TV
    Lets assume that the ablities do this damage (without mastery): CS, J and Exo 50, TV and DS 100.

    Now with mastery adding 50 % more damage to CS, TV and DS (increasing their damage to 75, 150 and 150 respectively).
    The first pull we did 75+50+50+75+150+50+75+150+0+75+50+50 = 850 damage.
    The second pull we did 75+50+50+75+150+50+75+150+150+75+50+150 = 1100 damage. (About 29.4 % higher damage with better luck on procs.)

    Now we add some more mastery and get 100 % increase to our CS, TV and DS damage (their damage with mastery is 100, 200 and 200).
    The first pull we do 100+50+50+100+200+50+100+200+0+100+50+50 = 1050
    The second pull we do 100+50+50+100+200+50+100+200+200+100+50+200 = 1400 (About 33.3 % higher damage with better luck on procs.)

    As you can see, with more mastery, the difference between good and bad pull is higher compared to lower mastery setup. This shows that mastery is indeed one of the stats that increases inconsistency.


    Now as far as crit rating goes it is a little bit more complicated, mostly because while my math skills overall are pretty decent, I always avoided statistics like a plague.
    What I think (and I might be wrong) is comming from this observation. Crit is very consistent when we are very close to either 0 or 100 %. If we get all the way to one of these extremes, all the randomness from crit disappears completely. Since critting with 20 % chance is statistically the same event as not critting with 80 %, the curve (if we made a graph) of inconsistency will be symetric (crit has the same consistency at 20 and 80 %, the same at 40 and 60 % etc.) And since I can't think about any reason why there should be any local extreme (minimum or maximum) of inconsistency other than 50 %, I believe that crit becomes more inconsistent as we increase the value of crit rating until we reach 50 % crit chance. After that point adding more crit should actually make it more consistent.


    Unfortunately I am unable to draw many conclusions from all of this because in most realistic setups we will probably under 50 % crit meaning that more crit rating will increase the inconsistency of our damage and more mastery will increase it as well.
    I am unable to tell which of the two stats will increase the inconsistency more at this point, personally I would recommend simply going for the higher average dps setup until we get more information.


    As a sidenote for the often cited example of siegecrafter belts, the best secondary stat to go for is haste all the way up to 50 % for these reasons:
    a) it gives you the highest consistency of all the secondary stats (you get more hits which means you get faster closer to you average amout of crits/procs)
    b) haste itself as a stat doesn't directly increase your inconsistency like crit and mastery do as far as I can tell
    c) haste is the strongest stat outside of heroism/bloodlust in most gear setups, the reason many people stop at around 40 % is that more haste is mostly wasted (you get a few more autoattacks in and that's pretty much it) during heroism/bloodlust where we do a decent part of our overall damage on the fight; yet you have to be able to kill a weapon on the belt without heroism of course

  8. #5928
    Here's what some people are missing. Mastery and strength (and haste to some extent) increases the lowpoint of your potential dps which is what you should aim to do (imo), while you can be insanely lucky and get crit after crit after crit with mastery build and do insanely stronger than you would normally you don't rely on that to happen to reach your estimated damage number. If you get crit over mastery you're putting "more eggs" into the "basket of please for the love of god crit" at the cost of mastery (or str/haste) the higher your crit rating / % is the higher the chance of success becomes for your build, yes. But you still have to lower mastery and/or other stats to get the higher chance to crit. You don't say you can do 400k dps if... or say you can do 400k dps on average, no way in hell I'd pick someone to do a job that they couldn't do all the time.

    I don't know if this was missed.
    Worst case scenario dps for mastery is higher than the worst case scenario for crit. (not by a insane amount but still...)
    Lets just put down some imaginary numbers shall we?
    100k dps. then say you get a 20% increase from mastery to 120k dps. (for the sake of argument we never get any crits in this mastery prio thingy)
    With the ups and downs of trinket procs, enchant procs, encounters and such let's say you manage to stay at 95% of the 120k dps point on average (95% will be for other one as well, not important)
    Then Crit.
    100k dps but no increase from mastery, let's say you get to 50% crit rating. Over 10000000000 tries you'll average close to that 50% crit rating and end up at a (theoretical) 150k dps, that's on average 30k dps higher than the mastery prio. But that does NOT change the fact that your low end / worst case scenario can be 100k or 20k lower than the mastery way, yes on average you'll do better but do you honestly want to be stuck one try not being able to crit on anything because RNG is RNG?

    Nobody is arguing the ups and downs about crit or mastery, everyone knows (or looks like it anyways) that crit on average is higher and mastery is more stable. The "argument" as far as I'm concerned is which of the two is "best suited" for raiding (progress in my mind), some value being higher (on average) while some (me) prefer being consistent from one pull to the next - or more in detail one belt jump to the next which mastery does better than crit.
    9thorder.com | Recruiting exceptional players!

  9. #5929
    Quote Originally Posted by Huntingbear_grimbatol View Post
    Worst case scenario dps for mastery is higher than the worst case scenario for crit. (not by a insane amount but still...)
    Yes. But worst case scenario never happens. And I mean it is so unlikely that it literally never ever happens even once in your entire life.
    Therefore thinking about worst case scenario is NOT a good measure of consistency.

    Quote Originally Posted by Huntingbear_grimbatol View Post
    everyone knows (or looks like it anyways) that crit on average is higher and mastery is more stable.
    No. This is exactly where you are wrong every time.
    We do NOT know that mastery is more stable, this entire discussion we are having is about trying to figure out IF mastery is more stable. So far we don't know. Some people are saying mastery is more stable (you are one of them), some (such as Neldarie) are saying crit is more stable.
    My personal belief is that their stability (or consistency) is very similar below 50 % crit (which is where we are).
    Did you even read my previous post?

  10. #5930
    It's like using the dumb assumption of Cheetah never catching up to turtle in the race. Turtle is ahead by default and whenever Cheetah reaches the point of where the Turtle was; Turtle has moved on further abit, when Cheetah gets to the point number2; Turtle has moved some yet again.
    Thats the very similar loophole in the reasoning of people stating that Mastery is more consistant than Crit. Turtle/Cheetah example = Time slows down. Mastery vs Crit debate people focus all their attention to extreme lows instead of looking at the entire picture.
    Last edited by Neldarie; 2013-12-26 at 01:00 AM.

  11. #5931
    Ok, so I ran 2 sims on my toon, one reforged towards mastery, one to crit. Both builds were at 40.8% haste. In only reforging my items I moved 4,608 rating. My base crit in the crit build got to 21.22% - meaning I would have 36.22% crit in raid.

    Mastery Build:


    Crit Build:


    Couple things to point out. First average simdps was higher on the mastery build, which was surprising (and also kind of puts an end to the discussion right?). Also, the worst case and best case dps was higher on the mastery build.

    Now we can all see that the dps histograms in sim have positive skews, which means mode<median<mean. This is why it is hard to hit your simdps (most of your attempts will be at the mode). Looking between the histograms you can see that the mastery build has a more symmetrical shape (the hump is closer to the middle), indicating that the mean is closer to the mode. This is consistency.

    Now, the question is - is there a crit rating where this changes? If we were to be able to get to 50% crit, or 60% crit, would we a) see a higher simdps and b)would the skew of the histogram be more symmetrical?

    My personal question (which ties into the previous questions) is this similar to what heroically geared players are seeing?
    Maegore @Maegoree Maegor#1377

  12. #5932
    I'd run the sim for a much longer duration, you could also sim by using BiS list gear or swapping out mastery / haste items with crit / haste items. Also if you're terrible bores run simc for 20 sec repeatedly and you'll find an answer to siegecrafter belt situations.
    9thorder.com | Recruiting exceptional players!

  13. #5933
    Deleted
    So how much difference in dps between crit build / mastery build are we talking about, concerning BiS gear level? People keep telling that the difference is marginal, is it in the 100s, or 1000s, or more? Would it be even viable to go all out crit instead of mastery, or would it be complete bs?

  14. #5934
    If you look at the difference between my character, 563 ilvl, the difference was 2k. Ramp that up to BiS, and 4-5k max I would guess.
    Maegore @Maegoree Maegor#1377

  15. #5935
    Deleted
    I simulated a haste > mastery > crit build, a haste > crit > mastery build and a haste > mastery = crit build (1,5-2k more crit) for my gear (577 - max haste). The difference was like 400 dps.
    haste > mastery > crit: 368176.9
    haste > crit > mastery: 367915.2
    haste > mastery = crit (1,5-2k more crit): 368313.0
    Not that much - 0,04% dps gain over haste > mastery > crit.

  16. #5936
    Its the problem of having very low (vs bis) ilvl on your gear. And its also a case of getting more and more crit devalues its value while pushing mastery higher. I've advocated the crit-mastery balance many pages back (by running sims 10k + its and reforging accordingly untill crit=mastery for you) thus having the most out of each. Generally I found that in best possible gear its around 16-17% (maybe 15-16 with more haste on gear) unbuffed crit max and after that mastery pulls ahead again.

  17. #5937
    Quote Originally Posted by Neldarie View Post
    Its the problem of having very low (vs bis) ilvl on your gear. And its also a case of getting more and more crit devalues its value while pushing mastery higher. I've advocated the crit-mastery balance many pages back (by running sims 10k + its and reforging accordingly untill crit=mastery for you) thus having the most out of each. Generally I found that in best possible gear its around 16-17% (maybe 15-16 with more haste on gear) unbuffed crit max and after that mastery pulls ahead again.
    Just to be clear on what you mean here, you are talking about having stat weights be equal between mastery and crit, correct? If so, I agree with you there. I think one of the contributing factors to the skews on the histograms my sims generated, was that on the crit build, mastery had a much larger weight than crit. I would suspect that anytime you have one stat weigh so much more heavily than another, this would only exaggerate the skew.

    I think eliminating as much of the skew as possible is the key to getting your actual dps closer to your simdps on a consistent basis.
    Maegore @Maegoree Maegor#1377

  18. #5938
    Quote Originally Posted by Maegor View Post
    Just to be clear on what you mean here, you are talking about having stat weights be equal between mastery and crit, correct?
    Yes, strength, haste, mastery and crit all increase the value of each other. Therefore if you blindly stack one of these, its value will go down.
    That is also the reason why mastery was straight up stronger than crit for you in your setup. As you get higher ilvl stacking more mastery, crit will pull ahead. Another reason for that is your Thoks trinket is low ilvl. It increases the value of crit by making your crits stronger as well as indirectly increasing it's value by giving you more mastery and haste.


    I looked in Simcraft and it actually shows you some form of consistency. What it does is discard 5 % best results and 5 % worst results and calculating the difference between your best and worst pulls from the remaining 90 %. It also then divides this number by 2 so the result shows how far from the middle are your very bad and very good pulls.

    I then simmed my own character (the default is haste > mastery > crit setup in 568 ilvl) and then simmed how would the results change if I had 3k more of one of the secondary stats.
    The results show average dps, how far from middle are your bad and good pulls (consistency), and the stat weights of haste, mastery and crit.

    Default setup
    dps: 340016
    range: 17308 (5.09 %)
    haste: 5.29
    mastery: 4.96
    crit: 4.98

    Mastery setup (+3k mastery)
    dps: 354791
    range: 18302 (5.16 %)
    haste: 5.39
    mastery: 4.91
    crit: 5.14

    Crit setup (+3k crit)
    dps: 354864
    range: 18096 (5.10 %)
    haste: 5.40
    mastery: 5.12
    crit: 4.87

    Haste setup (+3k haste)
    dps: 354672
    range: 17633 (4.97 %)
    haste: 3.24
    mastery: 5.09
    crit: 5.14


    As you can see, since all stat weights are very close in value, getting 3k more of any of them gives me almost the same dps increase.
    Also having more of one of them increases the relative value of the other 2 as we would expect.

    As I predicted, haste gives us more consistent damage (decreasing the difference between average and bad pull from 5.09 % to 4.97 %), while both mastery and crit made the damage more inconsistent (increasing the difference between average and bad pull to 5.16 and 5.10 % respectively).
    Mastery seems to be a little worse using this metric (it would help if some of you could do this experiment with your characters to get more data).

    Still, even though I think this metric used in simcraft is better than worst case scenario metric, it is not really ideal.
    I would prefer a metric that does something like this: ignore 1 % of worst results, then use the lowest dps from the other 99 % of results. This would eliminate the "almost never happens" absolute worst case scenarios but still focus only on the bad pulls in which we still want to be able to meet the dps checks.
    If someone has an idea how to find this number, please share.

  19. #5939
    What are your haste/mastery/crit values in your default setup? Seems awkward that you are swinging so far from default, but getting close numbers in the "extreme" (for lack of a better word) builds.

    And yeah, I agree that a better thok's will cause my crit build to sim higher than it did, since TTT is the key to crit being viable in the first place. My question would be - would a better version of TTT reduce the skew of the crit build or would the skew be close to what it was above? If the skew doesn't change then the mastery build is still more consistent since the mode is closer to the mean. I really wish simc would output the mode and median along with the mean, since that, IMO, is the correct way to measure real world variance.

    The only thing I can think regarding the throwing out of the high and low 5% is they are trying to control the data by disregarding values outside of a factor of std dev. Which in a bell curve, the high and low 5% would be close to everything outside of 1.8sigma.
    Maegore @Maegoree Maegor#1377

  20. #5940
    Quote Originally Posted by Maegor View Post
    What are your haste/mastery/crit values in your default setup? Seems awkward that you are swinging so far from default, but getting close numbers in the "extreme" (for lack of a better word) builds.
    http://eu.battle.net/wow/en/characte...ffert/advanced

    Quote Originally Posted by Maegor View Post
    My question would be - would a better version of TTT reduce the skew of the crit build or would the skew be close to what it was above? If the skew doesn't change then the mastery build is still more consistent since the mode is closer to the mean. I really wish simc would output the mode and median along with the mean, since that, IMO, is the correct way to measure real world variance..
    Why do you think that higher difference between mode and mean means more inconsistency?

    Let's look at these 2 sets of 10 numbers.
    a) 9, 9, 9, 10, 10, 10, 10, 11, 11, 11
    b) 6, 7, 8, 9, 10, 10, 11, 12, 13, 14

    Both these sets have mean, mode and median equal to 10. Yet a) is a lot more consistent for our purposes.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •