+ Reply to Thread
Page 2 of 2 FirstFirst 1 2
Results 21 to 36 of 36

Thread: Dodge vs Parry

  1. #21
    Join Date
    Feb 2008
    Location
    Aarhus, Denmark
    Posts
    73
    Quote Originally Posted by Ioseb View Post
    I have a DK who keeps telling me that parry is good for a host of crazy reasons. Mostly I think it's a lack of understanding of how DR works, but then he says crazy things like "Parry improves my threat".
    Is there any truth to this threat argument or can it's effect be seen as negligible in most/all models?
    This seems to be a very common misconception (unfortunetly)...
    I had this issue for quite some time as a newly startet DK tank in the guild i tank in... whenever a plate item with heavy parry dropped everyone pointet at me... Took some time to explain that the world did not evolve around parry for a DK...

    The thing about DK's and parry is that we have alot without even thinking about it.

    The only connection that a items with parry on and a DK item, is that items without "block rating" and "block value" just so happens to have parry on them in most cases.

  2. #22
    Join Date
    Oct 2008
    Posts
    4,930
    Technically, though I haven't had a chance to test it well, players do get parry hasted as well. That would represent a threat buff. There is a reason parry rating has a smaller conversion value to % than parry.

    It makes me sad that Naxx25 has as much or more block value/rating plate tanking gear than other pieces. =(
    The (Old) Book on Death Knight Tanking
    The New Testament on Death Knight Tanking
    -----------------------------------------
    Quote Originally Posted by Horacio View Post
    Who f-ing divided by zero?!?

  3. #23
    Join Date
    Jan 2008
    Location
    Yardley, PA
    Posts
    1,222
    Stacking parry for threat is wrong.

    There are plenty of other ways to gear and spec a DK that are less expensive, and get more returns.

    If you want avoidance, continue to stack defense. If you need a red gem slot and want some kind of avoidance out of it, put an expertise gem in.
    Luck is what happens when preparation meets opportunity. -Seneca
    Everyone marvels at a square egg, but only the chicken understands the PAIN.

  4. #24
    Join Date
    Oct 2008
    Posts
    4,930
    Agreed, though I've been meaning to do some rough gauging to see if I can figure out the parry-haste buff. I wouldn't gear for it, but it may actually be work out to be a decent buff to our threat as our gear improves, one I hadn't accounted for in the mix before.
    The (Old) Book on Death Knight Tanking
    The New Testament on Death Knight Tanking
    -----------------------------------------
    Quote Originally Posted by Horacio View Post
    Who f-ing divided by zero?!?

  5. #25
    Join Date
    Jul 2007
    Location
    Canadia
    Posts
    3,523
    Something here doesn't quite sit right with me, but I don't have time to look it over again. Someone will spot the problem, I am sure =)

    Napkin math: Assuming 0.24 extra swings per parry is correct, let's call it 4 parries to earn one extra swing. The boss attacks every 2 seconds.

    At 25% parry, we're looking at an average 16 swings to get 4 parries. That's an extra swing every 32 seconds, or 1.875 extra swings per minute. We gain (1.875 * 2.07 * D / 60) = 0.065D TPS - if your white damage swings average 350, that's about 22TPS

    30 % -> 0.078D (27 TPS)
    40 % -> 0.104D (36 TPS)
    47 % -> 0.122D (43 TPS)

    If I happen to be in an unlimited rage situation and can make all of those white damage attacks into Heroic Strikes, the threat output increases from 350 damage to 350 damage + 495 bonus damage + 259 innate threat. Average threat dealt is 1104, 3.15 times the amount of threat from a simple white damage swing. That increases the TPS gain at 25% parry to 72TPS and at 47% to 134TPS.

    That's all based on averages of course. You could get 4 parries in a row, giving you an extra swing in 8 seconds. Still only a blip on the TPS radar.

    Generalised
    1) Assuming 4 parries to get one extra swing: T = 0.25PDM/S , where T is the TPS gain, P is parry chance, S is boss swing speed, D is average weapon damage, and M multiplier for threat

    2) Roll your own: T = EPDM/S , where T is the TPS gain, E is "extra swings per parry", P is parry chance, S is boss swing speed, D is average weapon damage, and M multiplier for threat
    Last edited by Satrina; 02-04-2009 at 08:45 AM.
    Got a question? Try here: Evil Empire Guides and here: Tankspot Guides and Articles Library first!

  6. #26
    Join Date
    Feb 2009
    Posts
    12
    Quote Originally Posted by Croma View Post
    As Satrina says, the model in WoW is table based, so a roll to 'hit' could look like:

    0000-0900 : miss
    0901-2900 : dodge
    2901-4700 : parry
    4701-9999 : hit

    you dont calculate each avoidance seperately, in any order.

    ok, then if that is true my calculations for for nothing ><

    from my understanding i thought that each "event" was not related to each other as galushi said, they are all independent and each one does not rely on the other.

    basically each attack you take has a chance to miss, dodge, or parry.
    for each hit either none or all of the events can occur however the only one that is shown is the first event that takes place.

    so if an attack is dodged and parried, only dodge is shown assuming that the dodge check precedes the parry check.

    or at least.. that was how i thought it worked.

  7. #27
    Join Date
    Jul 2007
    Location
    Canadia
    Posts
    3,523
    Edit: They mutually exclusive.

    If you didn't read yet: http://www.tankspot.com/forums/f63/3...bat-table.html. Also, this one: http://www.tankspot.com/forums/f111/...-analysis.html

    They key to deciphering the puzzle was this, from Blizzard:
    [blizzard]The way WoW calculates crit rate is over ALL attacks. Crit rate is not based on hits only. In other words, if you have a 5% crit rate, that 5% chance includes misses. [/blizzard]

    They've since given up trying to keep the mystery and confirmed it's table based.
    Last edited by Satrina; 02-04-2009 at 04:56 PM.
    Got a question? Try here: Evil Empire Guides and here: Tankspot Guides and Articles Library first!

  8. #28
    Join Date
    Feb 2009
    Posts
    12
    ok, now I'm confused.

    From my understanding for an event to be independent would mean that it is NOT mutually exclusive.

  9. #29
    Join Date
    Jul 2007
    Location
    Canadia
    Posts
    3,523
    If the table is {miss, dodge, parry, block, hit}, you have 5 possible outcomes on one random check. Each one is mutually exclusive - if you parried, you were not missed, did not dodge, did not block, did not get hit.

    I'm an engineer, not a mathematician. My terminology may not be 100% correct in math-world =)

    Edit: Ah yeah, a little Google shows me where your confusion is, and it's due to me not being primarily a math person. Events on the table are mutually exclusive, by mathematic definition.
    Last edited by Satrina; 02-04-2009 at 04:58 PM.
    Got a question? Try here: Evil Empire Guides and here: Tankspot Guides and Articles Library first!

  10. #30
    Join Date
    Oct 2008
    Posts
    4,930
    /highfive Satrina

    I knew there was a reason your math/explanations made good sense to me. Familiar flavor. Engineers unite!
    The (Old) Book on Death Knight Tanking
    The New Testament on Death Knight Tanking
    -----------------------------------------
    Quote Originally Posted by Horacio View Post
    Who f-ing divided by zero?!?

  11. #31
    Join Date
    Jul 2007
    Posts
    16,409
    Software engineer's okay too?

    /dances

    READ THIS: Posting & Chat Rules
    Quote Originally Posted by Turelliax View Post
    I will never be a kaz.. no one can reach the utter awesomeness of you.
    http://i.imgur.com/3vbQi.gif

  12. #32
    WARNING THREAD HIJACK POSSIBLE

    Quote Originally Posted by Satrina View Post
    I'd imagine WoW uses a cryptographic RNG anyway, so the answer is most likely yes. (For all we know, they use a hardware RNG on each server)

    However, if they were using a simple pseudorandom RNG... If you take a PRNG and run its sequences for N consumers with thousands of consumptions per second in a completely random order - is each consumer's result still pseudorandom, or does the pseudorandom only apply over the sequence and each consumer's results are now truly random?

    I did some thinking on this, and due to the nature of the "linearity" of the system that is picking which consumer goes when. (I'm trying to convey the idea that the result passed through F(c) (function of consumer) is linear) I believe that if your base RNG is psuedo, then no matter the sample space you will never truly be random. That is to say, assuming that a psuedo random number generator does not have a set "error" (aka you know when it's going to make a mistake) but instead has a gaussian distrobution of when it is not truly random, then no matter what consumer gets what random number when, their possibility of getting a not truely random number is still the same.


    But there are so many assumtions in the above that we are playing in hypothisis land and have no real way to test if it's true. For all I know the server doesnt even generate the random number, the client does. I'd like to see a blue post on how they do this.. lol. (watch they are using diode noise on each server for their RNG and thus it's about as perfectly random as you can get... just to bite us in the arse).

  13. #33
    Join Date
    Jul 2007
    Posts
    16,409
    i hate talking about the RNG, lol i did an entire class on showing how you can never be truly random and how you can try but you can't. Man... memories.

    READ THIS: Posting & Chat Rules
    Quote Originally Posted by Turelliax View Post
    I will never be a kaz.. no one can reach the utter awesomeness of you.
    http://i.imgur.com/3vbQi.gif

  14. #34
    Join Date
    Jul 2007
    Location
    Canadia
    Posts
    3,523
    The client can't possibly be doing the random generating, that's an enormous security hole waiting to be exploited. If it were me, I'd just put a hardware RNG (diode noise, cesium decay, laser, what have you) that just contnually feeds a queue on each box in the server cluster and be done with.
    Got a question? Try here: Evil Empire Guides and here: Tankspot Guides and Articles Library first!

  15. #35
    Join Date
    Feb 2008
    Posts
    615
    I wouldn't be at all surprised if they just use a well-known "good" but simple PRNG. It's likely to be far faster and cheaper than a good cryptographic PRNG or a true RNG. And, given the large number of random values being pulled out by many many people all the time, even if it's theoretically possible to find a cycle, you're never *ever* going to be able to exploit it.

    So as long as they just make sure not to use a system that's known to produce really bad cycles, that's fine.
    Learn to science and stop theorycrapping in its tracks.

  16. #36
    Join Date
    Feb 2008
    Location
    Aarhus, Denmark
    Posts
    73
    Quote Originally Posted by Kazeyonoma View Post
    i hate talking about the RNG, lol i did an entire class on showing how you can never be truly random and how you can try but you can't. Man... memories.
    So true, and so wrong... This all depends on what our difinition of random is...

    Even if you where to use things as atmospheric noise, light sources, particles and what so ever. Many would argue that it is not random.

    For me, theoretically no, however since such pulls is working on so many variables (where by far billions of them are "unknown" to us), we may never be able to predict what comes out... (Hell if we got that far we would proberly be able to forsee the future...) and thats good enough for me to call it truely random.

    But i Honestly don't think Wow runs on a TRNG... for all that matters in this case, a PRNG is far enough.

+ Reply to Thread

Bookmarks

Posting Permissions

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