We have detected that cookies are not enabled on your browser. Please enable cookies to ensure the proper experience.
Page 3 of 3 FirstFirst 1 2 3
Results 51 to 73 of 73
  1. #51
    Join Date
    Jun 2011
    Posts
    8,354
    Quote Originally Posted by TiberiasKirk View Post
    You are afraid of what they would find. If you believed there was a way to prove me wrong it would have already happened. I welcome ssg looking at this.
    There is a probability I am wrong, but it's less than 1% based on the sample size I have (1000) and the fluctuation it would take to make a 5% change in partial avoidance appear when it really isn't supposed to be there.
    I already did on earlier post.

  2. #52
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by siipperi View Post
    I already did on earlier post.
    Care to assign a probability your answer is correct?

  3. #53
    Join Date
    Jun 2011
    Location
    UK
    Posts
    1,656
    Were you the guy in Leroy's group before he went mental?

    Mac
    See my Lotro gaming videos:
    Macdui on Youtube.

  4. #54

    Armageddon II

    Wait, back up....

    Quote Originally Posted by TiberiasKirk View Post

    I conduct experiments and statistical tests for the United States Air Force. My space defense work helps defend satellites from collisions and everyone on the planet including you from asteroid impacts. You better hope I can run a statistical test properly.
    Holy shmoley - you're protecting us from asteroid impact with statistical tests!! That is so awesome - they should make a movie. Maybe Bruce Willis can stop a giant asteroid from annihilating the earth by running a paired t-test.

  5. #55
    Join Date
    May 2017
    Posts
    307
    Due to the limitations of 5 images per post, I will be forced to make multiple posts in regards to the testing done.

    But for starters, the trait builds I will be using are Disorientation only and No Disorientation. Linked here is my common DPS set up on Bullroarer Test Server, with what I consider is optimal Finesse rating for a melee class. If needed, I will adjust finesse rating as people see fit, though I imagine since we're only testing partials, it wouldn't be needed, as in fact it could make things worse for the case of applying Disorientation debuff. I will also make 3 parses of each, try to clip them all into one image and post here for people themselves to see the results and deduct what they wish. Numbers can't lie.
    Only the following skills be used:
    - Guardians Ward
    - Sting
    - Stagger
    - Vexing Blow
    - Shield Blow
    - Shield Swipe
    - Bash
    - Shield-smash
    The manner in which the skills are listed is the rotation to follow, starting with Shield Blow to open a block chain. This rotation is a common trait in all tests.
    Additionally, if anyone feels like they want to assist me in this, have suggestions or want to contribute in any way, feel free to PM me and we'll talk through.

    Last edited by Decrepify; May 17 2018 at 01:32 AM.

  6. #56
    Join Date
    May 2017
    Posts
    307

    Initial numbers are in

    So, I have decided to extend my testing from 3 parses per trait build to 11, almost quadrupling the testing. This was done for multiple reasons, mainly because it would seem that BR dummies cannot (partially) block nor parry, and only partial evades were triggering. Unstripping my testing finesse to as low of a value as possible reveals that, unlike in testing parses, dummies can also fully evade as well as partially evade, but still cannot block or parry in any sense, with evidence parse presented here. At least it's good to know I hit a good spot with my finesse rating

    Now, the testing parses....
    I present tested data in 2 albums, each titled with what they represent, under the circumstances listed above, and repeated below:
    Quote Originally Posted by Decrepify View Post
    ...for starters, the trait builds I will be using are Disorientation only and No Disorientation. Linked here is my common DPS set up on Bullroarer Test Server, with what I consider is optimal Finesse rating for a melee class. If needed, I will adjust finesse rating as people see fit, though I imagine since we're only testing partials, it wouldn't be needed, as in fact it could make things worse for the case of applying Disorientation debuff. I will also make 11 parses of each, try to clip them all into one image and post here for people themselves to see the results and deduct what they wish. Numbers can't lie.
    Only the following skills be used:
    - Guardians Ward
    - Sting
    - Stagger
    - Vexing Blow
    - Shield Blow
    - Shield Swipe
    - Bash
    - Shield-smash
    The manner in which the skills are listed is the rotation to follow, starting with Shield Blow to open a block chain. This rotation is a common trait in all tests.
    First album contains testing parses without Disorientras....Disorientation trait skilled.
    Now, I'm not claiming I'm Einstein, but I'll just go with laymans math here, totalling all the partial evades and hits, dividing former with the latter and multiplying it all by 100, effectively giving a flat out average % of how many attacks were partially evaded in the parses.
    In the first testing, total number of attacks made was 2970. Total number of partial evades was 187.
    (187/2970)x100=6.2962% approx.

    Second album contains testing parses with Disorientation trait skilled.
    Same simple math being used as well.
    In the second testing, total number of attacks made was 2960. Total number of partial evades was 218.
    (218/2960)x100=7.3648% approx.

    Theory goes that Disorientation is supposed to affect partials like it does for full avoidances. Well, numbers in this case, prove someone contrary, it actually increased the avoided attacks by 1.1%.
    Wait, does this mean Disorientation is bugged? Well, no. Not really. We cannot test partial blocks and evades in this case, so numbers could be lower...but at the same time they could higher as well. I believe the test sample was large enough to cover it for now, pretty much proving that, at the very least, Disorientation doesn't reduce partial evasion avoidance. If one desires, they can extend this logic to others as well, or people can use their own maths to figure out. I have posted the parses and data, with detailed enough description of how I've got the said numbers.

    If there are any questions, please ask ahead.

    P.A.Q. (Potentially asked questions):

    Q:Why do some parses have more than 270 attacks, and yet some less?
    A:Lag. Plain and simple. Server was a bit jaunty for me, so I did this as best I could.

    Q:How can we tell that the partial evades weren't only the auto-attacks that happen between skill usage?
    A:Great question. Here is a perfect example mid-parse of a raw combat log where 3 skills partially missed.

    Q:There are some "Immune" numbers in the middle of the parses. Care to explain?
    A:Certainly. Apparently, testing dummies can be stunned...oddly enough, during which they receive same avoidance nullification as any other mob.
    Q:Won't that interfere with the testing?
    A:I firmly believe it shouldn't. In an actual combat, Guardian will cycle through all of its skills in a rotation, resulting in the same effect.

    That's all for now in my report.

  7. #57
    Join Date
    May 2017
    Posts
    307
    Well then, I take it we're done here. Due to lack of response from the OP, I can take it that he has either lost interest to defend his own theory/practice or he doesn't wish to admit the truth about the trait.

    So, to any new Guardians looking over for builds and traits, do not use Disorientation. It's not worth it in it's current form, and there are better traits all around.

  8. #58
    Join Date
    Jun 2011
    Posts
    4,651
    After reading this thread, I began finding myself involunitarily looking up at the sky every now and then, expecting meteorites to come crashing down on us from space.

  9. #59
    Join Date
    Jan 2010
    Posts
    607

    Disorientation Broken on Test Server

    Quote Originally Posted by Decrepify View Post
    Well then, I take it we're done here. Due to lack of response from the OP, I can take it that he has either lost interest to defend his own theory/practice or he doesn't wish to admit the truth about the trait.

    So, to any new Guardians looking over for builds and traits, do not use Disorientation. It's not worth it in it's current form, and there are better traits all around.
    I posted a new analysis based on the bullroarer test server data I had which shows Dissorientation completely broken on the test server.
    My analysis is against targets that can actually parry and evade (LOL) and contains error bars showing the statistical significance of the results. Your results are not professional at all. They show the lack of partial avoidance and any full avoidance gain, but instead of delving deeper to see what is wrong on the test server, you have no intention of making the game better and would be more than happy if they didn't fix the disorientation trait.

    Quote Originally Posted by TiberiasKirk View Post
    Had time to kill about 50 wolves out in the dale lands while the test server was up this week.

    Killed about equal numbers using a rotation of [Guardian's ward, shield blow, shield swipe, bash and shield smash].
    to trigger disorientation

    Melee auto attacks occurred in large numbers spread throughout the rotation (about 400 total over the 25 wolves for each case).

    Chart below show how DISORIENTATION ACTUALLY INCREASED FULL AVOIDANCE CHANCE instead of decreasing it (calculated by taking avoided auto attacks and then dividing by the total number of auto attacks not including missed attacks which there were very few anyway)



    The difference between the average is much greater than the margin of error depicted by the min and max values for the two cases. The increase in full avoidance for using disorientation on the test server is statistically significant.

    Now looking at partial avoidance, nothing changes. The difference between the average partial avoidance with and without disorientation is well within the margin of error represented by the min and max values that are +/- one standard deviation below and above the average values as seen on the chart below.



    chat logs for the two cases are available for download and represent the raw data used in the analysis below. Files are titled to indicate whether disorientation was traited or not.
    https://drive.google.com/file/d/14nz...ew?usp=sharing
    https://drive.google.com/file/d/1u96...ew?usp=sharing

  10. #60
    Quote Originally Posted by TiberiasKirk View Post
    My analysis is against targets that can actually parry and evade (LOL) and contains error bars showing the statistical significance of the results.
    I'm curious; How are you calculating the standard deviation on the targets avoidance when you don't know the targets actual avoidance?
    ~ I tank with a Beorning, my opinion is invalid. ~

  11. #61
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by Joedangod View Post
    I'm curious; How are you calculating the standard deviation on the targets avoidance when you don't know the targets actual avoidance?
    Standard deviation is a calculation based on the variation in the data. It in this case is based on the forumula N*p*(1-p) for binomial random variables.

    Estimates of p come from the data itself, the variance is an estimate as well based on estimates of p. I also computed variance of p by using subsets of the data and found it to be consistent with the formula above.

    The subset approach involves using smaller groups of attacks to estimate p many times and then look at the variation in that estimate. It agrees fairly well with the formula above using estimated values of p from the data.
    This second process doesn't require knowing what p actually is only changes in p are looked at to compute the variation.

    Bottom line is that statistical results always have a probability of being wrong, but those probabilities can be estimated as well and are very small.

  12. #62
    Quote Originally Posted by TiberiasKirk View Post
    <snip>
    So you're guessing?

    Your initial test only seemed to be based off two parses so you couldn't possible have worked out the mean value for the avoidance in that one. Your later tests were presumably comprised of 50 parses, one for each wolf with half being disorientation builds and half being non-disorientation. Let's assume a 50 attack per wolf average, not really based on anything just a rough guess at how many hits a wolf can withstand, this would mean you have a rough total of 1250 attacks for the Disorientation build and 1250 attacks for the non-disorientation build. You don't know the uptime on the disorientation effect on but seem to guess it's somewhere in the 1/3rd region which would leave your disorientation sample at a rough 417 attacks with the remainder essentially being a non-disorientation build. You also like to mix in a few non-BPE attacks like warchant so that number goes down further. There's also the case of you using stuns in the rotation whilst testing BPE so those stuns are gonna skew results.


    Overall it seems like you got a very small sample size with no idea of the actual mean avoidance in order to accurately calculate the standard deviation. I'm no air-force statistician but I think they might raise some issues with your methodology.
    ~ I tank with a Beorning, my opinion is invalid. ~

  13. #63
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by Joedangod View Post
    So you're guessing?

    Your initial test only seemed to be based off two parses so you couldn't possible have worked out the mean value for the avoidance in that one. Your later tests were presumably comprised of 50 parses, one for each wolf with half being disorientation builds and half being non-disorientation. Let's assume a 50 attack per wolf average, not really based on anything just a rough guess at how many hits a wolf can withstand, this would mean you have a rough total of 1250 attacks for the Disorientation build and 1250 attacks for the non-disorientation build. You don't know the uptime on the disorientation effect on but seem to guess it's somewhere in the 1/3rd region which would leave your disorientation sample at a rough 417 attacks with the remainder essentially being a non-disorientation build. You also like to mix in a few non-BPE attacks like warchant so that number goes down further. There's also the case of you using stuns in the rotation whilst testing BPE so those stuns are gonna skew results.


    Overall it seems like you got a very small sample size with no idea of the actual mean avoidance in order to accurately calculate the standard deviation. I'm no air-force statistician but I think they might raise some issues with your methodology.

    The probabilities I am ESTIMATING contain the chances disorientation is up and the chances the target is stunned as it was done consistently both ways (with and without disorientation up)
    Now you have outed yourself as not being a statistician. All estimates can be called guesses (they actually mean the same thing). Without estimates we don't do statistics. Good estimates also have estimates about how good they are (which I have done with error bars).

    You need to take a statistics class. If your estimates EVER had an estimate of their potential error associated with them you would have a leg to stand on. Please go on and continue with all the non-mathematically based reasons on why I am wrong, LOL

  14. #64
    Quote Originally Posted by TiberiasKirk View Post
    The probabilities I am ESTIMATING contain the chances disorientation is up and the chances the target is stunned as it was done consistently both ways (with and without disorientation up)
    You haven't shown any shred of evidence for that. You've given very little info on your methodology, in fact, you haven't shown any calculations at all. Ever.

    What you have shown is that you have a small sample size and some limited graph creating capability. You're the one making a claim here, you really need to actually try prove it to be true rather than just insisting it is true.
    ~ I tank with a Beorning, my opinion is invalid. ~

  15. #65
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by Joedangod View Post
    You haven't shown any shred of evidence for that. You've given very little info on your methodology, in fact, you haven't shown any calculations at all. Ever.

    What you have shown is that you have a small sample size and some limited graph creating capability. You're the one making a claim here, you really need to actually try prove it to be true rather than just insisting it is true.
    I posted my chat logs. I KNEW you were incapable of analyzing them.

  16. #66
    Join Date
    May 2017
    Posts
    307
    Quote Originally Posted by TiberiasKirk View Post
    I posted a new analysis based on the bullroarer test server data I had which shows Dissorientation completely broken on the test server.
    My analysis is against targets that can actually parry and evade (LOL) and contains error bars showing the statistical significance of the results. Your results are not professional at all. They show the lack of partial avoidance and any full avoidance gain, but instead of delving deeper to see what is wrong on the test server, you have no intention of making the game better and would be more than happy if they didn't fix the disorientation trait.
    Theory went that Disorientation reduced partial avoidances.
    Partial evasion counts as a partial avoidance.
    In both tests there were over 2900 attacks, using only a few select skills with high finesse. I believe that was a high enough sample size for this test. I even oncreased the number of parses just to make sure it's detailed enough.
    Results have proven the theory wrong, Disorientation did not reduce partials in any sense, with a 1.1% increase in partial avoidance even. That can be placed under "circumstantional error" to be honest. 1.1% means absolutely nothing when it comes to this test. Again, there were 2900+ attacks made, partial evasion didn't decrease, that was the goal.

    Disorientation is not bugged. It's trash and should be replaced with something else. But there are better traits anyway and you can easily skip it for no loss whatsoever.

    Again, 2900 attacks. 1.1% difference in favour of increased partials. Not a 5% reduction, not even close. 1.1%.

    Let this thread and theory die, it's done for discussion as far as I'm concerned. Until someone can run a more complete test, post parses and explain in greater detail than I have, I'll consider this disproven.

    Back to actually being useful and my real-life work.

  17. #67
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by Decrepify View Post
    Theory went that Disorientation reduced partial avoidances.
    Partial evasion counts as a partial avoidance.
    In both tests there were over 2900 attacks, using only a few select skills with high finesse. I believe that was a high enough sample size for this test. I even oncreased the number of parses just to make sure it's detailed enough.
    Results have proven the theory wrong, Disorientation did not reduce partials in any sense, with a 1.1% increase in partial avoidance even. That can be placed under "circumstantional error" to be honest. 1.1% means absolutely nothing when it comes to this test. Again, there were 2900+ attacks made, partial evasion didn't decrease, that was the goal.

    Disorientation is not bugged. It's trash and should be replaced with something else. But there are better traits anyway and you can easily skip it for no loss whatsoever.

    Again, 2900 attacks. 1.1% difference in favour of increased partials. Not a 5% reduction, not even close. 1.1%.

    Let this thread and theory die, it's done for discussion as far as I'm concerned. Until someone can run a more complete test, post parses and explain in greater detail than I have, I'll consider this disproven.

    Back to actually being useful and my real-life work.
    Disorientation is broken on bullroarer. It reduces neither partial or full avoidance. Testing on live is irrelevant since the plan is to implement the changes on bullroarer into the live server.
    I maintain you don't want to talk about it anymore because you don't want it fixed. Maybe when they go to fix it they could have it reduce partials instead of full avoidance making it useful.

  18. #68
    Quote Originally Posted by TiberiasKirk View Post
    I posted my chat logs. I KNEW you were incapable of analyzing them.
    I could very well go through and analyse them but why would I? I'd get a small sample size of results that is kinda flawed anyway since you decided to attack more than one type of mob in at least one of the parses. The whole chat log thing is dumb anyway, either upload your program that you use to analyse it along with the results of the analysis or upload some Combat Analysis screenshots since we already know combat analysis observes this information correctly.

    You keep saying you have done the calculations but you haven't shown what inputs you've used for the calculations. You probably know that you're making mistakes and don't want to admit you're wrong but it's honestly better than continuing to lie.
    ~ I tank with a Beorning, my opinion is invalid. ~

  19. #69
    Join Date
    May 2017
    Posts
    307
    Quote Originally Posted by TiberiasKirk View Post
    Disorientation is broken on bullroarer.
    Theory #2 to disprove then. Will post when BR comes back on tuesday/wednesday.

  20. #70
    Join Date
    Jan 2010
    Posts
    607
    Quote Originally Posted by Decrepify View Post
    Theory #2 to disprove then. Will post when BR comes back on tuesday/wednesday.
    You did your test on bullroarer, but you never said anything about full avoidance.

  21. #71
    Join Date
    May 2017
    Posts
    307
    Quote Originally Posted by TiberiasKirk View Post
    You did your test on bullroarer, but you never said anything about full avoidance.
    Well, I'll unslot every piece of finesse gear I got, including relics, setting my finesse to 0.
    Then I'll make 20 parses with and without disorientation. I think that's a good enough sample size. Unless you care to suggest what would be better.

    Also, please read my test results again. I have clearly stated why I've even done 11 parses, not 3 as originally planned. I even mentioned why I won't even bother with full avoidance, because that's not the part that mattered. You claimed Disorientation worked on partials. It doesn't. Now you claim it's bugged. Okay, I'll test that then. Without finesse. Good luck till then.
    Last edited by Decrepify; May 27 2018 at 11:18 AM.

  22. #72
    Quote Originally Posted by Decrepify View Post
    Okay, I'll test that then.
    Can you double check you're hitting the dummies from the front on the next set of tests? I don't think it would make much, if any, difference to the Disorientation test I just find it odd that the dummies only ever evade.
    ~ I tank with a Beorning, my opinion is invalid. ~

  23. #73
    Join Date
    May 2017
    Posts
    307
    Quote Originally Posted by Joedangod View Post
    Can you double check you're hitting the dummies from the front on the next set of tests? I don't think it would make much, if any, difference to the Disorientation test I just find it odd that the dummies only ever evade.
    I can comfortably say that I was attacking them from the front in ALL of my tests. I never had continuous Staggered debuff proc, always expired. And in my original testing rotation, stagger was one of the skills used.
    I will double-check and report to make sure when I parse in the future, thank you for bringing that potential issue up!

 

 
Page 3 of 3 FirstFirst 1 2 3

Posting Permissions

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

This form's session has expired. You need to reload the page.

Reload