We have detected that cookies are not enabled on your browser. Please enable cookies to ensure the proper experience.
Page 2 of 3 FirstFirst 1 2 3 LastLast
Results 26 to 50 of 73
  1. #26
    Join Date
    Sep 2010
    Posts
    348
    Quote Originally Posted by Calico View Post
    Just for discussion's sake, let's assume that a game actually had a minor bug in its random number generator.

    How would anybody ever report it as a problem? Every time someone complains, they're dog-piled by a dozen posts telling them that they don't understand how random numbers work.
    The bug report should look something like this: https://www.lotro.com/forums/showthr...THREE-THOUSAND (except that here, it was actually 'proof' that the RNG is really as random as you might expect). (Thanks bigsix66 for the link.)

    The thing is, one can never truly prove that an RNG is or is not correct based only on the results. (You'd need to look into the code.) What we can do instead is calculate the probabilities of each of these four situations:

    1 the RNG is functioning properly and the results indicate as much
    2 the RNG is functioning properly but the results look as if it isn't
    3 the RNG is malfunctioning but the results look as if it's working properly
    4 the RNG is malfunctioning and the results are suspicious (i.e. suggest that it's malfunctioning).

    Based on the results, either only situations 1 and 3 or only situations 2 and 4 are possible. The probability of each of these situations depends on the number of trials. By increasing the number of tests, we can be increasingly confident that one of the two situations is rather unlikely, and so conclude that the other is probably true.

    If someone ever posted a thread like the one I linked above, only concluding (based on the evidence) that the RNG was broken, I don't think that person would be 'dogpiled'. However, people claiming that the RNG is broken based on a sample size of 10... well... that last part (the sample size) pretty much shows that they don't understand how statistics work. (Knowing how an RNG works is irrelevant - you only need to know what it's supposed to do.)

  2. #27
    Quote Originally Posted by Svejakkes View Post
    In my experience the RNG is flawed in a 'color'-sense. It generates statistically way too many same results in a row. The fact that the true/false distibution flattens out in the long term, doensn't make it a fair RNG by definition.

    Here are two random sequences:

    Pattern 1
    101101010001101100101100001101 110010110100101

    Pattern 2
    111110001111100000100011111111 001000001110000

    Both sequences may have about an equal number of 0's and 1's, but the structures are different. The first pattern is 'smooth random', the second pattern is 'sticky random'. Should you use these patterns as a source for an audio synthesizer, then the first pattern would be called 'white noise' and the second pattern 'colored noise' (pink, red, brown, whatever).

    Most synthesizers have at least a white noise generator, because in combination with a filter it can cover a large audio spectrum. Colored noises are more limited, only useful for low frequencies and/or grainy effects.

    Back to LOTRO. It's RNG appears very colored, definitely not white. It's only 'fair' over long sessions. Technically it's too colored toward low frequencies. The several comments here about how fine everything turns out in the long term, actually confirms this (FWIW: period is inverse of frequency).

    But these comments are not really justifications. The fact that randomness works out well in the long term, doesn't mean it shouldn't work well (within statistical fairness etc.) in the short term. That's like saying "there is no financial crisis, it's just fluctuating, everyting will be good again in 50 years, just consider yourself rich".

    Bottom line: a fair RNG should produce results more or less like white noise. LOTROs RNG appears way too colored and it would be good if they could improve that.

    A very good hypothesis as to what might actually be happening, well illustrated and explained.

    This is kind of backed up by experiences doing the warbands bugud/kramp/gundul in rapid succession as was done frequently in the past and still done occasionally today. Commonly a random herd of players would slay bugud charge off and kill kramp and then charge up the hill to finish off gundul. I tended not to open the reward boxes until the last was slain and I was safely out of combat back in town at which point I opened the boxes in rapid succession and tended to find that if one box gave abundant rewards they all did and vice versa for poor rewards or just marks. I tended to notice this same pattern when people I was fellowed with opened their boxes.

    Back with crafting I've noticed patterns of "coloured noise" tend to happen, for good and bad outcomes, a heck of a lot too.
    Must remember to engage brain before using keyboard

  3. #28
    Join Date
    Jun 2011
    Location
    Germany
    Posts
    2,036
    I marked some numbers bold. Please look at them:
    Quote Originally Posted by Svejakkes View Post
    Here are two random sequences:

    Pattern 1
    101101010001101100101100001101110010110100101

    Pattern 2
    111110001111100000100011111111001000001110000
    So I get:
    Pattern 1a
    10110111
    Pattern 2a
    10111000

    This is every 6th number of each sequence.

    Still random? Of course. Equal number of 0's and 1's? No.


    What you all forget ist, there is no PRNG for you as one player. Instead, there is one PNRG for all random events. It doesn't matter, if the PRNG generates paterns like 1 or 2. You as a player will get the patterns like 1a or 2a, because the other numbers are assigned to events of other players. And it's not only crafting, there are a lot of randon numbers in game (damage, heal, block, parry, ....)



    At least, I hope you agree with me, that the LOTRO PRNG isn't implemented like this:

    (from: http://xkcd.com/221/)
    Wer Hilfe will, muss Informationen liefern.
    >> Lotro World transfer status <<

  4. #29
    Join Date
    Feb 2007
    Location
    Birmingham, AL
    Posts
    2,946
    Quote Originally Posted by Calico View Post
    Just for discussion's sake, let's assume that a game actually had a minor bug in its random number generator.

    How would anybody ever report it as a problem? Every time someone complains, they're dog-piled by a dozen posts telling them that they don't understand how random numbers work.
    It is actually very easy to test. T's engineering team takes the RNG program and runs several 10,000 of rolls many times and checks the results using standard math. This is reviewed by the QC department and they can actually see if there is an issue or not.

    For some reason I feel this has been done more than once. While players get a very small sample of the RNG results and make generalized statements based on that data.

    Who would you think has a better idea of what is really happening inside the RNG?
    I wonder what IS behind that waterfall?
    The Player Council knows!

  5. #30
    Quote Originally Posted by Calico View Post
    Just for discussion's sake, let's assume that a game actually had a minor bug in its random number generator.

    How would anybody ever report it as a problem? Every time someone complains, they're dog-piled by a dozen posts telling them that they don't understand how random numbers work.
    You've had the "how would it be reported?" question answered already. To the larger point, the reason you see these "dog-piles" is that a lot of people realize that Turbine's resources are finite. There are a lot of things we'd like to see these people doing...like fixing for-real critical bugs. Let's assume, for the sake of argument, that there were a minor bug with random generation....say, oh, over a sample of 10,000 or so, there's a notable, repeatable discrepancy of 3% or so. I'm not sure how you'd see that 3%...people who're better at stats than I am can tell me whether 3% is within the expected error for a test that size (I think it is), but let's pretend just for now that you really could see it, point to it, and say "yep, problem".

    Is that 3% (only truly observed over very large samples) worth chasing down, in the face of other things that could be getting fixed? Do we really want Turbine running off to re-validate their RNG every time somebody feels something off over a vanishingly small sample size? Seriously, think about the scope there....given the number of die rolls that happen across the platform on a daily basis, you figure somebody, amongst thousands of players, will encounter a local streak and /bug it....3 times a week? 4? At least?


    Personally, I wish Turbine would just move to a deterministic system, where if you had a 50% crit chance, the system would force every other roll into the expected range. It'd be a win for everybody:

    1) People who don't understand randomness will get what they expect to see, and these threads will vanish
    2) People who do understand randomness will realize that, over the long term, they're getting exactly what they had before, and not care about the change one way or the other
    The forums are not an accurate representation of the thoughts and feelings of the whole player base. Those who like a particular feature are in the game enjoying that feature. Those who don't like it log out to mention it on the forums. It is a relevant but biased source of feedback, and any claims of community desire should take this fact into account.

  6. #31
    Join Date
    Feb 2007
    Posts
    12,677
    Quote Originally Posted by Ailedra View Post
    I wish Turbine would just move to a deterministic system, where if you had a 50% crit chance, the system would force every other roll into the expected range.
    I don't. I find "deterministic" systems to be incredibly boring. What's fun for me is the surprise of getting a result I didn't expect, and I fully understand that means "unexpected" will mean "unexpectedly bad" as often as it means "unexpectedly good", at least over the long haul.

    Khafar

  7. #32
    Join Date
    Jan 2007
    Location
    Depends on which game I'm in :)
    Posts
    6,708
    Quote Originally Posted by Amrundir View Post
    What you all forget ist, there is no PRNG for you as one player. Instead, there is one PNRG for all random events. It doesn't matter, if the PRNG generates paterns like 1 or 2. You as a player will get the patterns like 1a or 2a, because the other numbers are assigned to events of other players. And it's not only crafting, there are a lot of randon numbers in game (damage, heal, block, parry, ....)
    This is exactly it. There is one RNG for the server and everyone doing anything that requires a random number is getting their number from the same RNG function. So the numbers that one player gets when crafting are not consecutive numbers. You get whatever happens to be the next number from the RNG. The four numbers the OP got from the RNG were, in all likelihood, not even spaced every nth number, but rather having n1, n2, and n3 other numbers between them.


    People posting things like this:
    Quote Originally Posted by Schaijian View Post
    There is absolutely nothing wrong here, it´s simple math, if you don´t have 100% chance for success, there is always the chance to fail and in your case 5% for each try - making it not very likely to fail 4 times in a row, but still possible wih a chance of 0.05^4 or 0.000625%. Or in other words: one time in 160.000 you statistically have this bad luck. Sucks but still no failure in the system.
    Quote Originally Posted by twittfounder View Post
    Yes...even IF you could accomplish that...it still IS possible to not crit 4 times in a row. 0.000625% -- for exactly four conecutive non crits in a row.
    However, as unlikely as that is...it is still possible.

    Are irrelevant since they only apply to four consecutive numbers. The four numbers the OP got while crafting were not consecutive numbers. In order to determine what the actual chance of the OP getting their four failures, you would need to know how many numbers the RNG gave out during that time (possibly other things, too). And we (the players) do not know that information. The OP never stated how they were crafting the four items, but the fewest possible numbers to come out of the RNG would be from selecting 4 as the quantity and craft them. But even so you are still not getting four consecutive numbers from the RNG, only minimizing the number of values the RNG is giving out to others. More numbers would be generated if you crafted one item at a time and made trips to the vault and vendors to get materials for the next item just before crafting it. Still all we could do is guess at how many numbers the RNG generated during that time.


    As Maxal stated above, the only real (and correct) way to test the RNG is for Turbine to run a test program that makes tens or hundreds of thousands of calls (or more) to the RNG function and analyzing the data they get (e.g. looking for randomness, trends). This would have to be done whenever anything that goes into creating the executable code is changed (e.g. compiler, CPU).

    Like everyone else, I've had good streaks, bad streaks, and results that made me go "hmmmm, that's weird...". But none was ever the fault of the RNG.
    "Everyone's got one. Better to be a smart one than a dumb one." - me
    "It's people like you that [I]think[/I] you know everything that give those of us that [I]do[/I] a bad name." - me
    "Eat, drink, and be Merry, 'cause Pippin was a fool of a Took."

  8. #33
    Join Date
    Jan 2007
    Location
    Depends on which game I'm in :)
    Posts
    6,708
    Quote Originally Posted by Khafar View Post
    Turbine gets these sorts of complaints all the time. If they get enough of them in one area, they'll go run tests on a large scale to see how it falls out, and of course, it's almost never an actual problem... just bad luck. I've seen at least a dozen threads over the years where Turbine folks have said pretty much that.

    There is one glaring exception, though... the infamous "Wi Flag". Poor Wi. I remember his complaints about this at the Crossroads of Dereth. Naturally, he caught a lot of grief from posters too, but over time, there were just too many people reporting (or seeing) the same thing. I witnessed it myself one night, where we were in a group going after olthoi in some dungeon I forget (BDC, perhaps), and one guy in our party was getting swarmed all night.

    Khafar
    But even the "Wi Flag" was not an exception. The "Wi Flag" was not a problem with the RNG. It was a problem in how the result from the RNG was being used. Big difference. RNG worked fine, the implementation of the algorithm was bad. You can't blame the RNG if you multiply the number by 2 instead of 7, or if you forget to add 12 to it.

    There was an issue during the first year of LOTRO where the Smithing Hammer was not adding the crit chance when used by a Metalsmith (worked fine for a Weaponsmith). Took some time to figure out what was happening, but they did find and correct the problem. Again, it was not a fault of the RNG. The fault was in how the algorithm that was using the value from the RNG was being implemented.
    "Everyone's got one. Better to be a smart one than a dumb one." - me
    "It's people like you that [I]think[/I] you know everything that give those of us that [I]do[/I] a bad name." - me
    "Eat, drink, and be Merry, 'cause Pippin was a fool of a Took."

  9. #34
    Join Date
    Feb 2007
    Posts
    12,677
    Quote Originally Posted by Penter-Kar View Post
    The "Wi Flag" was not a problem with the RNG.
    Sure, but that's not really the point. Players don't care if the Linux (or Windows) PRNG is basically sound. What they care about is how it's used in the game, and what implications that has for gameplay. In Wi's case, this was rather awful.

    Turbine has run tests on how they use random numbers in LOTRO repeatedly over the years, based on complaints about bad streaks. To the best of my knowledge, they've never found a problem with it. Yes, individuals are going to see streaks, which is completely expected. What some people are really asking for is an RNG which is less random (e.g. designed to reduce or eliminate any bad streaks).

    Khafar

  10. #35
    Quote Originally Posted by Maxal View Post
    It is actually very easy to test. T's engineering team takes the RNG program and runs several 10,000 of rolls many times and checks the results using standard math. This is reviewed by the QC department and they can actually see if there is an issue or not.

    For some reason I feel this has been done more than once. While players get a very small sample of the RNG results and make generalized statements based on that data.

    Who would you think has a better idea of what is really happening inside the RNG?
    First off, I'd like to say that my original question was half-way facetious, pointing out that these threads generate many replies that are more about patting oneself on the back for superior knowledge of math, and ignoring the idea of whether there might be a bug or not.

    To address the issue seriously: I completely agree that the Turbine code has a library function for generating random numbers, and that function has been tested and verified over and over again.

    But that library function is probably called hundreds, maybe thousands of places. Everywhere from how hard you hit in combat, to whether you crit in crafting, to how long Daffodilly takes to finish her muffin. Each of those places has to not only call the function, but it has to interpret the random value into a game result. Now we're talking hundreds or thousands of things that need to be tested, not just one. So the idea that there may be a bug in one of those places doesn't seem so far fetched.

    I'm not claiming that there's a bug, and I'm not saying the OP has a valid complaint. I'm pointing out that this problem is more complex than just how random numbers are generated, and that every time a game system is tweaked by adding a new buff or some such, there's the potential for a broken result to be introduced.

  11. #36
    Quote Originally Posted by Penter-Kar View Post
    Are irrelevant since they only apply to four consecutive numbers.
    This is not correct. They apply since all that matters is there are four independent random samples. Perhaps you are asserting there is a problem with the RNG after all. If not, think on the odds of a simple coin flip (a fair coin, etc, no tricks) after someone else has flipped it n times and got n heads (say n=2,3,10, or whatever you want; the answer is the same regardless).
    [CENTER]Othrandir[/CENTER]
    [CENTER]Resistance - Brandywine[/CENTER]

  12. #37
    Join Date
    Oct 2010
    Location
    Hamilton, NY
    Posts
    3,699
    Quote Originally Posted by clappi View Post
    They apply since all that matters is there are four independent random samples.
    That math is for four independent but consecutive samples. His point was that they're not consecutive, and so the math is off.
    [CENTER][IMG]http://img.photobucket.com/albums/v341/tindragon/Base.png[/IMG]
    Steam: Sneakeh Snake[/CENTER]

  13. #38
    Quote Originally Posted by TinDragon View Post
    That math is for four independent but consecutive samples. His point was that they're not consecutive, and so the math is off.
    I understand what he said. It's flat wrong. The math is not for four independent but consecutive samples.

    It does not matter if you flip a coin and get tails, hand it to me and I flip four more tails, and then hand it back. Your chance of getting tails on your next flip is still 50%. Just as if you never handed to me before flipping it for a second time. In this case consecutive samples are irrelevant, unless he's making a point that the RNG does not provide independent random samples. Good luck with that.

    Calico's point about the use of random values is entirely valid, in that just because an RNG may be reasonable the use of its resultant values are not necessarily also reasonable. However, my point was the posts made by people describing the chance of failing to crit four times in a row are also completely valid; a percent chance of anything, compounded over multiple samples, still gives you a chance, even though it may be small. And the fact someone "rolled" via the RNG between your first random number and your second one is completely irrelevant as long as the RNG provides independent random samples (the very purpose of an RNG).

    As correctly stated above by multiple others, the odds of not critting something with a 95% crit chance out of four tries is non-zero: (0.05 ^ 4) = 0.00000625 => 0.000625% (a.k.a. 1 in 160,000 cases).
    Last edited by clappi; Aug 11 2013 at 09:04 PM.
    [CENTER]Othrandir[/CENTER]
    [CENTER]Resistance - Brandywine[/CENTER]

  14. #39
    This bug was noticed since game launch. The way it works is that only the first roll in random, after that there is a much larger percent of getting the same result. So if you get the 5% crit, hurry up and craft 4 more of the same item and you have a much higher than 5% chance of critting over and over. Crafters first noticed this but after six years of samples its pretty well established. Now the chance of missing a 95% crit four times is 160,000 to 1, and it happens over and over again-- its not random, and it never has been.

    It would be better if this was fixed, and it could well be quite costly, though not as much as in the SoA days when the crafting components were hugely expensive. A single used recipe attempt in those days could well constitute 20 or 30% if a player's entire net worth. Multiple non-crits could and did put people out of the game. Its not so bad now that the relative price of a craft attempt is lower, or that the 85 level single used recipes are so expensive that people generally don't try to make more than one attempt.

    Now, this can work to your advantage. If you crit the first attempt, continue on until you get a non-crit. If you are working with expensive components and you fail the first attempt, simply log off and log into a different character to reset the string algorithim, then relog to your crafter and try again. You can earn back a little of what was unfairly taken from you if you can master this method. Still, it would be better if Turbine just fixed this bug.

    One thing I've noticed is the string usually gets broken or reset after 4 or 5 things of the exact type are crafted in a row, one after the other very rapidly.

  15. #40
    Join Date
    Apr 2007
    Posts
    532
    Quote Originally Posted by clappi View Post
    I understand what he said. It's flat wrong. The math is not for four independent but consecutive samples.

    It does not matter if you flip a coin and get tails, hand it to me and I flip four more tails, and then hand it back. Your chance of getting tails on your next flip is still 50%. Just as if you never handed to me before flipping it for a second time. In this case consecutive samples are irrelevant, unless he's making a point that the RNG does not provide independent random samples. Good luck with that.

    Just to follow up and clarify this a bit in an infinite number sequence (which is essentially what the RNG is) any sample whether consecutive or non consecutive will, reflect the larger sequence. With the coin flip it has no "memory" of what the last flip was so it does not care if you are recording the numbers sequentially or not its next flip has a 50% chance and no matter when, where, or how you record it over time the flips will all tend towards 50%. SImply put that act of recording or not recording has no affect on the overall pattern thus it does not affect the statistical probability of any particular sequence.
    "We can deny our heritage and our history, but we cannot escape responsibility for the result... we cannot defend freedom abroad by deserting it at home. " - E R Murrow

  16. #41
    Quote Originally Posted by MithrielWielder View Post
    This bug was noticed since game launch. The way it works is that only the first roll in random, after that there is a much larger percent of getting the same result. So if you get the 5% crit, hurry up and craft 4 more of the same item and you have a much higher than 5% chance of critting over and over. Crafters first noticed this but after six years of samples its pretty well established. Now the chance of missing a 95% crit four times is 160,000 to 1, and it happens over and over again-- its not random, and it never has been. ...


    Are there more specifics characterizing the nature of the problem somewhere? Is there already-collected data somewhere I can check out? I downloaded the 3K crafting crit data, but knowing more specifics about what people think they were seeing would be helpful. At a quick first glance so far the 3K crit data looks like a fairly reasonable binomial distribution. Even the "unlikely" scenarios (4 17% crits in a row, 35 83% non-crits in a row, etc) I think are more likely than postulated in that thread.
    [CENTER]Othrandir[/CENTER]
    [CENTER]Resistance - Brandywine[/CENTER]

  17. #42
    Join Date
    Jul 2007
    Location
    South Dakota
    Posts
    2,884
    Quote Originally Posted by Maxal View Post
    I was recently trying to explain this exact thing to another player. He cannot believe that the RNG is not broken. Even after the question was asked and answered in a 20 Questions onto so long ago.
    Yeah, it's the same thing with my roommate. Every few months he'll complain about this, and I've stopped trying to explain it to him because he never believes me.

    I will never understand why people will so readily ignore solid evidence in their face.



    Thanks for the links, though, to whoever posted them. I'll at least have something to show him next time.
    [B][COLOR=red]Turn[/COLOR] - Guardian
    [COLOR=red]Balandolen[/COLOR] - Lore-master
    [COLOR=red]Turnholm[/COLOR] - Minstrel

    [/B]

  18. #43
    Join Date
    Jul 2007
    Location
    Utah
    Posts
    418
    The OP does have a valid gripe because Turbine's RNG accuracy is most definitely skewed towards the greater frequency of numbers attempted over the lower frequency numbers attempted. However, the RNG will always correctly balance the proper percentages as the frequency of attempts increases. So, as far is Turbine is concerned the RNG is WAI.

    Therefore, you can expect low frequency attempts in using Turbine's RNG to have results that dont always match the rolling of actual dice to determine probabilities!

    Las Vegas would go broke if they were to use Turbine's RNG to determine the odds because there would be too many people sucessfully rolling a 1 in 20 chance in a row (the OP did it 4 times). And Las Vegas casinos couldnt count on the people to gamble long enough after cashing in their winnings for Turbine's RNG to balance the odds back in their favor.

    Welden
    Welden of Elendilmir

  19. #44
    Join Date
    Feb 2007
    Location
    Brandywine
    Posts
    2,226
    Can the Thread Starter please explain exactly what he/she was crafting, and what tools and items were used to achieve 95% crit chance.

    I am dying to know.

  20. #45
    Quote Originally Posted by welden View Post
    The OP does have a valid gripe because Turbine's RNG accuracy is most definitely skewed towards the greater frequency of numbers attempted over the lower frequency numbers attempted. However, the RNG will always correctly balance the proper percentages as the frequency of attempts increases. So, as far is Turbine is concerned the RNG is WAI.

    Therefore, you can expect low frequency attempts in using Turbine's RNG to have results that dont always match the rolling of actual dice to determine probabilities!
    I guess as far as math is concerned the RNG is WAI as well. This is exactly how pure randomness works; the behavior you describe matches actual dice. You have to get a fairly large sample size before you can really draw any conclusions or see any notion of pure randomness in the sample. Do you have such data somewhere that makes you think otherwise? I'd be happy to take a look at it with you.

    Let's assume the RNG/etc is working fine for the moment. If 100,000 people logged in per day, and lets say 1 in 10 craft just four gathering items (i.e. crit rate is 5%), we'd expect to see 22 quad-crits per year (by contrast we'd probably expect around 26K triple-crits per year). If we up the numbers of crafted gathering items to 40 then we'd expect to see at least 170 quad-crits per year (in this case upping the crafting count by 10x would actually provide a greater increase than simply increasing the crafters by 10x... but I haven't sat down to figure out by exactly how much). Now while these login/crafting numbers are completely arbitrary, hopefully they give some insight into how rare, or common, such events can be.

    *EDIT: For those interested, it looks like if we increased the number of crafted items to 40 in the above example we should expect 2-3 (more often 2 than 3) people see one or more quad-crits per day (or quad-fail to crit with a 95% crit chance)... approximately 867 per year.

    I'm not saying there is no bug. I have no way of knowing that. But I haven't seen anything here yet that would make me think something's necessarily broken. If we want to get Turbine to look at something we need to give them something more compelling than what's been posted in this thread (and almost all threads of this nature). Anyone feel like writing a plugin to log RNG-related stats? Folks could then more easily volunteer hard data for analysis.
    Last edited by clappi; Aug 12 2013 at 11:32 PM.
    [CENTER]Othrandir[/CENTER]
    [CENTER]Resistance - Brandywine[/CENTER]

  21. #46
    Quote Originally Posted by welden View Post
    The OP does have a valid gripe because Turbine's RNG accuracy is most definitely skewed towards the greater frequency of numbers attempted over the lower frequency numbers attempted. However, the RNG will always correctly balance the proper percentages as the frequency of attempts increases.
    Think about what you just said. Translated roughly, you're saying that over the long term, the results line up with the percentages, but they don't always over the short term. Put even more simply, you're seeing small streaks.

    Know what else works that way? Real-world RNGs, like dice.

    I just flipped this quarter, and got heads three times in a row. I should /bug this quarter, because it's clearly "skewed towards the greater frequency of numbers attempted over the lower frequency numbers attempted."

    Quote Originally Posted by welden View Post
    Las Vegas would go broke if they were to use Turbine's RNG to determine the odds because there would be too many people sucessfully rolling a 1 in 20 chance in a row (the OP did it 4 times). And Las Vegas casinos couldnt count on the people to gamble long enough after cashing in their winnings for Turbine's RNG to balance the odds back in their favor.
    Actually, Vegas uses complex RNG devices that do exactly this: they use high-tech devices such as "dice" and "roulette tables". Casinos count on streaks....they count on winner's feeling like they're on a streak, and pushing themselves right out of their own winnings....or continuing to gamble after a streak of losses, expecting that they're now "due".


    In order to consistently get the expected distribution over small sample sizes, you have to "force" an RNG to provide you those results....in which case, it's no longer random, it's deterministic.
    The forums are not an accurate representation of the thoughts and feelings of the whole player base. Those who like a particular feature are in the game enjoying that feature. Those who don't like it log out to mention it on the forums. It is a relevant but biased source of feedback, and any claims of community desire should take this fact into account.

  22. #47
    Join Date
    Aug 2010
    Location
    Germany
    Posts
    1,102
    As long as no one really knows how Lotro's RNG actually works, it is not really right to say it works correct :-)
    The flaws and pitfalls of computer-created random numbers are documented, and it is easier to do it wrong than right..

    IMHO, a general, server-wide RNG might work perfectly on the whole set of generated numbers, but that isnt helping an individual caught on the bad side of the RNG. And just as certain streaks become a near-certainty, a large enough sample provided, it also is a near-certainty that some players get shafted or are insanely lucky when it comes to getting the 'right' numbers from the RNG.
    I would assume even if several players would track a large amount of random numbers at the same time, there will be some more lucky than others, even after a million rolls.
    Let's assume a sample of a billion numbers add up to the average. the more people take numbers at random, creating a subset of these numbers, some are bound to have picked numbers adding up to less or more then the average. of course, all totalled up would be average, but still the individual subsets are not.

    I would like to see a character-based, 'intelligent' RNG, which even outs the rarer streaks, tweaking chances to provide a success of failure when the odds turn too low. Of course that wouldn't be random, but i think it is more important to the individual player's perception on how lucky / unlucky he is.

    otoh, i might be totally wrong in my assumptions, which, tbh, seems rather probable :-)


    edit:
    http://archive.gamedev.net/archive/r...ess/page4.html
    shuffle bag and some more here, interesting read..
    Last edited by Rapunzel666; Aug 12 2013 at 12:01 PM.

  23. #48
    Quote Originally Posted by Hazelthorn View Post
    Im trying desperately understand something Turbine, why is it when i have a 95 percent chance to critical success on an item do i fail 4 times in a Row?? Seriously, can anyone understand that one?

    thats like being told to scribble anything on a paper and somehow getting it wrong or missing the paper completely over and over.

    Theres got to be something wrong. Please Fix Turbine.
    Here's a guess:

    You want an event generator based on a minimum percent success rate with extreme results eliminated. Others are satisfied with the status quo.

  24. #49
    This is just a tad off-topic, I apologize.

    Quote Originally Posted by Maxal View Post
    I was recently trying to explain this exact thing to another player. He cannot believe that the RNG is not broken. Even after the question was asked and answered in a 20 Questions onto so long ago.
    Should you refer to the discussion we (I think) had in the temprary forums, you are misrepresenting what has been said there.
    The issue there was the current individual loot system (aka lottery loot), which has the pRNG as input. The explicit message there was (and is): the pRNG appears to be working fine, but the way its results are applied seem to be off and could use a close look.
    And again, the question that was asked (and answered) in the july 20 questions didn't address that particular problem.

    I think it is important to get that distinction right, or the discussions will continue to revolve around that bit without helping the matter.

    FWIW, Penter-kar (and Khafar, too), already brought that up wrt. the Wi-Flag:
    Quote Originally Posted by Penter-Kar View Post
    But even the "Wi Flag" was not an exception. The "Wi Flag" was not a problem with the RNG. It was a problem in how the result from the RNG was being used. Big difference. RNG worked fine, the implementation of the algorithm was bad. You can't blame the RNG if you multiply the number by 2 instead of 7, or if you forget to add 12 to it.
    SNy
    LotRO on Linux! http://SNy.name/LOTRO/
    Also home to the LI progression diagram.
    Find the new forums unreadable? Try my forum theme.

  25. #50
    Join Date
    Apr 2007
    Location
    Brooklyn
    Posts
    3,713
    Quote Originally Posted by MithrielWielder View Post
    This bug was noticed since game launch. The way it works is that only the first roll in random, after that there is a much larger percent of getting the same result. So if you get the 5% crit, hurry up and craft 4 more of the same item and you have a much higher than 5% chance of critting over and over. Crafters first noticed this but after six years of samples its pretty well established. Now the chance of missing a 95% crit four times is 160,000 to 1, and it happens over and over again-- its not random, and it never has been.

    It would be better if this was fixed, and it could well be quite costly, though not as much as in the SoA days when the crafting components were hugely expensive. A single used recipe attempt in those days could well constitute 20 or 30% if a player's entire net worth. Multiple non-crits could and did put people out of the game. Its not so bad now that the relative price of a craft attempt is lower, or that the 85 level single used recipes are so expensive that people generally don't try to make more than one attempt.

    Now, this can work to your advantage. If you crit the first attempt, continue on until you get a non-crit. If you are working with expensive components and you fail the first attempt, simply log off and log into a different character to reset the string algorithim, then relog to your crafter and try again. You can earn back a little of what was unfairly taken from you if you can master this method. Still, it would be better if Turbine just fixed this bug.

    One thing I've noticed is the string usually gets broken or reset after 4 or 5 things of the exact type are crafted in a row, one after the other very rapidly.
    Don't forget your lucky rabbit's foot and your horseshoe! And never craft after you've walked under a ladder or broken a mirror.
    Work like no one is watching, dance like you don't need the money...

 

 
Page 2 of 3 FirstFirst 1 2 3 LastLast

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