test content
What is the Arc Client?
Install Arc

Everything you never wanted to know about Infinity Lock Boxes - 2019 Edition

2»

Comments

  • voodoopokeyvoodoopokey Member Posts: 244 Arc User
    leemwatson wrote:
    The main factor we do not know is how the game's probability tables are formatted. It would be pretty easy to fit all the infinity prizes in a basic 1-100 table which would then exclude the possibility of the odds of getting a T6 being lower than 1%. Yes it could be 1-1000 but (assuming a 0.4% chance) that would in theory improve the chances of getting a T6 as the number range is greater than a 1-100 table. As I mentioned earlier, the fact that has been claimed a couple of 'draws' gave approx 0.40% only backs up a 1% chance, it does not disprove it. A programmed 0.40% chance would invariably and reliably produce a lower draw rate.

    There is an assumption that the probabilities are static, and certainly that need not be the case... but it is by far the easiest thing to make, and there's no great reason to apply additional bias except to favour the user... and most of the evidence suggests there's no user-benefitting bias. I'm comfortable with the assumption.

    Based on the OP's posted data the chance that the drop rate for T6 ships is 1% falls outside the 95% confidence interval.. quite a ways outside of it.
    leemwatson wrote:
    Another factor that people who have never done any programming before will not know is that if there is one thing a computer does badly, it's producing a truly random number. Back in my early days of programming the venerable Spectrum and Commodore, programming them to produce 'random' often ended up them being 'pseudorandom' to the point of being predictable, and it's typically the same this day and age.

    That's an argument typically made by people who don't do any serious programming (at least any that relates to randomness). The algorithms computers use for generating random numbers are typically of very high quality - higher quality than physical dice (where bias can, and often is, introduced in the manufacturing process). The only time the quality of an algorithm's randomness is a serious consideration is when dealing with cryptography... but the only time you see people debating random number generators as a concept, is in video game forums.

    The main realistic consideration with computer based random number generation is ensuring that the generators state is opaque to anyone who relies on it being random. That's not an issue with games where decisions are made server-side.
    leemwatson wrote:
    One of the things missing in the OP's results is the Standard Deviation which the scientific method demands. With this sort of testing on the RNG statistics, SD is required, as are multiple samples of the same 10000.

    Again, mathematics is not science, so it isn't a question of "the scientific method". If all we're talking about is the chances of getting a T6 ship, then we can look at it as 10000 trials in which we either get a T6 ship or don't. In academic circles doing that in spite of not having set out to gather data for that, might get you some sideways glances... but we're on a forum, not writing a masters thesis. No standard deviation is required - we can calculate our confidence intervals by treating the numbers as proportions of the total.
    leemwatson wrote:
    From the couple of times I've seen this test done, all that can be indicated is there is a 'minimum' chance of 0.4%, which, as I said disproves the 0.4% claim.

    None of that is true.

    The 0.4% proportion is the mean value for our distribution of possible values for the odds of getting a T6 ship. It represents the maximum probability point (since it's what we actually got in the data) around which a normal curve can be drawn representing the total area of probability.
    ltminns wrote:
    The key to a pRNG is the seed. What do they use as the seed for generation of numbers?

    It really isn't. A random number generator's seed is simply a starting value to initiate the state of the generator. Without access to the algorithm itself, the actual raw output from the generator, and access to all of its outputs... we could seed the generator by asking a 5 year old to pick a number between 1 and 10... and it'd still be perpetually unpredictable.
  • ltminnsltminns Member Posts: 12,569 Arc User
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.
    'But to be logical is not to be right', and 'nothing' on God's earth could ever 'make it' right!'
    Judge Dan Haywood
    'As l speak now, the words are forming in my head.
    l don't know.
    l really don't know what l'm about to say, except l have a feeling about it.
    That l must repeat the words that come without my knowledge.'
    Lt. Philip J. Minns
  • warpangelwarpangel Member Posts: 9,427 Arc User
    edited August 2019
    ltminns wrote: »
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.
    The seed doesn't actually have much bearing on how "lucky" the result is. The only important thing is that it can't be always the same, because that would make the "random" stuff happen the same way every time (and yes, I have played games that have done that, it's ridiculous and very, very obvious). As long as the database ID was only used once to initialize the RNG in the beginning, it's as good a source as any.

    Every game with any significant luck component will have endless speculation on the RNG being bad because "other players are more lucky than me." It's almost always just clustering illusion.
  • foxman00foxman00 Member Posts: 1,481 Arc User
    warpangel wrote: »
    ltminns wrote: »
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.
    The seed doesn't actually have much bearing on how "lucky" the result is. The only important thing is that it can't be always the same, because that would make the "random" stuff happen the same way every time (and yes, I have played games that have done that, it's ridiculous and very, very obvious). As long as the database ID was only used once to initialize the RNG in the beginning, it's as good a source as any.

    Every game with any significant luck component will have endless speculation on the RNG being bad because "other players are more lucky than me." It's almost always just clustering illusion.

    Clustering Illusion? First I have heard of that term. Really interested to see where that came from, as i am curious. Anyway, these sort of tests done by players always make me nervous. Dont know why.

    Additionally, in relation to what is going on. Even if Cryptic posted the odds of winning, the next thing you would have is threads from players saying "I opened TRIBBLE boxes and didnt get anything, your odds state etc etc" Without knowing anything about how odds and random systems work.

    Its an interesting test, but we would need several like it conducted across multiple platforms, areas and times to get a decent result to minimise but not elimate the variables to even get a rough indicator of odds.
    pjxgwS8.jpg
  • warpangelwarpangel Member Posts: 9,427 Arc User
    foxman00 wrote: »
    warpangel wrote: »
    ltminns wrote: »
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.
    The seed doesn't actually have much bearing on how "lucky" the result is. The only important thing is that it can't be always the same, because that would make the "random" stuff happen the same way every time (and yes, I have played games that have done that, it's ridiculous and very, very obvious). As long as the database ID was only used once to initialize the RNG in the beginning, it's as good a source as any.

    Every game with any significant luck component will have endless speculation on the RNG being bad because "other players are more lucky than me." It's almost always just clustering illusion.

    Clustering Illusion? First I have heard of that term. Really interested to see where that came from, as i am curious. Anyway, these sort of tests done by players always make me nervous. Dont know why.
    https://en.wikipedia.org/wiki/Clustering_illusion
    Additionally, in relation to what is going on. Even if Cryptic posted the odds of winning, the next thing you would have is threads from players saying "I opened TRIBBLE boxes and didnt get anything, your odds state etc etc" Without knowing anything about how odds and random systems work.
    People already complain about opening N boxes and not getting the ship anyway. Not like it matters.
  • voodoopokeyvoodoopokey Member Posts: 244 Arc User
    ltminns wrote: »
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.

    There's literally no reason to ever do that.. and the idea is really, really stupid. The average user is an authority on absolutely nothing but has strong opinions on everything. Best to ignore their speculation and focus on the available data.
    warpangel wrote:
    As long as the database ID was only used once to initialize the RNG in the beginning, it's as good a source as any.

    It would not be a good seed as it would produce the same output in the same sequence every session. This might not be obvious to the user if they didn't take the exact same actions (since the output of an RNG is not the same as what they end up seeing... it is converted into some usable form for them, and that conversion method may differ from action to action) but if they did, they'd find they had the same outcomes over and over.

    Seeding a generator is cheap and easy. Storing generator state is generally not (it depends on the algorithm) so having a persistent state between sessions would be uneconomical. Since re-seeding requires no storage, it's always more economical to re-seed between sessions unless there's a very specific reason not to. Note that "session" in this case isn't user session, generally, it'll be server session for shared generators... which is preferred as it prevents output recording for sequence analysis (not that they'd be getting the raw output anyway, so best of luck to those who try it either way).
    foxman00 wrote:
    Its an interesting test, but we would need several like it conducted across multiple platforms, areas and times to get a decent result to minimise but not elimate the variables to even get a rough indicator of odds.

    No, we don't. We can (and have) already obtained a rough indicator of odds based on the stated set of assumptions. If you want to vomit out a set of new, complex assumptions then that's a fun make-work project for you, but it has very little value and mostly involves constant goalpost moving in an effort to maintain one's comfort with generalized ignorance.

    "...but what if its different on tuesdays... but what if its different if you're a ferengi.... but what if its different if you bought a key in the last 24 hours... but what if... but what if...."
  • davefenestratordavefenestrator Member Posts: 10,511 Arc User
    edited August 2019
    Its an interesting test, but we would need several like it conducted across multiple platforms, areas and times to get a decent result to minimise but not elimate the variables to even get a rough indicator of odds.

    Tests like this (1,000 - 10,000 and recording results) have been done before and reported to this forum and (for an older lock box) to the wiki. The .04% result has been consistent over years of tests now.

    It also fits with human nature where humans prefer to pick a "round number" like 1:250 over 1:197 or 1:281 when setting up an odds table.
  • markhawkmanmarkhawkman Member Posts: 35,231 Arc User
    ltminns wrote: »
    Yes, but years ago in another game there was speculation that the Character's database ID number was used as the seed. That could lead to some Characters being super lucky and others almost never winning.
    This game doesn't have the randomizer client side, and probably uses a single randomizer for the entire game. I personally suspect that re-initializing the RNG is part of what they do on server resets.
    -=-=-=-=-=-=-=-
    My character Tsin'xing
    Costume_marhawkman_Tsin%27xing_CC_Comic_Page_Blue_488916968.jpg
  • oldkhemaraaoldkhemaraa Member Posts: 1,039 Arc User
    edited August 2019
    (Trolling comments moderated out. - BMR)
    Post edited by baddmoonrizin on
    "I aim to misbehave" - Malcolm Reynolds
  • dukedom01dukedom01 Member Posts: 462 Arc User
    After some thought I would like to point out that the Infinity Prize Pack: Kit Modules has diluted to a sucker deal at this point too.

    It's still 'decent' when you go for one of the profession choices, a random roll out of 11 each.

    However 22 different universal modules are in it already, and once the s31 box gets added there will be 5 more.

    Feel free to speculate how the numbers game out for 27 modules you have to gamble on.
    Ceterum censeo Otha supplendum in praemiis.
  • davefenestratordavefenestrator Member Posts: 10,511 Arc User
    dukedom01 wrote: »
    After some thought I would like to point out that the Infinity Prize Pack: Kit Modules has diluted to a sucker deal at this point too.

    It's still 'decent' when you go for one of the profession choices, a random roll out of 11 each.

    However 22 different universal modules are in it already, and once the s31 box gets added there will be 5 more.

    Feel free to speculate how the numbers game out for 27 modules you have to gamble on.

    Perhaps Cryptic should let us pick the kit or trait "family" (like section 31) or even better let us pick the exact one?

  • ltminnsltminns Member Posts: 12,569 Arc User
    Or the Weapons Packs where you can pick type and then Cannon/Beam/Ground. The actual item is random, Omnis being the 'White Whale'.
    'But to be logical is not to be right', and 'nothing' on God's earth could ever 'make it' right!'
    Judge Dan Haywood
    'As l speak now, the words are forming in my head.
    l don't know.
    l really don't know what l'm about to say, except l have a feeling about it.
    That l must repeat the words that come without my knowledge.'
    Lt. Philip J. Minns
  • darkbladejkdarkbladejk Member Posts: 3,712 Community Moderator
    /necro
    "Someone once told me that time was a predator that stalked us all our lives. I rather believe that time is a companion who goes with us on the journey and reminds us to cherish every moment, because it will never come again." - Jean Luc Picard in Star Trek Generations

    Star Trek Online volunteer Community Moderator
This discussion has been closed.