- something rotten in Denmark?

Discussion in 'General Discussion' started by LoserSlick, Apr 20, 2017.

  1. LoserSlick

    LoserSlick Devotee of the Blood Owl

    let me be clear from the outset: this is not any kind of veiled threat or criticism with any degree of malice whatsoever. quite the opposite, this is me providing feedback to indicate a problem to help improve a game that i enjoy playing. god that sounds corny

    upon my returning to this game i think i had gone some 27 extended set packs with absolutely zero hits. it then dawned on me that i'd had a long line of very similar results long before said return. i'll admit i was a little frustrated when i posted what i believe was some kind of 'public' criticism to poxnora's help function mechanism - i even went so far as to offer calculations on other players experiences in this regard who may have had similar issues. following that complaint: the very next box i purchased, for the first time ever, i hit the mother load (or comparative mother load, rather) - two exotic and two legendary runes.

    the box i purchased after that was not as great though still better than any previous experience - again, comparatively.

    tonight, i purchased another box and appear to be back in the zero hits zone. moreover, the single guaranteed exotic was one of the oldest and lowest valued in the entire game.

    i'm not going to go and dust off the ol' probabilities and combinations textbook, but clearly somethings... goofy. like mathematically goofy. well really mathematically, logically, and chronologically goofy.

    prior to coming back to the game, much of a very similar purchase experience had left a bad taste in my mouth. i can't be certain, but i'm sure this undoubtedly played a factor in my initial poxnora hiatus in the first place.

    while i did drop 50$ into that '27 pack' figure, i'll admit the rest of it was through campaigning. i can see how this might frustrate a gaming company - though, consider: i think it's very possible the aforementioned 'bad taste' in my mouth might be a leading factor as to why i lack the constitution to drop money into this game on a more regular basis. in my defense, the reason i campaign so much is because it doesn't require very much brainpower and is an alternative to playing minesweeper while i listen to my lectures.

    if my personal purchase experiences are simply an anomaly (which, again, is highly improbable, mathematically speaking--like reeeeally improbable to the point of absurdity if i'm being completely honest) than you might consider finding a better means to make yourself aware of them when they occur. i make a very decent living and would love nothing more than to give you my money. but i don't want to be gamed you realize. i can't imagine anyone who would.

    your pricing models are flawed basically is what i'm getting at. this constrains your playerbase which in turn limits future possibilities. you guys have done a good thing here and poxnora shouldn't be a niche game. dream bigger!
     
    super71 likes this.
  2. themacca

    themacca Master of Challenges

    Packs are about to be redone. it should be easier to justify dropping dollaridoos once the new packs come out.
     
    Kampel and LoserSlick like this.
  3. Hierokliff

    Hierokliff I need me some PIE!

    I know how you feel, i used to buy one or two boxes earlier, but after coming back and noticed that the store/packs are still the same i havent bought any. just with the gold from daily skirmish games. but i do really love that gold, the free runes and the possibility to use the forge to craft almost any rune now. its such a great improvement compaired to when i left. The packs are being remade as @themacca said above, with the chance of getting exotic/legendary improved with a new formula that sounds better. so atm iam hording my gold and just use runes to craft new runes.
    slaying a bunch of skirmish games while watching a movie/tv gives me a few hundred shards per day, i did write it down in excel for the first 600games and i did average ~17shards/game, somedays i did only get common/uncommon and others i could get 2 legendary and a few exotics in the same day.
     
    Last edited: Apr 21, 2017
  4. Thbigchief

    Thbigchief I need me some PIE!

    - No silly...it's "prolly" the next box that will balance out the odds.... this is how the model works bud. All kidding aside unless you opened a larger sample of boxes it just "feelsbadman" as opposed to rotten in some Danish hell hole
     
  5. BurnPyro

    BurnPyro Forum Royalty

    denmark?
     
  6. LoserSlick

    LoserSlick Devotee of the Blood Owl

    sorry but that is not a rational - you can make that same argument of any sample size. 27 rares not hitting on ten percent is more than an outlier. moreover, hitting 2 five percent legendaries on four ten percent exotics immediately following my complaint, while very much appreciated, is almost exponentially more suspect. -which of course ended in another statistically improbable downward trajectory back to zero hits land.
     
    super71 likes this.
  7. Sokolov

    Sokolov The One True Cactuar Octopi

    This is a common issue people have with randomness.

    So let me try and explain why the results are not so surprising.

    First, we have to understand that randomness is NOT evenly distributed.

    upload_2017-4-21_16-45-1.png

    Image A is more random than Image B, while Image B is more evenly distributed, but people, looking at Image B, will feel it is random or fair.

    In fact, it is described as follows: "The two images above model different natural distributions. Image A is supposed to look like the night sky – the dots are stars, randomly distributed. Image B models a colony of glowworms on a cave roof, which looks much like a star-filled sky, but have an underlying organization."

    [​IMG]

    Another one, where the bottom seems unfair, but is actually more random, while the top is something someone would make up as a "good shuffle" for their MTG deck, but it's a lie - you only get a distribution like that if you cheat.

    The source of the image is a Magic article discussing deck shuffling and talks about similar things that I will be discussing.

    ~

    For example, if you ask someone to make up some coin flips, you are likely to get something along these lines:

    HTTHT THHTH

    Where there are very few "runs" of Heads or Tails - because humans think of 50% this way. But the reality is that a probability of 50% converges on 50% in large samples, not small ones.

    But the reality is that you are just as likely to get HHHHH as you are to get TTTTT, HTHTH or THTHT because each coin flip is INDEPENDENT of each other and doesn't care about the results of the former.

    ~

    You really can't make that argument of any sample size. Sample size ALWAYS matters when it comes to probability and 27 is just not enough. Allow me to demonstrate.

    In Pox terms, each pack may have 10% chance, but it doesn't care at all whether you got one before (which might be a design change we make, where we do guarantee an exotic if you don't get one after a certain number of tries, but that's a different discussion which I will talk about in a different post).

    So in your case, you might think it should look something like this for 20 packs:

    RRRXR RRRRR RRRRR RRRXR

    But the probability of anyone getting 27 packs in a row without an exotic is not 0 or even extremely small. In fact, it's roughly 6% (0.9^27).

    What about the probability of getting X in 27?

    0 in 27: 6%
    1 in 27: 17%
    2 in 27: 25%
    3 in 27: 23%
    4 in 27: 15%
    5 in 27: 8%
    6 in 27: 3%

    If we double the number of packs to 54 (also doubling the Exotics just to compare "expected values")?

    0 in 54: 0.3%
    1 in 54: 2%
    2 in 54: 6%
    3 in 54: 11%
    4 in 54: 16.3%
    ...
    6 in 54 (which is the "same" as 3 in 27): 16%
    ...
    8 in 54 (which is the "same" as 4 in 27): 8%
    ...
    12 in 54 (which is the "same" as 6 in 27): 0.004%

    (I am not going write out the math in each case, but this is done using what is called a Binomial Distribution.)

    This means that out of 100 people who buys ~27 packs, approximately 6 of them will get 0 Exotics, while 17 of them get 1 Exotic. So out of 100 people, almost 1/4th will get something less than what they would expect out of fairness.

    Notice how even just doubling the packs makes the likelihood of 0 out of 54 to basically nothing, this is because we have increased the sample size to a large enough point where we have converged on the mean more, a property known as The Law of Large Numbers. Increasing the sample size changed something from 6% to just 0.3%. That's a dramatic change, I think you will agree.

    So is it uncommon to get 27 out of 27 no exotics? Sure. Is it actually mathematically wrong or extremely unlikely? No.

    And when you buy a box that had 2 EX (one of which was guaranteed) and 2 LEG, well, that's well within the range of "average" and certainly not anomalous.

    Specifically, if we "pretend" that you basically got 5 random EX, and plugged that into the calculator has 10% chance of success.... we get 5 out of 30 at a 10% chance.

    And if we take it as 5 out of 57... that's 17.5% chance of happening. Along with 4 out of 57 being 15% and 6 out of 57 being 17% you basically had almost a 1/2 chance of getting that kind of pull from that sample.

    You might say, "it's still weird I got nothing initially and stuff later!" But again, that's because we are humans and like evenly distributed even though randomness doesn't care. In fact, this type of "averaging out" has a technical term called Regression toward the Mean.

    We can do more specific math on this too (since the LEG has a different probability chance and is a second order calculation), if we'd like, but I think the above numbers will suffice as a general indication.

    Now, this doesn't mean nothing weird isn't happening, there might be a problem but the specific data provided is not really demonstrative of an issue.
     
    Last edited: Apr 21, 2017
    Etherielin, Tweek516 and Hierokliff like this.
  8. Sokolov

    Sokolov The One True Cactuar Octopi

    Now, that's the math, but is this good? It's certainly a model that card games have used with success, but it certainly doesn't feel good. One thing that can be done (and several games do this) is modify the chances if you aren't getting anything. For example, we might fill a meter that says "guaranteed exotic in your next pack" which resets as soon as you get one, but when filled, your next one is guaranteed to have one and set this to a value of 10. This means that your MINIMUM chance will be 10% but you are likely going to get something slightly higher in the long run. This has almost no revenue impact, but makes the pack buying experience much nicer (so in fact might positively impact revenue).
     
    Last edited: Apr 21, 2017
    Kampel, Etherielin and Fentum like this.
  9. DiCEM0nEY

    DiCEM0nEY I need me some PIE!

    Try leveling alts, you can get alot of shards. Beats paying for things.
     
  10. super71

    super71 I need me some PIE!

    In other words sok try's to fool people again with a long list of hog wash.

    The boxes are broken and will award you Bane Shift when buying with gold and decent stuff when buying with cash. It's been this way forever and really hasn't change with the exception of the forge and a more f2p model, granted the f2p model isn't the best because we're drawing from a pole of thousands of cards. Paid boxes award better Bane Shift to get you to keep paying, it's the same thing in hearthstone.

    The new pack system sounds promising so best to hold onto your gold.
     
  11. Sokolov

    Sokolov The One True Cactuar Octopi

    What, specifically, do you consider hogwash in what I said? It's just math.

    Note that there is no difference in the packs whether you get them via gold, credits, or system gift. In the database, we create one pack with the parameters, and then assign costs (we can even make them cost Shards, if wanted). The fulfillment system, once it goes into "deliver pack" mode, doesn't know how it was obtained and doesn't care - it just runs thru the math the pack is built with. Same with getting cards from campaigns or after ranked matches - those are basically the same system.

    Also, note that OP's 27 pack figure of no exotics appears to have been packs purchased with $$$:
    So it's not exactly evidence that there is a problem with gold packs. In fact, it's evidence against this assertion - since his $$$ purchased packs yielded 27 packs of nothing.

    @LoserSlick, do let me know if I have misunderstood you, but I believe you are simply discussing the inconsistent nature of the randomness and are concerned that it might not be working properly (whether by chance or by design). In any case, I hope I was able to sufficiently demonstrate that mathematically what you experienced is some kind of super unlikely event that necessarily indicates a problem.

    That said, if anyone has been tracking gold vs credit packs and truly believes there is a problem instead of just what appears to be an unfounded speculation from @super71, I'd be interested in looking at that data and seeing if there's actually an issue to investigate.
     
    Last edited: Apr 21, 2017
  12. LoserSlick

    LoserSlick Devotee of the Blood Owl

    lol see this is what i didn't want. honestly would not have at all minded if this thread would have been closed after the first response being that i was unaware of any upcoming changes to purchasing in the marketplace. perhaps there was a different private sub forum i could have posted this in (i apologize for that). but chief poked me! = his fault :)
    (sorry - had to respond to em)
    of probability and randomness: the point is that i'm still an outlier in terms of any relative frequency. for arguments sake, twenty seven packs in addition to "similar past experience" would lower that 6% figure even further, presumably (or even the 1 in 27: 17% figure).
    i do understand anecdotal criticism in terms of probability is silly given the premise of sample size and that i have that going against me. but on the grounds that i have no evidence of controls and you do (or may) the point is that i griped about a possible statistical anomoly out of frustration only for that to be followed up with an even more improbable one - which i then would have presumably taken as validation in spite of whether or not i had been statistically justified formerly.
    that is to say...
    for arguments sake (not for the marginal gains, but to simplify), let's round those 7 individual packs to ten--4 sets of ten packs totalling 40. sampling four sets of ten subjects from the same pool, what is the probability of the first three coming back with zero hits (zero x at 10%, and zero y at 5%), followed by the probability of the fourth resulting in 4 hits (2 x and 2 y). the fourth pack alone is an anomoly in and of itself let alone in conjuction with the first one (even if not figured in terms of sets insofar as how they are actually purchased - not sure why i chose to think about it that way). lol but regardless obviously that is an argument that is certinaly not in your favor - at the same time, however, that is not a fair argument on my part at this point all things considered. sorry - it's late and my head hurts (also i don't have a calculator on me) - but even 4 or more hits in the fourth in general: 0.85^10, 0.85^10, 0.85^7, 0.1^2*0.85^6*0.05^2. might want to double check that but i think i'm close
    lol that said, i don't believe there is a mathematical explanation for what's just happened here ;)
    but there is a logical one.
    either way, whether it was a deliberate cautionary function or a human response to the red flag i threw up, it's not something you guys aren't aware of. which is good :)
    lol in the meantime, could you modify me a couple more legs??
    just kidding!
    (seriously tho)
     
  13. LoserSlick

    LoserSlick Devotee of the Blood Owl

    p.s. i reserve the right to refigure those numbers on a day in the future of my choosing - when i'm awake and have coffee in me.
    long story short, given the assumed inclination of your average player to opt for the box instead of individual packs (for the guaranteed exotic), current probability models easily have the potential to make for an unpleasant consumer experience.

    also this:
     
  14. LoserSlick

    LoserSlick Devotee of the Blood Owl

    sorry man. i really didn't mean to dump a complaint on you in this way. i hope it came off as more of an inquiry than a complaint at least initially :/
     
  15. themacca

    themacca Master of Challenges

  16. LoserSlick

    LoserSlick Devotee of the Blood Owl

    oh wait. i do have a minor correction to make:

    concerning the wording in my second post, i seem to have conflated boxes with individual pack probabilities
     
  17. Sokolov

    Sokolov The One True Cactuar Octopi

    What did you want then? I would have thought that you wanted a reply from someone who had the power to investigate the issue?

    The box having 2 EX and LEG is actually very likely, as I demonstrated, rather than "even more improbable."

    So... if I am understanding, you are asking about a hypothetical example of:
    1. 10 packs of 0 EX
    2. 10 packs of 0 EX
    3. 10 packs of 0 EX
    4. 10 packs with 4 EX?
    Well, this is dramatically different, as previously you had 30 packs with 4 EX, so this is EXTREMELY unfair. But let's do it anyway.

    10 packs of 0 EX has a rate of resulting in 0 EX around 35% of the time - every 3 groups of 10 you buy will have over 1/3rd chance of not having any EX. This is quite high due to SAMPLE SIZE.
    30 packs of 0 EX has a rate of resulting 4.2% of the time.

    Not much different than the 6% for 27 since we are adding just 3 more packs.

    The last pack, 4 out in 10, is the really crazy one - with just a 1.1% chance of happening.

    In total though it is 40 packs with 4 EX, which has a 20% chance of occurring - which is pretty much expected.

    Of course, the reality is you bought a BOX, with a guaranteed EX and having 30 packs instead of 10, which would be 60 packs with 4 EX, which is 13%.

    (Again, we could do go deeper with the LEG calculation, which would put the %s slightly higher, but this suffices, I think.)


    I literally gave you a mathematical explanation :(

    What about it did you not understand? What part of the explanation doesn't make sense?[/QUOTE]
     
    Last edited: Apr 22, 2017
    Lushiris and Tweek516 like this.
  18. LoserSlick

    LoserSlick Devotee of the Blood Owl

    bid's up!

    lol wow thanks man. but yea it definitely sucked

    i'm going to try and avoid putting sok in a tough spot here, so i will simply say that he did address an awareness of outliers, which is a good thing. at the end of the day, whether it was human or algorithmic, that fourth box did at least somewhat compensate those previous 27 packs. and the fact that they're implementing something different is probably a good sign
     
  19. Sokolov

    Sokolov The One True Cactuar Octopi

    Feedback/discussions are always welcome! :)

    I am not sure what you mean by putting me in a tough spot - I am simply explaining the math to demonstrate that there doesn't seem to be something goofy going on, but you seem convinced there's something wrong with what happened from a mathematical perspective, so I am not sure what the issue is.

    (We both agree that the system can "feelsbadman" but that's a separate discussion as to whether it is working as described or normal from a math perspective.)

    Which is exactly what you'd expect in the long run, as I explained. It's how probability works. You look at the 27 packs as a group because they fit some pattern to your human brain, but randomness literally doesn't care. You can cherry pick which segments to group together, but the more packs you look at, the more likely you are to find that it is closer to the "expected" or "average" while the smaller the groups you select, the more likely you will find something that is not going to match the expectation.

    Note that the pack meter thing is something I'd like to do, but it isn't being developed at the moment. I do think we'll get in the smaller packs (at a lower cost) which mathematically raises the EX/LEG to Gold ratio, however.
     
    Last edited: Apr 22, 2017
  20. Sokolov

    Sokolov The One True Cactuar Octopi

    Made a Google Sheet to demonstrate.

    I generated 100 "packs" with a random value of 1 to 10, with 1 being an EX.

    Test 1:
    upload_2017-4-22_1-3-4.png

    upload_2017-4-22_1-0-7.png

    Test 1 had one run of 35 packs without an EXO.

    Test 2:
    upload_2017-4-22_1-3-19.png
    upload_2017-4-22_1-0-36.png

    Test 2 had a run of 30 packs without an EXO.

    Test 3:
    upload_2017-4-22_1-3-39.png
    upload_2017-4-22_1-1-32.png

    Test 3 had a run of 28 without an EXO, followed by an EXO, then a run of 20 without an EXO.


    In each case, the % converged around ~10% by the end of 100 packs.
     
    Last edited: Apr 22, 2017

Share This Page