Dice Algorithm Sucks



  • Re: Dice Rolls Not Random

    Alright, I have been playing Catan online for years now and I have never been so disgusted and disappointed by the complete mockery of dice patterns. I have played several games within the past 2 days wherein a 6 or an 8 were not rolled once for at least 3 consecutive barbarian attacks. All the while, 11s and 3s were rolled exhaustively. Such inconsistency with probabilities only favors the poor strategists or the gambler. There needs to be better consistency with the dice mode and the roll patterns.

    The old catan online had the option to play with the dice mode Cardstack which I found to have a perfectly balanced distribution of rolls per roll probability. Like a perfect bell curve with 7 in the middle. As such, strategy and placement were paramount. I typically won 60-75% of those games. For CaK auto games, I can't see an option to put Cardstack as a dice mode. I rarely win now and not because my strategy has changed. It almost seems that the CPU picks a winner even before the game begins and favors them throughout with robbers, progress cards, and die rolls.

    Either change the dice patterns or allow me the option to choose dice mode in auto play games!!!!!

    Otherwise I refuse to play on this platform again. It is too remarkably frustrating to see the CPU favoritism toward the weakest, strategic player.



  • First round of barbs on my current game. 2x 7, 2x 11, 1x 3 and 1x 4. Previous game, 10x 11, 9x 3 roll and only 3x 5s. This game has turned into a pure luck game of the dice. So irritating!!!!


  • administrators

    Hi @nemi-asis,

    to verify the validity of the RNG you have to track about 2.000 consecutive dice throws. It is actually very likely that any random 50 2d6 distribution does not look like the perfect bell curve. If you add them up you will see what you expect. That's unfortunately what (close to) true randomness looks like.
    If you keep on playing it will certainly normalize! Thank you very much for playing Catan Universe as much as you seemingly do. Dedicated players like are the ones the keep the community flowing and your feedback is especially appreciated.
    Have you though about funding or joining a guild so that you can play custom matches with the dice mode activated?



  • @Administrator I think it would be a good idea to ad a dice :game_die: mode option to automatch. Even on Playcatan, some players refused to join a game without cardstack. Custom games are still hard to find. :mag_right:



  • Are the dice rolled locally? Is there some weirdness being introduced by having 4 clients roll the dice independently?

    I realize that pseudo-random should mean it doesn't really matter, and patterns we see aren't necessarily betraying statistical probabilities, I just didn't know if this was a contributing factor to the weirdness people think they see.

    Of course, dice are dice and sometimes they just aren't fun. It was more just a random thought, one that I thought perhaps you guys had explored before as you've been working on this longer than us :-)



  • Well, it has happened a lot that the seed provided for the RNG can sometimes repeat so the dice sequence becomes predictable if you find out the seed. Also, the order of cards in the deck can similarly have the same sequence.

    And the previous Catan App created by the same developers did have a pattern which prefers even numbers (10000+ rolls so it can't really be a coincidence)



  • @Stroom But if by off chance more than one client starts with the same seed on a local RNG/local dice, does this result in an increase in repeated numbers? Not so much about knowing the sequence vs 4 random sequences interlaced producing a lot of clustering of numbers? [I know I should just to python and write up a quick test :-P ... or I could play another game, ha ha ha ]

    [Edit: and I do appreciate the developers in so far as tolerating all our questions/concerns, given the challenge of implementing on multiple platforms/fixing bugs/implementing features/budgets/time/etc., ]



  • I think the game vs only AI has local seed (because if the game crashes and starts at a previous point in time, the rolls will be the same).
    In online games, the RNG is probably on the server side. But there could be some way to find out the seed or to hack the game based on some recent bug reports.

    I think there is some game logic run on the client side even in online games and it could be hackable. Based on how some people are in the top of the ELO ladder, it seems like they either found out a way to freeze the game for other players or to bypass the game rule logic somehow.



  • @Stroom @MonkeyZ
    I don´t know about the tech stuff, but I since I play a lot, I noticed that certain numbers have a kind of predictabílty in certain games.
    For example, if 3 and 11 roll a lot at the start, I can predict they will roll often in the rest of the game. That is against all statistic probability, but it is what I have observed.
    Or if 6 or 8 don´t roll much at the first few rounds, they won´t start doing so for the rest of the game. As if the numbers had just shifted, but still certain numbers are luckier than others. You just have to wait a while to see what the lucky numbers in this game will be.



  • Maybe you should write down the numbers and see if there is a specific pattern?

    One confirmed pattern that consistently happened a year ago was this:

    9,7,4,5,10,5,7,5,12,4,12,12,10,7,8,7,5,3,7,10,6,6,5,4,5,2,7,9,9,9,9,9,7,8,2,9,3,3,9,8,11,9,12,9,8,5,8,6,6,12,3,6,10,3,5,11,2,8,12,4,8,8,10,3,3,5,6,5,2,5,8,8,9,8,9,5,3,7,6,7,9,8,4,11,9,5...



  • This post is deleted!


  • @Stroom You are right, but it doesn´t bother me enough to put in so much :sweat: :game_die: effort. I simply try to adjust my strategy.



  • @Rumpelstilzchen said:

    @Administrator I think it would be a good idea to ad a dice :game_die: mode option to automatch. Even on Playcatan, some players refused to join a game without cardstack. Custom games are still hard to find. :mag_right:

    cardstack and friendly robber options for auto match! Players with the same options will be matched. Please!



  • Card stack would also have to follow a RNG seed and that could still keep following a specific pattern. If it isn't already.



  • @Stroom said:

    Card stack would also have to follow a RNG seed and that could still keep following a specific pattern. If it isn't already.

    I thought they had cardstack working fairly well on CatanOnline, it would be RNG on the order but reasonably tight to the proper distribution within the span of a normal game, with a few numbers thrown in/pulled out to make it not 100% to the curve (if I remember it correctly).

    I just want to see it a bit more reasonable. I've had more than enough times where, for example I'll have a 5 that should be common and only have it roll 2-3 times in a game. @administrators (@Developers) are you keeping metrics on the dice rolls in online games? (storing the dice rolls for all the games to watch for problematic patterns/lack of sufficient randomness within the length of a normal game over many real games, rather than just "random over 10 000 rolls" of the algorithm in a standalone test case. I realize our "feelings" on dice often don't reflect reality)



  • The distribution would be better but you can still have the same sequence or numbers. This happens because the shuffling of the deck would still have to be performed by the same RNG that rolls the numbers. And if there is the predetermined sequence, you can abuse it.

    Let's say you know that the sequence will be 6,7,8,8,8 - If you roll the 7, you can easily decide to place the robber on the 8 and block the resource income for other players.


Log in to reply