dice used in catan is biased

[Mathew5000]

Are we assuming that the program is right?
In fact, PHP's random generator had many problems.
Are you still using the latest PHP generators and programming properly?
Creating random numbers is very difficult.
We should not assume that our programs are correct, but we should examine them carefully to make sure that we are doing the right thing.

Already discussed extensively on this forum during the last two years. Here's an interesting thread from 2022: viewtopic.php?t=27679 Also lengthy discussion in this "bug" report: https://boardgamearena.com/bug?id=66082

[Jellby]

Also lengthy discussion in this "bug" report: https://boardgamearena.com/bug?id=66082

That's an neverending source of laughs!

"I haven't run any official statistics analysis, but...": so anything after that is worthless
"I got a 1 in 50,000 event": wow! with more than 3 million games played, I guess that proves there's a bug
"Finally a game with truly definitive evidence that the dice are not coded well": finally a truly definitive example of selection bias

[Suoivax]

If only they would release the random function, we could sample it a billion times...
Unless you make it open source, the suspicion will never go away.

there are a billion (ok, maybe not a billion) rolls already recorded. write a script to scrape the data from played games.

[Nekojin]

I agree...my friend and I in the regular board game switched to an app that gives a perfect distribution every 36 rolls called catan dice. After the 36 rolls it starts over so if you end in the middle of the next 36 rolls the distribution will be off some.

Even regular dice in any given game will not give you a good distribution....if you could have a game that had 10,000 rolls you would get a good distribution.

I am fairly new to the bga online platform but from what I see is 11's and 4's roll way more often than they should. EVEN 2'S and 12's are vastly more represented than they should be. A 2 and 12 should only on average be rolled 1 time each in 36 rolls. Games offen have these rolled 2-3 times or more in bga.

The problem with generating a truly random number is you need a way to randomly generate a random number for your algorithm so that it generates a unique number. I watched a video once about a powerball gambling generator that used the pressure of the weather(if I remember correctly) in a given place that was always changing to generate the powerball numbers.

This is why I rate catan (even the normal board game) with probably 15%-25% luck...

So you've got a computerized dice calculator that is guaranteed to roll one 2 and one 12 out of every 36 rolls? That's not random! And that's absolutely gameable, if anyone tracks what rolls have already happened.

People overestimate how un-random randomizer scripts are. If a calculation repeats after every 4096 events pulled*, that is absolutely random enough for a d6 roller, or even a poker card roller - because even if the sequence repeats, it doesn't guarantee that it starts at the same place every time!

*4096 is a ridiculously low sequence for a randomizer. Most are several orders of magnitude bigger.

[Dan Funk]

Random number generator in catan is not random. I can say this with 100% certainty. Why? Because I have played 3 games with ***ZERO*** 7s! I calculated the odds of this happening once by chance as one in 50 million. The odds of it happening 3 times must be staggering indeed.

[twsansbury]

Random number generator in catan is not random. I can say this with 100% certainty. Why? Because I have played 3 games with ***ZERO*** 7s! I calculated the odds of this happening once by chance as one in 50 million. The odds of it happening 3 times must be staggering indeed.

After scraping all of your games, I can affirm the distribution of dice values in your games converge towards the normal distribution of 2d6. The sample is too small to assert random or non-random rolls per NIST's statistical test.

[euklid314]

Random number generator in catan is not random. I can say this with 100% certainty. Why? Because I have played 3 games with ***ZERO*** 7s! I calculated the odds of this happening once by chance as one in 50 million. The odds of it happening 3 times must be staggering indeed.

Can you give the game numbers of these 3 games? It is important to know the number of rolls in these games to truely assess how rare these events are.

Your games have 60 rolls on average. If you take one of your shorter games with only 40 rolls then the probability of having no 7 in a single game is approx. 0,07%. Since you have 2000 games it is not surprising that you have one or two games with zero 7s. Three such games is above average but there must be hundreds of CATAN player with 3+ such games on BGA - with millions of games played.

As twansbury wrote, your dice value distribution is as expected - thus you can safely reduce your 100% certainty to 0%.

[Anthena]

Just got done playing another game of Catan and I am becoming more and more a believer the algorithm is DEFINITELY biased. This particular game the player consistently rolled 8, 9, 10 or 7. Naturally, they won easily in under 10 minutes. Either some players have figured out how to adjust the dice system or they are just EXTREMELY lucky. PLEASE FIX

[Anthena]

I apologize, game time was 17 minutes. I would like to provide evidence of bias OR here are the numbers to strive for when setting up the game.
Dice roll: percentage of 3 (~6%)
6
Dice roll: percentage of 4 (~8%)
6
Dice roll: percentage of 5 (~11%)
8
Dice roll: percentage of 6 (~14%)
10
Dice roll: percentage of 7 (~17%)
17
Dice roll: percentage of 8 (~14%)
25
Dice roll: percentage of 9 (~11%)
17
Dice roll: percentage of 10 (~8%)
8
Dice roll: percentage of 11 (~6%)
0
Dice roll: percentage of 12 (~3%)
2

As you can see, in this particular game, 7, 8 and 9 were the dominating rolls. Coincidence?

[Nekojin]

Evidence? Or just more "feels"?

Edit: The above post wasn't there when I started writing this post. That's 109 rolls in one game. You realize you'd need a LOT more than that to show anything, right?