dice algorithm is clearly flawed

Forum rules
Please DO NOT POST BUGS on this forum. Please report (and vote) bugs on : https://boardgamearena.com/bugs
Knibbeltje
Posts: 6
Joined: 07 September 2026, 22:05

Re: dice algorithm is clearly flawed

Post by Knibbeltje »

This game: https://boardgamearena.com/table?table=944675465

-Just one 6 in the last 30 turns
-Six 2's, but no 3's at all in 49 turns
- Ten(!) 11's

The game before that: https://boardgamearena.com/table?table=944658976

twenty turns without a 5, 6, 8 or 9

Honestly, I know nothing about statistics (so correct me if I'm wrong) but it seems to me this can occur sometimes, but i experience these statistically ridiculous sequences all the time. I can keep posting links to games..
User avatar
Jontia
Posts: 268
Joined: 06 March 2014, 22:59

Re: dice algorithm is clearly flawed

Post by Jontia »

Knibbeltje wrote: ↑08 November 2026, 14:42 This game: https://boardgamearena.com/table?table=944675465

-Just one 6 in the last 30 turns
-Six 2's, but no 3's at all in 49 turns
- Ten(!) 11's

The game before that: https://boardgamearena.com/table?table=944658976

twenty turns without a 5, 6, 8 or 9
These things are not even very unlikely.
1 six in 30 turns has a one in 18 chance.
Six 2's in about one in 500.
No 3's is about one in 17.

You're picking these after the fact too, so really that no threes is a Or condition of no 3's, no 11's or any other whacky thing.
The chances of any individual game having something 'odd' happening are actually really high as long as you don't specify what 'odd' means before you start looking.
User avatar
Simsat
Posts: 1
Joined: 20 June 2026, 08:17

Re: dice algorithm is clearly flawed

Post by Simsat »

Goforit22 wrote: ↑30 October 2026, 17:49 I completely agree with poster that a game of Catan can extremely frustrating due to biased dice. Its probably also true that the dice is random, but for many games the dice is still the deciding factor.

I dont get why BGA havent implemented a card deck dice generator. Imagine dice determined by a number of card decks, number of dice sets determined by their probability. When the deck(s) is depleted, it start over. Will still be randaom, but much closer to a normal distribution for every game. Will put strategy in the high seat, not lucky dice.

It will take 10minutes for a AI to write code, max. I can provide it :-)
But there is the variant for using a dice deck. Why dont you use it?
User avatar
euklid314
Posts: 718
Joined: 06 April 2020, 22:56

Re: dice algorithm is clearly flawed

Post by euklid314 »

Knibbeltje wrote: ↑08 November 2026, 14:42 This game: https://boardgamearena.com/table?table=944675465

-Just one 6 in the last 30 turns
-Six 2's, but no 3's at all in 49 turns
- Ten(!) 11's

The game before that: https://boardgamearena.com/table?table=944658976

twenty turns without a 5, 6, 8 or 9

Honestly, I know nothing about statistics (so correct me if I'm wrong) but it seems to me this can occur sometimes, but i experience these statistically ridiculous sequences all the time. I can keep posting links to games..
Because you state that you do not know a lot about statistics I will briefly explain:
When rolling 2 regular dice the probability of sum 2 is 1/36, sum 3 is 2/36,..., sum 6 is 5/36, sum 7 is 6/36, sum 8 is 5/36, ... sum 12 is 1 /36.

That is, the expected value for 36 consecutive rolls is 1 2s, 2 3s, 3 4s, 4 5s, 5 6s, 6 7s, 5 8s, 4 9s, 3 10s, 2 11s, 1 12s.

But the expected value does not mean that this outcome is likely. Quite on the contrary! If you roll 36 times the above result will "never" happen. It is as unlikely as winning in (several) lotteries.

The reason is that all these dice rolls are independent of each other. If you roll two 8s in the first 5 rolls then the dice do NOT give you less 8s in the upcoming rolls to even out what has happened in the past.

Thus, for such a short sequence of 36 rolls, it will almost always happen that one number is hugely overrepresented and other numbers are hugely underrepresented. But since you cannot predict in advance which numbers that will be and since it will not always be the same numbers, the process is still perfectly random.

Just try it out with some ordinary dice at home. Roll them 36 times and note the results. Then roll them another 36 times. And a third time. It is extremely likely that you will get several extreme results. And even after 3x36=108 rolls some numbers will definitely be hugely overrepresented. That is just a normal behavior with random, independent dice.
jollyCrocodile
Posts: 3
Joined: 08 July 2026, 23:33

Re: dice algorithm is clearly flawed

Post by jollyCrocodile »

Jontia wrote: ↑08 November 2026, 14:59
Knibbeltje wrote: ↑08 November 2026, 14:42 This game: https://boardgamearena.com/table?table=944675465

-Just one 6 in the last 30 turns
-Six 2's, but no 3's at all in 49 turns
- Ten(!) 11's

The game before that: https://boardgamearena.com/table?table=944658976

twenty turns without a 5, 6, 8 or 9
These things are not even very unlikely.
1 six in 30 turns has a one in 18 chance.
Six 2's in about one in 500.
No 3's is about one in 17.

You're picking these after the fact too, so really that no threes is a Or condition of no 3's, no 11's or any other whacky thing.
The chances of any individual game having something 'odd' happening are actually really high as long as you don't specify what 'odd' means before you start looking.
I always find, after the fact, that the winner was unfairly favoured!
jollyCrocodile
Posts: 3
Joined: 08 July 2026, 23:33

Re: dice algorithm is clearly flawed

Post by jollyCrocodile »

By the way mathematically interesting is the effect of end of game:

For example say some one is very strong on 2 (not the best choice except for in this game): 2 rolls early and rolls say three times. This give the player such an advantage in the early game that they can win quickly. The result is 2 rolls disproportionately because there were not enough dice rolls following on to normalise for these early 2s
User avatar
Paddles
Posts: 112
Joined: 25 March 2012, 12:04

Re: dice algorithm is clearly flawed

Post by Paddles »

jollyCrocodile wrote: ↑10 November 2026, 02:49 By the way mathematically interesting is the effect of end of game:

For example say some one is very strong on 2 (not the best choice except for in this game): 2 rolls early and rolls say three times. This give the player such an advantage in the early game that they can win quickly. The result is 2 rolls disproportionately because there were not enough dice rolls following on to normalise for these early 2s
That's not mathematically interesting, it's the expected outcome. As euklid314 correctly pointed out earlier, subsequent dice rolls are (or should be) independent (often expressed as "the dice have no memory"). If the dice are fair, then we shouldn't expect the accumulated results to revert back to the theoretical distribution after an unusual starting event (*). Similarly, if you're flipping coins and you happen to start with 18 heads in a row, then returning to an accumulated result of 50% heads, 50% tails is statistically unlikely, not almost certain. The coin (or the universe) isn't sitting there thinking "we've got lots of heads, we need to start landing on tails more to balance things out". Making sure the results even out in the end comes from adding bias, not eliminating bias.

(*) Unless you're rolling dice for an incredibly long time. If we're rolling thousands of times or more then it becomes more likely that the impact of those initial unusual rolls disappear. But it's not something we'd expect within the number of rolls usual in a game of Catan.
User avatar
Coleus1
Posts: 20
Joined: 22 April 2025, 14:51

Re: dice algorithm is clearly flawed

Post by Coleus1 »

Users who are loud to defend the "random" BGA Catan dice do not even play the game :lol: the game is also much more than dice rolls over a number of games.
Paddles wrote: ↑06 November 2026, 12:59 So how can you tell that it really was biased towards a player, rather than the simpler hypothesis that one of the players happened to be lucky in the first few rounds with a fair random number generator?
By playing the game a lot both on BGA, other sites and real life and noticing how easily one can spot the favored player after first few rolls on BGA Catan? When was the last time you played BGA Catan?

Create an agent on Claude that analyzes the game for you live and you'll see.
User avatar
Suoivax
Posts: 397
Joined: 13 April 2022, 00:27

Re: dice algorithm is clearly flawed

Post by Suoivax »

Coleus1 wrote: ↑10 November 2026, 10:04 Users who are loud to defend the "random" BGA Catan dice do not even play the game :lol: the game is also much more than dice rolls over a number of games.
Paddles wrote: ↑06 November 2026, 12:59 So how can you tell that it really was biased towards a player, rather than the simpler hypothesis that one of the players happened to be lucky in the first few rounds with a fair random number generator?
By playing the game a lot both on BGA, other sites and real life and noticing how easily one can spot the favored player after first few rolls on BGA Catan? When was the last time you played BGA Catan?

Create an agent on Claude that analyzes the game for you live and you'll see.
You're right, I dont play Catan on BGA. I'll defend the dice because I know how computer programs and numbers work.
User avatar
Paddles
Posts: 112
Joined: 25 March 2012, 12:04

Re: dice algorithm is clearly flawed

Post by Paddles »

Coleus1 wrote: ↑10 November 2026, 10:04 Users who are loud to defend the "random" BGA Catan dice do not even play the game :lol: the game is also much more than dice rolls over a number of games.
Paddles wrote: ↑06 November 2026, 12:59 So how can you tell that it really was biased towards a player, rather than the simpler hypothesis that one of the players happened to be lucky in the first few rounds with a fair random number generator?
By playing the game a lot both on BGA, other sites and real life and noticing how easily one can spot the favored player after first few rolls on BGA Catan? When was the last time you played BGA Catan?
You didn't actually answer the question of how you can differentiate between a player being favoured vs. a player being lucky in the early rolls i.e. what makes this bias rather than luck - all you did was repeat your original assertion that you can. If your definition of "favoured player" is the post hoc observation that a player was lucky on the early rolls therefore they must have been favoured by a biased algorithm then you need to read and understand about logical fallacies, specifically "affirming the consequent".

I will admit I don't play Catan here as much as you do, but if you'd checked my profile you could see exactly how many games I'd played here and how recently (142, and September). You could also see my history of other games that include dice or similar random selections, including other games where people make the same claims about bias without any evidence, let alone statistical rigour.

The integrity of the underlying RNG used in BGA has been validated by numerous mathematically skilled people and teams; the specific BGA code that uses that RNG for Catan can be inspected by anyone who joins the BGA development program, so any actual coded bias should be relatively easy to prove (and yet, no one has yet, despite numerous claims). But your claim is, so far, an unprovable/unfalsifiable "I know it when I see it". Even if you can't prove it, I'd be happy to see something that could at least be tested experimentally.
Post Reply

Return to “Catan”