change the dice generator.

[Romain672]

What I see in many, MANY games is that 1 or 2 numbers are RED HOT!! while the others roll way, way below expectations (like my example above). This is because there's an error in the way they're seeding the random function. Of course these 2 "hot spots" change every game and over 100 games the numbers "look right" but inside any individual game it doesn't "feel right". I think this is the core of the complaints.

When you see a game where 3's and 4's out roll 6's and 8's by a margin of 4-to-1 it COULD be bad luck, but when you see it over dozens or hundreds of games you realize it's just bad programming. Kinda like the person in your game group who doesn't ROLL the dice, they just sorta drop them out of their hands. EVERYONE shouts "roll it again and roll them RIGHT"

Thing is, human is bad for that. It will try to find pattern where there is none.

And I saw it by doing some calculations. If you assume that in most/many games 2 numbers will be rolled much more than what you expect compared to the average, then you already got (edit) 11*10/2 possible duos of numbers which can be weird. (even if you don't count 2/3/11/12, you got 7*6/2 duos)
So if you find that the probability to roll too much 4&8 in your next game was like 1%, then this is totally normal. You got more than 50 duos of numbers you could pick from. Once you take into account all of that, you can very fastly find that anything is normal.

That's pretty easy to check too, you don't have to trust my calculations, just ask in another math forum, they will even be better than me for that.

I checked many individuals games, and I find nothing weird.


Since you seem to understand statistics, just do it yourself, create a new post about your next 10 games, and share the results. I could then simulate 1 or 10 random other sets of 10 games not on bga, and you can see the results. Or if it's done nicely, you will got answer from people which are much better than me.

Or just trust that this is a php function of random, which if it's faulty, should have much more bad consequence than having bad rolls in a board game site.

[Gulchen]

@ URjustSOL :

You of course are correct regarding Rcomer727's comment.
However, regarding the rest of your comment:



I looked at

your game just before your post
and
your games after that

, ​ and I get that these had

96 rolls, game completed
7 rolls, game cancelled
0 rolls, game cancelled
44 rolls, game completed

respectively.


I'm guessing things like ​ 3's and 4's out roll 6's and 8's by a margin
of 4-to-1 ​ occur when the number of rolls is closer to 44 than to 96.
(If my guess is wrong, then: ​ Do you have a table number?)

_For {3,4} vs {6,8} specifically_, I get that the chance of the left pair outrolling the right pair
by a margin of at least 4-to-1 out of 44 rolls total is strictly between 1/10058 and 1/10057.


Romain672 overestimated the number of duos - {6,4} is the same set as {4,6} -
but also was not accounting for you having used _two_ duos: ​ {3,4} and {6,8}.

(There are 2970 ordered pairs of disjoint duos.)

[Henk21]

Or just trust that this is a php function of random, which if it's faulty, should have much more bad consequence than having bad rolls in a board game site.

I'd like to know if random_int() is used. rand() or mt_rand() can be predicted by those who know how to. In other words, they can know and perhaps even influence the dice rolls.

And btw, statistically you can have a game with 40x 2 and nothing is wrong with it because statistically it happens. Doesn't mean anything. Doesn't mean the dice generator is any good.

[Romain672]

And btw, statistically you can have a game with 40x 2 and nothing is wrong with it because statistically it happens. Doesn't mean anything. Doesn't mean the dice generator is any good.

At one point, you need to be real. If you assume that 8x 2 is weird if the game has only 8 rolls and is from a game of someone who talked by the past on the forum (and that's the case already), then if you increase that to 10x 2 and 10 rolls, you would already need investigation. If you increase that to 15 or 20 (and all dices rolls of that game were 2), then something will be wrong for sure.

But the thing is, you need to know where the limit is (it's around 1/10^10 to something which happen frequently), so you increase that to 10^15 for bga who got lots of games going all of the time. Then you increse that further to be much more convinced because of taking any number possible instead of 2, but at one point, you can't increase it anymore.

You can't said, I rolled that dice 50 times, got 50 6 but that dice is maybe not rigged. That's the difference between theory and practice. In practice, even if in all the claims of the world done by all humans one time it is false, we should not care. (edit: you can said it for the theory, and that's the point ahah)(edit²: it's possible too that it is in another reason. Maybe you roll your dice in a way which make it more likely to roll in a certain way which increase the chance of rolling a 6. But still, on an internet website, that would be interesting to investigate, maybe a game started at time X got bugged, and so weird things happen on it, while it doesn't happen in other games per example)

I'd like to know if random_int() is used. rand() or mt_rand() can be predicted by those who know how to. In other words, they can know and perhaps even influence the dice rolls.

Write 'random_int' in the search tab in the top right of that page.

[euklid314]

Write 'random_int' in the search tab in the top right of that page.

... or read BGAs FAQ. Indeed, random_int() is called for random events like dice throws.

[Suoivax]

best and only way to determine if the dice have issues is to aggregate all rolls from every single Catan match on this site..
There is no doubt in my mind that if someone did that, the breakdown would be almost exactly
2 /12- 2.78%
3 /11 - 5.56%
4/10 - 8.33%
5 /9- 11.11%
6/8 - 13.89%
7 - 16.67%
One game will never give you any insight.

What I see in many, MANY games is that 1 or 2 numbers are RED HOT!! while the others roll way, way below expectations (like my example above). This is because there's an error in the way they're seeding the random function. Of course these 2 "hot spots" change every game and over 100 games the numbers "look right" but inside any individual game it doesn't "feel right". I think this is the core of the complaints.

When you see a game where 3's and 4's out roll 6's and 8's by a margin of 4-to-1 it COULD be bad luck, but when you see it over dozens or hundreds of games you realize it's just bad programming. Kinda like the person in your game group who doesn't ROLL the dice, they just sorta drop them out of their hands. EVERYONE shouts "roll it again and roll them RIGHT"

You...... just described EXACTLY how randomness works. It's how real life games go if you care to track them.

And if someone can "cheat"a real life roll, they should be in Vegas. A die doesn't have to tumble to be random, unless the person is picking the side and dropping it with two finger from 1cm high. And even then it will likely skew when it hits.

[Guillem_Itsejam]

I played a Catan game where not even an 8 came up. No one 8 in all game is almost impossible.

[Gulchen]

Do you remember how many rolls there were? ​ ​ ​ As I mentioned earlier in this thread,
the most recent game by the user URjustSOL, as of when I checked ​ had only 44 rolls.
For that many, the chance of no 8s is strictly between 1/721 and 1/720.
No 6s ​ has the same probability, and from the same number of rolls, ​ no 7s ​ is between 1/5 and 1/4 as likely.

[Jellby]

For that many, the chance of no 8s is strictly between 1/721 and 1/720.
No 6s ​ has the same probability

So, with 1100 Catan games played, assuming all were around 44 rolls long, finding one or two games with no 8 or no 6 rolled is kind of expected. If that had never happened you could say the dice are rigged!

[brunajmeller]

Look at this game: https://boardgamearena.com/8/catan?table=666623439

38 TIMES THE NUMBER 8!!!!! only 3 times the number 6. This is insane!