Catan Is Broken. Unplayable

[ElNSTElN]

The developers of CATAN on BGA could just add statistics for each number (2 - 12). That should easily prove over time that the distribution of numbers is just as expected.

[yfimc]

You clearly don't know how randomness works

There are probably hundreds of Catan games being played on BGA. And a few of them have skewed distributions. Those are what are posted about. People get angry but they don't mention the majority of somewhat OK distributions.

You're being rude and condescending. If I say I know how random numbers work, believe me. I find the games I play on BGA do not match the distribution of rolls of the games I play in real life.

If you do a coin flip 4 times and you get 3 heads, 1 tails or even 4 heads, it would seem very irregular. But in the grand scheme, it is expected to happen quite often. Similarly, a game of Catan lasts for about 70 rolls. Which is about 2 sets of combinations for 2 die rolls. Very few turns so you won't get a perfect bell curve in every game.

70 trials is not very few. 4 is few. 70 is a lot and you should expect the numbers to approximate a bell curve.

And like I said that is not the first game where I got a strange distribution, that's just the one where I finally came to the forums.

Humans believe that the dice roll patterns have to be as close to the average in every game. And that's wrong. People who can't get over it, should play some other games.

These are some wild generalizations.

For sure, over thousands of games, the averages of dice roll results will resemble a bell curve. Enough has been done already to display the fairness of RNG. But every few weeks another person comes and asks us to re-do these tests. It's like hitting refresh and expecting a different result.

I'm not saying it's not fair, I'm saying it's not seeming to behave as though it's random based on my experience in 150 games. And what I'm hearing from folks responding is you have faith in the RNG technology on BGA, and no faith in people's experiences so you blindly trust that the RNG will produce a normal distribution over time, but nobody has actually verified that.

[yfimc]

I'm not sure why the folks who have responded here keep trying to belittle the extraordinariness of the distribution I posted. It's not that 7 and 8 are rolled less than 5, 6 or 9. That's fairly unsurprising. They're rolled less than half as much as 11 and 3!! That is not the same as them rolling less than 5 or 9.

The only thing it tells anyone is that you do not understand random numbers. Unfortunately like many people you use your lack of understanding to try and interpret things - there is nothing odd about such a distribution in this situation because as has already been pointed out to you if you generate enough parcels of random numbers some will inevitably deviate by this amount from the expected values.

I could really dispense with the condescension, thanks.

Of course, if you generate enough parcels some will deviate. If the game I posted was the only game where this has ever happened, sure. But it's not.

[Romain672]

Your game is weird for sure. Problem is it's something you note after the fact, and not before.
On itself, if you look at the most commun complains, per example the number the further from what is expected, you will find two value which are pretty far, but none of the two are extremely far.
Then if you take the opposite, you will find in many games no occurence of 2 or 12, but again, all numbers outside of 2 were rolled enough times to not be weird.

You can totally find a game where 2/5/8/11 will be rolled much more than others, but even if they are all 3n+2, that's just one factor you could look at in the serie between thousands of others.

[yfimc]

Here is some random distributions with 60rolls:
https://www.zupimages.net/up/22/29/xu8k.jpg

these are great, and also, not one of them approaches what i posted

[Romain672]

Per example, the chance of rolling at least eight 3 in 56 rolls is 1,18% ( https://docs.google.com/spreadsheets/d/ ... 1846324187 )

If now you do my fast formula 1-((1-0.0118)^11) to try to multiply my formula by 11, you find 12%.
Which mean in 12% of the games you should have a game where a specific number is rolled more than what is expected compared to your number of 3s in that game.


(edit: forgot a 1, 18% => 12%)

[Stroom]

You clearly don't know how randomness works

There are probably hundreds of Catan games being played on BGA. And a few of them have skewed distributions. Those are what are posted about. People get angry but they don't mention the majority of somewhat OK distributions.

You're being rude and condescending. If I say I know how random numbers work, believe me. I find the games I play on BGA do not match the distribution of rolls of the games I play in real life.

You are not the first person and you won't be the last. And every time I see people like you claim that they understand randomness. And it appears that they have no clue how many biases they have about how randomness "must" work.

This is a common bias that you are expressing. I am pretty sure that you do not track dice rolls in real life games. I have had plenty of games where the same number is rolled 4 times in a row. I have had plenty of games where one number is not rolled for the whole game.

If you do a coin flip 4 times and you get 3 heads, 1 tails or even 4 heads, it would seem very irregular. But in the grand scheme, it is expected to happen quite often. Similarly, a game of Catan lasts for about 70 rolls. Which is about 2 sets of combinations for 2 die rolls. Very few turns so you won't get a perfect bell curve in every game.

70 trials is not very few. 4 is few. 70 is a lot and you should expect the numbers to approximate a bell curve.

And like I said that is not the first game where I got a strange distribution, that's just the one where I finally came to the forums.

70 rolls IS very few for Catan. I specifically used 4 coin flips vs 70 die rolls. Because if you roll 2 dice then you get one full set of all possibilities in 36 die rolls. x2 = 72 so it is comparable to 4 coin flips as one coin flip has 2 possibilities. If you roll dice 36 times, would you get every possible combination? No. 2 or 12 can clearly be missing at such a small set of results. And even with x2 as many rolls, some of those can be missing.

For 2 die rolls, to CONSISTENTLY get a normal distribution, you'd have to roll maybe 360 times. And even then you can get outliers.

Again, this is another explanation of you not knowing how randomness works and that you have the human bias of expecting the randomness to beave in a pattern that you expect... for clearly too small number of observations.

You seem to expect that previous dice rolls somehow influence future rolls too. They don't. Bell curve does not have time to balance itself if you have already had an instance of one number being rolled more than usual. The game usually ends around 70 turns in so you don't have it balancing out.

Humans believe that the dice roll patterns have to be as close to the average in every game. And that's wrong. People who can't get over it, should play some other games.

These are some wild generalizations.

Not wild at all. This is how people who don't understand randomness, often think.

For sure, over thousands of games, the averages of dice roll results will resemble a bell curve. Enough has been done already to display the fairness of RNG. But every few weeks another person comes and asks us to re-do these tests. It's like hitting refresh and expecting a different result.

I'm not saying it's not fair, I'm saying it's not seeming to behave as though it's random based on my experience in 150 games. And what I'm hearing from folks responding is you have faith in the RNG technology on BGA, and no faith in people's experiences so you blindly trust that the RNG will produce a normal distribution over time, but nobody has actually verified that.

I have learned enough about different RNGs, including the one BGA uses, from the previous 10 times I have had this conversation on different platforms. RNG systems are random enough for a dice roll that is rounded down to 1-6 while it could be used for a random number in the billions. You should be much more worried about your physical dice and the way you roll them.

RNG normal distribution capabilities have been tested hundreds of times. You coming here asking us to run the test one more time does not change anything. You can go and test it yourself too if you want to. There are plenty of pages online that use the same basic RNG that most online pages use. It can have theoretical issues only in the general pattern in such a miniscule way that you can't see it for a rounded down version.

All your concerns are human bias talking and it shows. And sorry for being condescending, it is just getting old and no-one has come up with anything new.

[Meeplelowda]

This is a common bias that you are expressing. I am pretty sure that you do not track dice rolls in real life games.

This reminds me of a discussion a couple month back in the backgammon sub-forum where someone claimed certain rolls *never* happened with "real" dice. Well, it just so happens that in real life tournament backgammon they do track the rolls (just as you record moves in tournament chess) and not surprisingly the "anomaly" was also found in recorded games using real dice.

[Jellby]

70 trials is not very few. 4 is few. 70 is a lot and you should expect the numbers to approximate a bell curve.

[...]
They don't. Bell curve does not have time to balance itself if you have already had an instance of one number being rolled more than usual. The game usually ends around 70 turns in so you don't have it balancing out.

Just a minor nitpick: the limit distribution of 2-dice rolls is not a bell curve, it's a perfect triangular shape; no "round" top, no long tails, just a trivial 1-2-3-4-5-6-5-4-3-2-1 shape. If it's a bell, it's a bell like this: https://fortandfield.com/products/5-1-4 ... ass-finish

[Cile]

The developers of CATAN on BGA could just add statistics for each number (2 - 12). That should easily prove over time that the distribution of numbers is just as expected.

I don't know if this would help or cause more issues. On the one hand, people could see the "global" statistics, but if they think someone has gone through the effort of making the Catan dice not truly random for some reason, despite "RND(1,6)" being incredibly easy to code, they're just going to question the global stats.

On the other hand, considering it is near impossible to get a "normal" distribution of results in 70ish dice rolls it would pretty much be telling everyone at the end of their games "here's how your dice rolls were weird this game!" which would result in more of these threads.
"I look at the stats of game after game and NONE of them follow a normal distribution! Clearly these rolls aren't truly random!"

The developers would be much better off adding an option to use a deck of cards instead of dice.