Board Game Arena

MsterDestroyer wrote: ↑10 April 2023, 17:39

Tisaac wrote: ↑10 April 2023, 16:28

MsterDestroyer wrote: ↑10 April 2023, 13:17 Really I have experienced it. But only an unique game where the dice rolls stats are like I experienced it is enough for question the system.

The 8 and 6 are the second in the probability table when you roll 2 dices ofr 6 numbers and you sum the results. It is very easy to see that it is quite impossible. Try to repit that.

Another thing to take in account is that I habe only played two games in the last year and one ofthe two games have an improbability result.

I expect this is more than enough to review the dice roll system

It is clearly not enough. Provide stats !!!

You are asking me to observe stats in the same manner than throwing a coin to the air 10 times twice in a year and see that 7 times it was the same face. I dont need to spend the time looking stats for the rest of games.

Sorry but I dont get your example. What are you trying to say here ?
No one hold all the data easily at hand, but everyone that did actually try to gather datas instead of saying "it's obvious that..." always come to the same conclusion : datas are following random distribution. Just search the previous post with similar topics

Well, I suppose i didnt explain well. How many games do you need I review to see if the stats are wrong? 50,100, 1000?

You are seeing that the obvius is that the dices roll well. Where are that stats? I think are the same. Then you are talking about obvius without data?

Explain what you need for example of bad stats...

The dice stats in this game look pretty "reasonable" to me.

Here you have the distribution compared with the normal one. Look the deviation that you can find in the frequency observed.

Given that exactly 34 rolls produced a result in {6,8}, the chance of
number of 6s ​ , ​ number of 8s
being at least as extreme as what you reported is slightly greater than ​ 1/341 .

https://paste.ee/p/UCgqI

Do you have the total number of rolls in each of the games from your charts?

MsterDestroyer wrote: ↑10 April 2023, 21:26 Here you have the distribution compared with the normal one. Look the deviation that you can find in the frequency observed.

And how often did each number appear in these games? How's the totals?

Gulchen wrote: ↑10 April 2023, 21:34 Given that exactly 34 rolls produced a result in {6,8}, the chance of
number of 6s ​ , ​ number of 8s
being at least as extreme as what you reported is slightly greater than ​ 1/341 .

https://paste.ee/p/UCgqI

So a 1 in 341 chance. And over 1 million games of Catan have been played on BGA so results like this are bound to have come up for some people, hundreds in fact, even if the randomisation is working perfectly.

MsterDestroyer wrote: ↑10 April 2023, 21:26 Here you have the distribution compared with the normal one. Look the deviation that you can find in the frequency observed.

Can you add the plot of the average of these 31 games for the number of occurences ?
I am not sure how to read the second graph, it's so small. Can you also just plot the overall standard deviation of all the rolls aggregated ?

The first graph actually looks pretty good to me if I expect such a random law, but let's wait for the aggregated result

https://boardgamearena.com/table?table=365300677

So in the shared game, there was 65 rolls. There was 17 sixies.
The probability to do that or more sixies is 0,645% (my Bernouilli table: https://docs.google.com/spreadsheets/d/ ... 1846324187 )
But now, this claim could have been done for any number. So I will 'multiply' that probability per 11 (1-(1-0.00645)^11), which give 6.87%
I usually do the same for the number of players+1 since you could do that claim for any player or the whole game: (1-(1-0.0687)^4) which give 24.8%

So congrats, if you take any game, and look for the most unlikely number you, another player, or the whole table rolled the most out of what you expect, you would need 4 games to find something more unlikely

To start to change my opinion on that, you would need way more than 4, more like 1k and do it twice, or do proper stats.