I gotten black 12 times in roll from autopairing in 9x9. These are my pairing settings
This could simply be probability, and not a bug.
2 Likes
I looked at their profile, out of their last 60 games they’ve only been white 6 times, before that it was roughly 50-50.
Does indeed seem odd.
1 Like
Maybe you should go and buy lottery
2 Likes
Probability 4e-11 or 1 in 20,569,864,075
2 Likes
You played 1483 games. I ran 100000 simulations and estimate that the probability to get black 12 times in a row at least once in a series of 1483 games (assuming equiprobability of colors) is about 16%.
3 Likes
That’s the probability that one particular user, chosen before the games happen, gets this run of colours assuming each colour choice is 50:50 and independent. But that is not the event which has happened. The user was selected after the “unusual” event happened. Therefore a more appropriate event to consider would be more like “if 10,000 users are playing on OGS, what is the probability at least one of them gets a run of the same colour 54 or more times in 60 games considering their game settings and pairing pool, notices, and reports it”. This probability is much higher.
This kind of mistake with probability is very common, and was behind several famous miscarriages of justice around infant deaths.
7 Likes
What about an improved pairing which could avoid this situation? If players expect to have more and more chance to get the color looking more and more missing. Pure luck may be not what players expect.
Then we may ask what would be the maximum of the number of successive games with the same color each of us will accept to play with fun?
Most players don’t care about the color. If you really want to play with a certain color, you can always create a custom game.
Not surprising given that most of these games were handicap games against stronger players.
3 Likes
Tada!
5 Likes
I dunno. At least that seems what OP is asking for.
You can’t “always” if you want automatch functionalities.
In the case there’s a bit about squaring the probability and assuming it’s independent.
There’s a bit about it maybe being an issue of conditional probability does. As in given this happens what’s the chance she’s innocent vs given she’s innocent what’s the chance of this happening.
Unrelated to the above:
I do think when it comes to Bernoulli trials like this though and you’re trying to figure out if there’s a bug or other explanation it’s not always very helpful to say well you’ve played 1000 games or 10,000 games were played today and so somebody was going to get this.
It can be, but if you asked what was the chance of getting a double 6 in a dice roll, it’s not helpful if someone says well millions of people have rolled dice in the last year, somebody was going to roll double 6’s.
There has to be a reasonable premise and you can’t pick arbitrary cutoffs (you’ve played 1000 games or 1001 games and get different answers, or 10,000 were played today), unless that is the right question you’re asking about.
In fact, the exact same thing is popping up with Kramnik and his cheating accusations in chess. He would try calculate the likelihood of someone winning 50 or 60 chess blitz games in a row.
I am curious, and I think usually there’s assumptions being made in the calculations which are probably about independence of results of games etc, which might not be true.
Did you count just the ones without handicap? Since we know those will have a determined color 
The api is saying around 762, but I’m not sure how to tell how many of those would be automatch. You could always be playing against an opponent in a custom game that chose a colour
Maybe around 343, if we assume all automatch game names contain the word “vs.”
With a series of 343 games, the probability to get black 12 times in a row at least once is 4%.
1 Like
Sure but we might as well get the numbers right.
I’m not doubting that getting 12 games in a row happens, I’m just curious
A) what’s the right premise/question to ask
B) specifically if you were bug testing, what might tip you off that something is up (streak length, or frequency of certain streaks)
Just genuine interest in the question.
I realised that if you can get the same colour 12 times in a row you can probably pass the glass panel game in Squid Game in one go