...

[Giordano Bruno 1974]

You aren't ever going to be happy with online backgammon or even games against a computer.

Yep most likely. At least not without having access to significant numbers of history logs to be able to undertake the required analysis.
Another poster upthread stated that computer RNG's simply are not purely random. It remains my experience that this is true everywhere I have tried.

Even on my own PC, running GNUBG I can beat GNU far more times when rolling real physical dice than when using GNU's own RNG. It is what it is.

Then why not stick to that?

[Romain672]

Game No---No Of---Total---%
----------- Doubles -Throws
#109661286----12----50 (24%) => 81,2%
#109671565----4----- 44 (9%) => 82,0%
#109671796----8----- 55 (14.5%) => 35,9%
#109677539----10----41 (24%) => 83,8%
#109696866----9----- 35 (25%) => 85,9%
---#109826928----3----- 42 (7%) => 85,6%---
#109840092----12----41 (29%) => 95,8%
#109849486----8-----36 (22%) => 66,0%
#109844128----9-----64 (14%) => 43,9%
#109856820----5-----43 (11%) => 63,2%
#109859222----7-----61 (11%) => 75,7%
#109851074----8-----50 (16%) => 7,2%
---#109969081----4-----46 (8.6%) => 87,1%---
#109963403----4-----53 (7.5%) => 90,2%
---#109968324----9-----75 (12%) => 65,8%---
---#109965336----5-----57 (8.7%) => 86,3%---
---#109963218----9-----43 (21%) => 54,1%---
---#109981226----9-----30 (30%) => 97,1%---
#110013686----6-----71 (8.4%) => 92,0%
#110012369----18----69 (25%) => 97,0%
#110129592----5-----62 (8%) => 91,4%
#110129835----10----45 (22%) => 73,2%

Oh you gave us stats. Let's analyse them.
So I did some modification to my document, and in short he will do 256 simulations, he will take the closest expected value, and will add the number of simulations of that value, then the value +1, then the value -1, then the value +2... Until he find your number (and he remove half of the occurences of that value)(and start with -1 before +1 if it's pertinent).
That's complicated but in short, it give a number which in average will be exactly 50% and will have nearly completely random value. More that number is low, and more you are close to the average. More it's high, and more you are far from the average.
And let's see what we get.

Ok so the result is on the citation, and these numbers are weird (74,6% in average), so let's see how you get these number. Bga is nice, you can get history
Apparently, you take in consideration all conceded games, expect 2 (I put the one where there was a concede with '---' ).
https://boardgamearena.com/table?table=109843364 - 3/12 - 39,8%
https://boardgamearena.com/table?table=109842375 - 4/24 - 12,1%
I suppose you didn't add them because there wasn't enough data, but it's two games which are totally what you expect.
The average goes down to 70,5% which stay enormous.

I can't take random numbers and compare it with that, because of how I find my pourcentage, there is 0% chance to have a number lower than 5% because the most likely occurence will happen at least 10% of the time. (I did anyway, and I needed 20*256, which is 5k simulations to find the first probability of above 70%.)
So ouch it was pretty unlikely.

But never forget you start to take in consideration double when you begin to see it. So this mean the data is slightly biaised.
And your post was in the 6th september, and you stopped your data at the 2nd september. I dunno if you stopped on purpose or not, but if yes, there is another biais there.

[CaractacusPots]

I dunno if you stopped on purpose or not, but if yes, there is another biais there.

I just got bored of having to manually look at each game in turn instead of being able to do a simple export of my game histories. I would happily analyse all my games but doing it manually is just a non starter. There needs to be a game export function.

[euklid314]

Well no need to comment on that since you have already found out your truth.

If you roll 5-5, 6-6, and 4-2 in your moves 3,4,5 and the computer analysis calls that a lucky streak, then you answer that you have skillfully prepared for such a sequence (in moves 1+2?).
Note that the probability of you getting a 5-point board within only three moves (while your opponent plays only correct moves and does not "help you") is well below 1 in 10.000 = 0.01%.

If your opponent enters 5 out of 7 times against your 5-point-board the dice must be rigged.
Note that the probability of him achieving this feat is exactly (7 over 5)*(1/3)^5*(2/3)^7 = 0.5%.

... and now I will work very, very hard on myself not to continue this rather fruitless discussion. I think we both made our points.

Correction: The probability is (7 over 5)*(1/3)^5*(2/3)^2 = 3.7%

[Jest Phulin]

His odds are 1 in 3 to produce a 2.

Dunno why I'm being picky on this when you aren't listening to anyone else and only providing hand-picked oddities to support your claim, but...
You are incorrect in this statement. There are 11 ways out of 36 to produce at least one 2. While very close to 1 in 3, it is not exactly 1 in 3.

The system giving him that 2 on 5 occasions out of 7 attempts is the issue and doesn't help anyone have more faith in the RNG here imo.

Quite the opposite. If it had been done twice, then the argument could have been made that the RNG must have a memory in order to produce the aggregate expected result in such a small sample. If it had been done 500 out of 700 times, then we might question it.

Just doesn't happen with such frequency imo.

May I see the logs of your physical games and recording every dice throw? If you have not kept those, then the well-documented condition of observational bias is a more likely explanation.

[Cappie]

I do have to say... that following this discussion I wanted to play a game of backgammon for the first time in a while.
And of course, during this one game there were also a lot of doubles in the end.
Which, of course, doesnt mean anything is wrong with the way the dice are being rolled by BGA.
It mostly means I had this discussion in mind playing the game. And so I was biased.

[CaractacusPots]

Latest bit of BG fun

These rolls just happened

2-1, 5-5, 2-1, 2-1, 2-1,


It's random Jim, but not as we know it

[Jest Phulin]

It's random Jim, but not as we know it

Showing perfectly well how little you know randomness.

[Romain672]

I often see some french 'Among Us' streams, and lots of people said that X is too much impostor.

It really show how bad humans are to understand randomness.

[CaractacusPots]

More BGA "randomness"

Latest game 4 consecutive dice throws,

6-3, 6-3, 6-3, 6-3,

That's a 1 in 5,832 chance of happening

Of course it dosn't matter because here anything can happen as it would not be random if it didn't right ?!!! lol

So 10 consecutive dice rolls all the same would be just fine ! In BGA world . . . .

It's random Jim, but not as we know it