Randomness
complete
A
Alfie Kirkpatrick [Developer]
I have done a quick analysis on the last 2,000 or so matches on Hub, looking at the distribution of dice rolls. I often see in the match comments that Hub produces more doubles than expected and/or more than other online platforms. I hope the numbers below show that this is not the case.
Some players also claim that Hub favours one player over another, presumably by adapting rolls according to how a match is going or a player's PR. The pre-determined dice should be enough to disprove this particular conspiracy theory. When you pick A-E at the start of a match you are choosing from a set of pre-determined rolls for the whole match. You can download the rolls to a password protected zip file before the match starts, and verify after the match that the rolls weren't tampered with. If anyone requires more technical detail, I'm happy to provide.
The probably of any single roll is 1/36 or 0.0277777. Below you can see that doubles, particularly 5-5 and 6-6 are right in the middle of the table with the expected probability. Note that some variance around the mean is inevitable, even in what appears to be a large stastical sample. If anyone is a statistician and wants to help me tighten this up, let me know - stats isn't my thing!
matches: 2098
rolls: 532359
2,2: 0.02838 (15107)
5,4: 0.02805 (14935)
2,1: 0.02805 (14934)
4,2: 0.02802 (14919)
4,3: 0.02800 (14904)
6,2: 0.02797 (14890)
5,1: 0.02796 (14886)
1,4: 0.02796 (14885)
3,5: 0.02791 (14856)
1,6: 0.02790 (14854)
6,5: 0.02788 (14841)
2,5: 0.02786 (14833)
5,6: 0.02781 (14803)
4,4: 0.02781 (14803)
4,1: 0.02780 (14801)
4,6: 0.02779 (14792)
6,6: 0.02778 (14788)
2,4: 0.02777 (14786)
5,5: 0.02777 (14785)
3,4: 0.02776 (14777)
1,3: 0.02772 (14759)
1,2: 0.02771 (14753)
6,4: 0.02769 (14739)
3,2: 0.02766 (14726)
1,1: 0.02764 (14715)
3,1: 0.02763 (14711)
2,6: 0.02762 (14706)
6,3: 0.02761 (14700)
5,3: 0.02760 (14692)
1,5: 0.02759 (14689)
4,5: 0.02759 (14686)
5,2: 0.02757 (14679)
6,1: 0.02757 (14677)
2,3: 0.02754 (14661)
3,6: 0.02752 (14649)
3,3: 0.02750 (14638)
B
Backgammon Hub Admin
complete
Superceded by https://ukbgf.com/online-dice-random-or-not/
M
Martin Brown (maartjebabes)
With my (historic!) statistical background, I’ve looked at this.
• The 95% confidence interval (2 standard deviations) in Alfie’s numbers is between 14547 and 15107
• Only double 2 falls outside that. It’s not a shock that in 36 rolls, one or two falls in the outlying 5%. (At 5 you’d start questioning).
• The double 2 is not so far away from the mean to be significant – if you rolled a pair of dice 532359 times, one in every six times one of the 36 rolls would be this far away from the mean. I think every backgammon player would agree that a 1 in 6 chance occurring is within normally expectation!
• The total number of doubles (88,727) is very much within the bounds of what you’d expect.
As for getting 12 doubles in 34 rolls, you’d expect this to happen ~0.2% of the time, 1 in 500. The last 7 pointer I played was 362 rolls. If you look at this as ~11 sequences of 34 rolls o you’d expect that to happen in 1 in 45 matches.
Actually there are 329 such sequences, but they're connected.
I could put a longer explanation if anyone wanted.
S
Soruche Saajedi (dopestick)
last night i had a game with 12 doubles out of 34 rolls. we rolled double 5 three times in a row. the last 4 or 5 opponents have accused me of cheating, i'm starting to feel i am getting more of the luck. can you please take a look at my last few games and make sure i'm not crazy
A
Alfie Kirkpatrick [Developer]
Soruche Saajedi (dopestick) I'm sure your luck will change!
You would expect 6 doubles in 36 rolls on average. I can't work out the probability of 12 or more doubles in 36 rolls but am sure it's not that low.
J
John Pollard 🇬🇧 (johnypoll)
It is a bug in the brain, not a bug on backgammonhub
A
Alex Zamanian (z)
Every backgammon site/program ever made has had people complaining about unfair dice. Therefore people complaining about unfair dice is not evidence that the dice are unfair.