Board Game Arena

if abnormal probabilities can never happen, then no-one would ever win the lottery.

You have to take a random selection of games, not specific ones and analyse the results.

The RNG has no idea that doubles mean anything, in fact it doesn't know anything about backgammon or even that it's a dice that it's being used for. This association is all in your head.

Even if it seems fishy, what do you suggest to improve?
You'll look in the code and there is nothing fishy in there. Developers can check most of the code, if there was some hack it would become known at some point.

Ze Monstah wrote: ↑20 April 2021, 22:52And again, that part of the algorithm... again, it is not the full algorithm

What part seems hidden to you?
The game just calls the function bga_rand() everytime it needs a dice result, bga_rand() just calls random_int(), which is a PHP RNG.

Ze Monstah wrote: ↑20 April 2021, 23:06 Game 165874452 (https://boardgamearena.com/table?table=165874452): 10/35 doubles
Game 165943746 (https://boardgamearena.com/table?table=165943746): 13/43 doubles

These 2 games are from the last 13 games, after the last one i posted. So 13 games (11+2) out of last 104 (91+13) are double-fishy.

According to my simulation, there is roughly a 23% chance of finding at least 13 games out of 104 with as extraordinarily large a number of doubles as the least impressive of those games. Combined with the fact that this threshold for unlikelihood was completely cherry picked and the fact that I am not even accounting for the higher variance in the proportion of doubles in a game (as dschingis27 noted), there is almost certainly nothing to be worried about.