The short summary of the situation was that both players were using automatic rolls and moves, a situation that was most likely to occur during the bearoff, when players had toggled double, automove, and greedy.
I did some back of the envelope calculations, and predicted that if my hypothesis was indeed correct, an analysis of a very large number of FIBS rolls would turn up approximately 16.83% doubles, rather than the expected 16.67%. (Indirect measurement was the only route possible, since it's not feasible to monitor individual games to determine board positions, etc.)
The results were inconclusive... doubles showed up at a rate of 16.71%, which is higher than expected, but far lower than my hypothesis predicted.
It should be made clear that even if there is a very slight bias in the dice, this bias would affect all players equally, and would not lead to situations that are unfair for any player or group of players.
The rec.games.backgammon article that started it
The outcome of five million rolls
Badly-written PERL script to print a dice report
An article by Gary Wong on 'unusual' events with online dice (added 1/11/1999)
You are the visitor to this page.