Statistically, it's not possible to judge whether a dice roll is random or not from a small sample such as a few games. The sample should overcome about 200 dice rolls and the distribution of each dice roll (each should show appproximately the same frequence and therefore randomly distributed in a large sample) should be checked, before claiming it 'non random'. On the other hand, people are convinced that in some cases the server seems apparently to play "against". Two questions: 1) obviously, why should it? 2) can you imagine how powerful a computer should be to control and predetermine dice rolls for each game played? Unless you believe he is supporting only your opponent...but that is paranoia.
Unfortunately, I am sure that people will continue to look suspiciously to dice rolls....