Quote from: natri on June 22, 2014, 07:22:10 PM

Hey guys, I'm back!

I've gotta say I spent quite some time working things out, and this time I've manually obtained numbers from exactly 1000 rolls! Yes, 1k rolls, every single one recorded and written down in my excel file for later analysis. And I've gotta say the results don't surprise. Running chisquare test gave again a quite confident rejection of the null. Numbers probabilistically don't seem random.

Some info you might wonder (my test size of 1000 rolls):
MAX value: 9986
MIN value: 3
Median: 4815
Mean: 4832,296

9056 values never occurred, 885 singles, 53 duplicates and 3 (!!) triplicates.

Some numbers are even complete obvious duplicates of what I got in my past rolls. I am VERY strongly inclined to believe the probability of certain values is much MUCH higher than some others (here we esp. wonder about 9998,9999 and 10000 - who has ever even seen those?!).

Quote from: wetsuit on June 21, 2014, 05:50:38 AM

The javascript based roll verifier is now online at http://roll-verifier.s3-website-us-east-1.amazonaws.com/. I shall link it via the website shortly.

You indeed did not change the roll outcome and you used the same server and client seeds (all correct), however, even with completely random inputs (client seed, server seed) the final outcome is most likely not a random number ranging from 0-10000, I am quite confident in that.

Any updates? Have you performed any more tests to ascertain the randomness? Based on my own sample after playing for about 6 months, they do not look entirely random (I still need some time to fine tune my analysis though).