147 and 997 Random Generator Algorithm

[EdS2]

With 100 values sampled 1000 times, you shouldn't expect exactly 10 of each value: that would indicate a non-random process, because it's a very improbable result.

(Although you should expect the average to be somewhat close to 10.)

Someone with more knowledge than I have might be able to say what the expected distribution is, and what its characteristics are. You'd then be able to compare your results with the expected, and with even more knowledge, say something about "how random" your process is.

In the absence of analytical techniques, I think it would be fair to repeat the overall experiment, 10 or 100 times, and see how the means and standard deviations vary. You would of course use the same algorithm but a different seed.

It's certainly interesting that some choices of parameter, for the very same algorithm, can give results which are better or worse than others.