LoL, After re-reading that the correct answer is 50. So you were closer that I was.

The thing about pseudo-random is that it is NOT random and hence the 'pseudo'. But if you are using a human (the delay in pressing the button) as the input entropy and that input entropy is greater (a significant multiple of) the selection permutations (number of entries) then all is good.

Keep in mind that because of the looping feature of your code that the input entropy must be divided by the loop frequency. For example; if you poll the button every 20mS then a 1000mS period has a entropy of 200 and not 1000.

Micro controllers cannot generate random numbers period. If they could then you would be able to write a formula for 'random' which because it is coded, would be predictable. The definition of random is 'unpredictable'.

I haven't looked at the library that you are using but I suspect that it may be software implementation of a linear feedback shift register. Check the entropy specifications as these types of pseudo-random generators have a very short sequence repeat period. If the sequence repeat period is NOT orders of magnitude bigger than the permutations then you are NOT being fair to some competitors as their particular permutation may be excluded.

You mean 500?

You could always throw in the ADC value on an "antenna" attached to a roughly floating pin...