A Random Question

[Paul Dale]

Both of these will be biassed. Which one is closer to uniform will depend on <num>.

Off the top of my head, something like this would be better:

Code:

<< 1E11 -> n l 
    <<
        l n IQUOT n *         # Largest value below which the roll is fair
        DO RAND l * IP        # Roll once
        UNTIL DUP2 < END      # Check within range for fair roll
        NIP n MOD 1 +         # Clean up the stack and produce the result
    >>
>>

I've not checked this code, so there are likely errors.

I've fixed the point David noted below (IQUOT instead of IDIV).


Pauli