Lots of rand()s

Some of the coolest programs are the ones with random results... russian roullette, initial placement in Life, etc.

I wanted to see how close to even rand() would get over a large sample size, so I ran 100 million tests, and published the results.

If this gets moved to Code Catacombs, that's fine by me. It's not exactly cool, but it is cool by extension. :)

Also, if someone would like to comment on how it could be made more efficient, that would be awesome! I'm specifically interested in the while block, because it's the workhorse, though other comments are good too.

Code:

#!/usr/bin/perl -w
use strict;

my ($counter, $tests) = (0, 100_000_000);
my %results;

my $rand;                #declare once, save time
while ($counter++ < $tests) {        #not creating with .. operator, m
+em efficient
    $rand = int(rand(10) + 1);
    $results{$rand}++;
}

my $total;
$total += $_ for values %results;    #only 10 results possible

print "\n\n";
print "Total\t$total\n\n";
print "$_:\t$results{$_}\t\t" . $results{$_} / $total  * 100 . "\n" 
                        for sort {$a <=> $b} keys %results;
print "\n\n";
[download]

Output:

Total   100000000
 
1:      10003440                10.00344
2:      9996091         9.996091
3:      10004370                10.00437
4:      9996926         9.996926
5:      10000607                10.000607
6:      9998471         9.998471
7:      10002412                10.002412
8:      9997738         9.997738
9:      9999782         9.999782
10:     10000163                10.000163

Comment on Lots of rand()s Download Code

Replies are listed 'Best First'.
Re: Lots of rand()s by Albannach (Monsignor) on Mar 08, 2004 at 14:22 UTC
Your comment `#not creating with .. operator, mem efficient` is somewhat outdated, as for several major releases now (at least since 5.005_03) the `..` operator has not created a masive temporary array as it was wont to do in earlier times. So unless you are using an old version of Perl, please feel free to use `for(1..1_000_000)` without fear, and avoid clumsy alternatives! Update: Back to your topic, this little exercise isn't a new idea, but it is a good thing to try in order to get some understanding of rand. You might try different pseudo-random number generation functions (a quick google should turn up lots) or indeed some CPAN options) and compare the results, especially if you are going to rely on rand for any meaningful purpose (e.g. we use Monte-Carlo economic simulations on occasion). You might also consider the distribution being generated - the built-in functions typically generate uniform distributions, but for many purposes that isn't immediately suitable. In the real world however, if you do need lots of pseudo-random numbers you're most likely better off with a non-default algorithm coded in C for speed, or some real randomness. -- I'd like to be able to assign to an luser	[reply] [d/l] [select]
Re: Re: Lots of rand()s by David Caughell (Monk) on Mar 09, 2004 at 21:04 UTC
Re: the .. operator. Thanks for pointing that out (this is the kind of thing I wanted to know). I'm not a stats major, and this isn't a life-or-death thing, (there are other nodes dealing with true randomness on PM), I just wanted to know whether the results would come up distributed relatively evenly. Thanks for the reply, Dave	[reply]
Re: Lots of rand()s by Jaap (Curate) on Mar 08, 2004 at 12:46 UTC
Here's an optimization: `$results{int(rand(10) + 1)}++;` [download] There's no need to write it to `$rand` first. This should be a little bit faster. What are you trying to prove with this? That rand() is not really random?	[reply] [d/l] [select]
Re: Re: Lots of rand()s by David Caughell (Monk) on Mar 09, 2004 at 21:00 UTC
Hey Jaap, I'm not trying to prove anything. I just wanted to see what the results would be, and since I figured other people might want to also, I posted the results. Thanks for the optimization... makes a lot of sense. :)	[reply]


Perl-Sensitive Sunglasses
	PerlMonks