Re: [Boost-users] Boost random not looking so random

Subject: Re: [Boost-users] Boost random not looking so random
From: Lars Viklund (zao_at_[hidden])
Date: 2012-10-28 02:06:02

On Sat, Oct 27, 2012 at 09:30:45PM -0700, Robin Rowe wrote:
> I'm using boost to shuffle an array, but it's not looking as random as I  
> would expect. Am I not using it right?

Randomness doesn't give a hoot about discernable patterns. A generator
that avoids "non-random patterns" would be properly defective.

If you want to measure the randomness of something, consider using
the proper statistical tests designed to measure the uncorrellation of
sources.

> The code below outputs sequences that look random mostly, but I see
> patterns such as 7 7 12 12 11.
>
> 	boost::random::mt19937 rng;
> 	boost::minstd_rand lcggen;
> 	lcggen.seed(clockTime);
> 	rng.seed(lcggen);
> 	unsigned min=1;
> 	unsigned max=15;
> 	for(unsigned i=0;i<5;i++)
> 	{	boost::random::uniform_int_distribution<> dist(min,max);
> 		const int j = dist(rng);
> 		... code that shuffles stuff using j ...
> 		min++;
> 		cout<<j<<' ';
> 	}

You may of course be using the results of an uniform distribution in a
way that causes the resulting shuffle have some configurations that are
more likely than others, but ocular observation is still not a reliable
metric for that.

-- 
Lars Viklund | zao_at_[hidden]