reto msdn 4: http://blogs.msdn.com/b/esmsdn/archive/2014/10/17/retosmsdn-reto-4-mezclando-con-eficiencia-en-c.aspx

Shuffler.cs

/*
We are asked to give a random derangement: http://en.wikipedia.org/wiki/Derangement

We use a slight modification of Fisher-Yates shuffle algorithm to make sure it's a derangement. 
This algoritm works because all elements are moved one or more times to the left so at the end
all elements are in different positions than the original one.

But with this modification, random properties are not great. For example some permutations are more
likely than others, and there are some permutatations which can never happen.

An algorithm with better random properties (but slower) is:
http://epubs.siam.org/doi/pdf/10.1137/1.9781611972986.7

Also we're using Random, if we want stronger randomness we should use 
crytographic random generation such as RNGCryptoServiceProvider.

*/

using System;
using System.Collections;
using System.Collections.Generic;

namespace Shuffle
{
     public static class Shuffler {
         
          public static List<T> Shuffle<T>(this List<T> list) {
               var rnd = new Random();
               var n = list.Count;
               var copy = list.ToArray();
             
               for (var i = n-1; i >= 1; i--) {
                    var j = rnd.Next(i);
                    var tmp = copy[i];
                    copy[i] = copy[j];
                    copy[j] = tmp;
               }
               return new List<T>(copy);
          }
         
          
          
          
          public static List<T> ShuffleOriginal<T>(this List<T> list) {
            var rnd = new Random();
            var copy = new List<T>(list);
            var n = copy.Count;

            for (var i = n-1; i >= 1; i--) {
                var j = rnd.Next(i);

                var tmp = copy[i];
                copy[i] = copy[j];
                copy[j] = tmp;
            }
            return copy;
        }
    }
}