Now I've looked into this I was definitely wrong. Probably the most informative discussion is the subsection on Psuedorandom generators in the wikipedia entry for the Fisher-Yates Shuffle.

This would apply if you used only one random number, but why would you do that? For example, take a random permutation of a pack of cards, with 52! possibilities. It's easy to generate a sequence such that every ordering has an equal probability, despite the huge size of 52!, because multiple...

If there are any R users around I've just put out some software as an R package. The description and CRAN link are below. The average solve time for the Kociemba solver is 27 milliseconds on my laptop because I only use tiny pruning tables (due to CRAN limits on package size).
cubing: Rubik's...

One more. This only requires the number of cycles as an input, and not the length of each cycle.
Fortunately this still fixes the permutation sign. The number of targets is given in the preamble.
The hard part is to generate the cycle lengths to ensure each conditional state is equally...

As below. Written in the language R which you can regard as pseudo-code, but it is runnable; download R software (search for CRAN) and just copy and paste. Uses cycles and not targets because the number of targets depends on where your buffer is. Probably contains lots of bugs.
# Function for...