On Apr 13, 2007, at 2:50, h_kociemba wrote:

>
>>>
>>> So, prompted by the discussion that ensued from Chris's extended
>> cross:
>>>
>
>>> Cube Explorer's random set:
>>> {4, 0, 298, 0, 2407, 0, 4543, 0, 2437, 0, 308, 0, 3}
>>> {402, 1559, 2846, 2585, 1878, 486, 242, 0, 2}
>>> {3579, 3708, 1859, 640, 174, 35, 4, 1, 0, 0, 0, 0, 0}
>>> {3699, 3710, 1750, 639, 163, 33, 5, 0, 1}
>>>
>>> JNetCube's random scrambles
>>> {1, 0, 292, 0, 2204, 0, 4412, 0, 2700, 0, 368, 0, 23}
>>> {83, 1508, 2922, 2530, 1900, 493, 257, 0, 7}
>>> {3668, 3697, 1871, 603, 128, 28, 4, 1, 0, 0, 0, 0, 0}
>>> {3690, 3690, 1840, 594, 142, 39, 5, 0, 0}
>>>
>
>
> I got a bit suspicious after I sent my last message - could JNet be
> really so bad in a probabilistc sense?
> Adding the numbers for CO for JNet I only get 9700 and not 10000. So
> Lucas made an error. I suspect, that instead of 83 it should be 383
> and then JNet gives *good* values for the corners.
> It still remains the problem with the 23 cases with correctly edge
> orientation, which is extremely unlikely.
> Nevertheless I think we can live with the fact that within 10000
> cubes you may get maybe 20 too many cubes with a completely correct
> edge orientation and JCube does its job well.
>
> Herbert

So can someone run these numbers for all mix lengths between, say,
20 - 30?

So we can finally answer the old question of how many moves it truly
takes to fully
scramble a cube?

There's lots of glory for whoever discovers the magic number. You may
even be responsible for invalidating all current cube records!


- - - - - - - - - - - -
"He who refuses to do arithmetic is doomed to talk nonsense."
--- John McCarthy

Lars Petrus - lars@... http://lar5.com