How do you do move analyses?

[not_kevin]

I'm trying to do a science project comparing the move efficiencies of Petrus and Roux. I know that there are numerous studies on cross move efficiencies; how, exactly were they done? I spent some time thinking about them, but they kinda didn't make sense to me (let's see, brute force? "Case 134,254..." No). If a computer code was used, how exactly was it used? I don't need to see the source code or anything; just the idea on how it was approached. Thanks in advance!

By the way, this isn't asking for the actual move analyses, but the process on how to do it.

 

[Lotsofsloths]

what do you mean?
For my science fair last year I just did 5 solves and counted the moves with different methods xD

 

[edwardtimliu]

muy confusado? (thats confused or ed-spanish)

 

[not_kevin]

Pardon for the vague description, here's what I mean:

Lars Vandenbergh did a cross study for move count (http://www.cubezone.be/crossstudy.html). I was going to use this idea to try to find the average move count for the Petrus method as described on Petrus' website and the Roux method as described on Roux's website. I was wondering how to do that, exactly.

@Lotsofsloths: that's something to verify my claim, but not quite what I was shooting for. Instead of an experimental approach, I would like a concrete, empirical way of finding the average move count for Petrus and Roux (looking for move efficiency, not speed efficiency, however).

 

[Lucas Garron]

You just solve all cases, by symmetry...
(So yes, like "(let's see, brute force? "Case 134,254..." No)")

Stefan has already gotten some good numbers on it: see page 32.
Cross has been fully done (Vandenbergh), and though I don't remember any full analyses of the 2x2x2 block, it's probably done. (Wasn't somebody going to modify Lars' code?)

 

[cuBerBruce]

Cross has been fully done (Vandenbergh), and though I don't remember any full analyses of the 2x2x2 block, it's probably done. (Wasn't somebody going to modify Lars' code?)

Solving a specific 2x2x2 block or a 2x2x3 block (in one step) has been done ( http://cubezzz.homelinux.org/drupal/?q=node/view/116 ), but not for "color neutral" solving as that requires analyzing over 900 quadrillion cases (or all 43+ quintillion scrambles if you don't apply symmetry reduction). By the way, I was not the first to do the 2x2x2 block analysis.

As far as learning how to use a computer to analyze such things, you might first read this page: http://www.geocities.com/jaapsch/puzzles/compcube.htm. Also, the documentation for Cube Explorer has some good information. There is also Stefan's thesis that Lucas mentioned.

Some care needs to be taken when trying to analyze an entire method to make sure you get accurate move counts for comparison purposes. If a step involves an "AUF" or setup and "undo setup" moves, you can't forget to include those moves (weighted by the probability of occurrence if computing average move count). Don't forget to account for skipped step cases as well.