41024Re: Statistical analysis of (mostly megaminx) scramblers
- May 2, 2008So you do sticker-cycles for Megaminx BLD as well?
As for scramble-randomness calculation wishlists... they always grow
faster than any of us want to code them up. Whenever I read
something like that - say about 5x5 centers, I get chills down my
spine and nausea. I wonder if it's just me or do a lot of other
programmers feel that way. It's like I did it enough while in
college and at work, that coding drives me crazy even if it's for
something "fun" like cubing.
Although, I a couple months ago I was bored enough to write a solver
program for the Pentultimate. It's a 12-sided puzzle (dodecahedron),
face-turning down the middle such that each pentagon face has 6
stickers. Or you can look it up if your curious. I guess I found
that to be an enjoyable excersize.
I'd say sticker distribution approach is *okay*. For sufficently
high trial-length/sample-size it'll still yield very useful results.
It's just that somehow I find that it's not entirely *ideal* that's
-Doug (why am i up this late... /early?)
--- In email@example.com, "Stefan Pochmann"
> Same suggestions as Doug, plus...
> - I think sticker color distribution is alright. Causes triple-
> counting of corners, but also triple-right counting. I believethis
> evaluation is fine.3x3
> - Might be possible to do *exact* computations of averages and
> standard deviations if you use dynamic programming similar to my
> scramble analyzer here: http://stefan-pochmann.info/spocc/other_stuff/
> - When blindsolving the megaminx (or the 3x3), pieces not at the
> correct place simply have no "correct/wrong" orientation for me.
> those at the correct place are obvious. However, you can simplyfor
> each piece pick any reference orientation you like. Gosh, I wish Imany
> had written that article already (it's on my to do list).
> - I had actually thought about analyzing the 5x5 centers, that's
> something else I'd like to see. So after N scramble moves, how
> moves does it take to solve 1) the white center 2) the easiestpieces
> center. This is something you could do here as well, i.e., analyze
> the solution length or the position distribution of a group of
> or stickers. But as mentioned above, I believe single stickercolor
> distribution to be fine.
- << Previous post in topic Next post in topic >>