Gripper: Rubik's Cube Algorithm Analyzer

[Stefan]

it managed to use double layer turns in beautiful ways that eliminated all kinds of ugly turns in ways that a human would never be able to conceive of, at least from what i've seen.

Can you show some?

 

[badmephisto]

ok well i played around with it for a while, but i cant get any real analysis done in the few minutes that i tried. I had to cut off the recursion at depths like 3 or 4 as opposed to 14 that would fully optimize an algorithm. I am also in process of doing a LOT of tweaks to the costs. Obviously there is still work to be done there.
I also found a few bugs that changed up the algorithms, and I suspect there are a few more still lurking around.

Here are a few interesting things i found so far: (these are only the raw algs, not the solution paths. you have to figure out what grip to hold the R slice in beginning at. Most of the time its obvious though)
funny t perm: R2 U' R2 D f2 R2 D R2 U' y R2 D
(start with thumb on D fingers on U)

interesting v perm: R2 U' B2 d r2 F U' y' R U R' D R D' R
old v perm with new twist: R' U z' U' r' U' b' R2 U' R' U R' F R F (b' is done with middle finger, keeping steady grip)

the n perm it seems will never become faster. Here are few that can potentially be considered good? I think more depth is required for N perm.. and even then i doubt that we will find something good. Thank god it only occurs 1/108 times...
R' d b' R2 x' U R' D U' R D' R2 U R' D
R' z R U' R2 D R' U D' R U' R2 z' R U' r

the ugly left-handed R perm variations that are promising maybe?:
R U' B r' U' R' z R U R' F' D F2 u'
R U2 R' z R y R U' R' D R U R' F u' <-- i like better

here are two interesting e perm variations:
z' R U' R2 F' R D R' F R2 U R' F' u' R
u' R' F R u z' U' R2 F' R D' R' F R2 U


anyway, I will do more later. I am really burnt out and tired, I may even take a break from this project for a day or two. If you want to get a crack at it, you have to go through the code on SVN, only that is up to date.
you want to dump algs from opticube into a file, like a.txt one per line, then put it into same directory, and then go

python rotgen.py a.txt 4

4 is the depth (maximum # of rotations it will insert. Ideally you want much more i think). 4 is pretty much most I could do on this PC.

that will print like 50 best algs, and in addition it will create a nice table called a.txt_out.html in the same directory that you can open and see the algs.

Also note M/S/E is still not supported for this. Oh and there may still be bugs

 

[fanwuq]

I remember there was a web-based version, but I can't find it. Also, the download links aren't working.
I was going to use Gripper to see if there's a better version of:
http://www.speedsolving.com/forum/showpost.php?p=160373&postcount=29

Edit:
Found it http://www.badmephisto.mine.nu/cgitests/outtime.py/handlehi
I had to search quite a few times. Is there going to be a downloadable version?

 

[suhas2112]

Hey meph, the links aren't working!! When i try to download the exes, it says 404 not found and when i try to go to your wiki link, it redirects me to the speedsolving wiki page... Can you please fix this soon, because i really wanna check this program out...

 

[badmephisto]

aaah this is an OOOOLD OOOLD thread.
yes i think i took everything down since
now that the school is over, maybe one day ill get this working again.
It was all web based, no download required.

 

[Scigatt]

aaah this is an OOOOLD OOOLD thread.
yes i think i took everything down since
now that the school is over, maybe one day ill get this working again.
It was all web based, no download required.

Next time include 'L' moves, for lefty cubers. I don't see them on the page fanwuq linked.