3x3 scrambler experiment

[Deleted member 36445]

after some folks confirmed that you can scramble a 3x3 like a megaminx (https://www.speedsolving.com/thread...ory-question-thread.56699/page-8#post-1442214) I tried to make my own scrambler with react

check it out: https://www.filipeteixeira.com.br/scrambler/

Spoiler: how it was made

 

[PapaSmurf]

That's pretty cool. I doubt it will be used in any large capacity at all (current scrambling is already pretty good), but it's really fun to scramble with that. I wish it was better than what we have atm.

 

[Deleted member 36445]

maybe some maths or trial and error can define a better length of the sentences for an optimal scrambling

 

[Deleted member 51471]

This is really cool! Not very useful, but still epic!

 

[xyzzy]

On a completely objective level, this is pretty useless – it doesn't provide the same guarantees of scramble quality that the existing methods do, and it doesn't seem especially ergonomic either.

But still, it seems cool, and it's not like everything we make have to be maximally useful for everything anyway.

 

[rokicki]

Forget scrambling. Do a regular scramble, then solve a 1x1x3 row on the UL edge (ignoring centers), then try to solve the rest of the way using just r and d. It's pretty challenging . . .

 

[Cubing Forever]

Forget scrambling. Do a regular scramble, then solve a 1x1x3 row on the UL edge (ignoring centers), then try to solve the rest of the way using just r and d. It's pretty challenging . . .

Corner permutation should be solved for that to be possible right?

 

[Deleted member 36445]

Corner permutation should be solved for that to be possible right?

I understood that no
if you're able to scramble CP/EP/EO/CO with <rdU> then the opposite is true

 

[Deleted member 36445]

Forget scrambling. Do a regular scramble, then solve a 1x1x3 row on the UL edge (ignoring centers), then try to solve the rest of the way using just r and d. It's pretty challenging . . .

a new method is born
(but I think you should be allowed to throw some U moves here and there...)

 

[Cubing Forever]

I understood that no
if you're able to scramble CP/EP/EO/CO with <rdU> then the opposite is true

Yes but Tomas was talking about <r, d>not<r, d, U>
It's definitely possible to scramble and solve EO, CO, EP and CP using <r, d, U> but not <r2, d2, U> like megaminx

 

[Deleted member 36445]

Yes but Tomas was talking about <r, d>not<r, d, U>
It's definitely possible to scramble and solve EO, CO, EP and CP using <r, d, U> but not <r2, d2, U> like megaminx

my scrambler is not just double turns like megaminx

 

[rokicki]

Oops, you're right; you can't solve everything with just <r,d>; the centers and corners are mutually restricted.

Okay, scramble with just <r,d> and then solve with <r,d> (no U). I think it's pretty challenging!

 

[Christopher Mowla]

Oops, you're right; you can't solve everything with just <r,d>; the centers and corners are mutually restricted.

Okay, scramble with just <r,d> and then solve with <r,d> (no U). I think it's pretty challenging!

It looks like it's not possible to do a single 3-cycle of corners. (I should have known that!)

But here are sequences I found for all other basic cases (in less than an hour):

2 corner twist:
[d' (d2 r' d2 r d2 r d2 r')3 d, r]

4 corner twist:
[d' (d2 r' d2 r d2 r d2 r')3 d, r2]

Corner 2 2-cycle
(r' d2 r' d2 r d2 r d2)3

2 Edge flip:
[(r d r' d' r' d)3 , (r d2 r d2)3]

Edge 3-cycle
[[(r d2 r d2)3, d2], r]

 

[rokicki]

Great job! I have an edge 3-cycle in 16 moves, an edge two-flip in 30 moves, a corner two-twist in 28, and a corner double-2-cycle in 24 moves. None of these are necessarily optimal; all are commutators or repeated algorithms. I cheated; I used a computer to generate the algs. Still, quite challenging.

 

[Christopher Mowla]

Great job! I have an edge 3-cycle in 16 moves, an edge two-flip in 30 moves, a corner two-twist in 28, and a corner double-2-cycle in 24 moves. None of these are necessarily optimal; all are commutators or repeated algorithms. I cheated; I used a computer to generate the algs. Still, quite challenging.

I new you were cooking a recipe! My goal was to post first!

I guess this is the analogue to 2-Gen type 5 for the 4x4x4. (Hence my interest!)

P.S.

You couldn't find a shorter solution than 24 moves for the corner 2 2-cycle? (That's what I got too.)

 

[rokicki]

Indeed, there are no shorter corner pairs of 2-cycles. Indeed, the coset of corner perms that preserve edges, centers, and corner orientations (for one particular corner orientation convention) looks like this:

   1 0
   4 23
  18 24
  14 25
  14 27
   9 28

so quite deep.

Fixing edge orientations *starts* at 28 moves (and there is indeed a 28-move 2-edge flip):

   1 0
  76 28
  62 29
254 30
198 31
195 32
185 33
  48 34
   5 35

Fixing corner orientations is easier:

   1 0
  84 24
  16 25
  30 26
  20 28
  86 30
   6 31

The deepest position I've found overall (so far) is 36 moves but I suspect there's at least a distance-37 position.

 

[Christopher Mowla]

Fixing edge orientations *starts* at 28 moves (and there is indeed a 28-move 2-edge flip):

.
.
.

Fixing corner orientations is easier:

Wow, I thought it would be the other way around! Nice analysis! Where is @xyzzy ?

 

[Deleted member 36445]

I'm glad my silly experiment led into an interesting new method
you guys are awesome

 

[xyzzy]

Where is @xyzzy ?

Busy, but since I was summoned…

@Ben Whitmore generated a bunch of algs for this before. Also, I feel like I saw some activity on ⟨Rw,Uw⟩ solving methods not too long ago, although now that I'm looking for it again, I can't find it.

 

[PapaSmurf]

Busy, but since I was summoned…

@Ben Whitmore generated a bunch of algs for this before. Also, I feel like I saw some activity on ⟨Rw,Uw⟩ solving methods not too long ago, although now that I'm looking for it again, I can't find it.

That's because it was on the FTO discord server. Yeah, <rd> is just a transformed <ru>, so it's the same thing. We could reverse engineer this to make a potentially more ergonomic RS scrambler. Something like reduce to <RrUu> to <RU> could work, although I know next to nothing about this kinda thing.