SSC (Shadowslice Snow Columns) 3x3x3 Method

[TyeDye]

http://youtu.be/KeLoHqZInWs
First cubing video, they'll get better (I hope).

Thank you so much! It all makes sense now. This method is very interesting and fun. It combines many different aspects of other methods into one. Sort of a semi-belt method, with elements of winter variation, roux, edge orientation, and even square one. This is an awesome method and I applaud you shadowslice

 

[sqAree]

I propose a slight variation of SSC. Let's have a look at this example solve:

F' L D' B' D2 F //EO + line
R2 //separate RB/RG
U' R2 // RB pair
U2 D R // RG pair
D' R // CO = 2001 on U/D
U' D2 R // CO + belt
D' R2 D R2 // CTL
U' R2 D' R2 D R2 D' R2 D' R2 D2 R2 //CPBL
D2 //DLDR
M2 U2 M2 U //ULUR
D2 M' D2 M //M perm

39 ATM

Steps 3 and 4 are the same as in the original method although I'm pretty sure we can make use of a wider variety of algs for CPBL to influence LEE during that step.
Instead of EO + 3 edges followed by orienting 3 corners on D, setup and do WV/SLS I propose doing only the EOline (on left) as the first step (which is easier to see in inspection and has less moves) and then solve the remaining 2 edges of the belt and CO of both layers in one intuitive step. I believe the latter can always be done in 10 moves ATM but will take significantly less moves in most of the cases. With the first step being a bit simpler it's maybe possible to predict the location of the unsolved belt edges as well as CO on D so recognition would be fine (only required to look at U).
I'm not yet sure if that step can be performed fast using the intuitive way. I basically observed that by using SLS we always break up the 3rd belt edge anyway and that the algs finish in the same manner all the time so we can save a lot of moves by not putting it in the first place.
As for how to perform that intuitive step, we would first of all separate the two edges into the U and D layers and then manipulate the CO (in relation to those edges) in both layers by setup <U,D> followed by <R>. Directly before solving belt and CO there is always a pretty easily recognizable pattern, that is, by starting from the edge and going clockwise: ccw, solved, solved, cw (hence CO = 2001 in the comments of the example solve by assigning numbers modulo 3 to the respective corner orientation).
So this comes down to the quest of getting this pattern in 6 moves ATM. For now I'm not able to figure it out fast but I suspect we can find some tricks, orders and standard cases similar to how LSE is solved in Roux nowadays.

My example solve has almost the worst case for the first two steps and yet only 17 moves as opposed to the projected movecount of 23 for the original method.

 

[bobthegiraffemonkey]

I propose a slight variation of SSC. Let's have a look at this example solve:

F' L D' B' D2 F //EO + line
R2 //separate RB/RG
U' R2 // RB pair
U2 D R // RG pair
D' R // CO = 2001 on U/D
U' D2 R // CO + belt
D' R2 D R2 // CTL

l U' R D2 R' U l' //CPBL
U' R2 D R2 D' R2 D R2 D' R2 //different CPBL

Dunno after that. I initially wrote this method off, but some variation of it might have potential with some more tricks like different CPBL algs which affect E-slice. I think permuting some of E layer at the end (which means as long as belt edges are in the belt and oriented you don't need to permute them at first) along with PBL tricks could be worth looking into, would need some work though.

I've been working on some L6C algs for sq-1 recently (DFL/DBL solved -> solve rest without separating first). Can cross-post if there's interest when I'm posting to a sq-1 thread.

 

[shadowslice e]

I propose a slight variation of SSC. Let's have a look at this example solve:

F' L D' B' D2 F //EO + line
R2 //separate RB/RG
U' R2 // RB pair
U2 D R // RG pair
D' R // CO = 2001 on U/D
U' D2 R // CO + belt
D' R2 D R2 // CTL
U' R2 D' R2 D R2 D' R2 D' R2 D2 R2 //CPBL
D2 //DLDR
M2 U2 M2 U //ULUR
D2 M' D2 M //M perm

39 ATM

Steps 3 and 4 are the same as in the original method although I'm pretty sure we can make use of a wider variety of algs for CPBL to influence LEE during that step.
Instead of EO + 3 edges followed by orienting 3 corners on D, setup and do WV/SLS I propose doing only the EOline (on left) as the first step (which is easier to see in inspection and has less moves) and then solve the remaining 2 edges of the belt and CO of both layers in one intuitive step. I believe the latter can always be done in 10 moves ATM but will take significantly less moves in most of the cases. With the first step being a bit simpler it's maybe possible to predict the location of the unsolved belt edges as well as CO on D so recognition would be fine (only required to look at U).
I'm not yet sure if that step can be performed fast using the intuitive way. I basically observed that by using SLS we always break up the 3rd belt edge anyway and that the algs finish in the same manner all the time so we can save a lot of moves by not putting it in the first place.
As for how to perform that intuitive step, we would first of all separate the two edges into the U and D layers and then manipulate the CO (in relation to those edges) in both layers by setup <U,D> followed by <R>. Directly before solving belt and CO there is always a pretty easily recognizable pattern, that is, by starting from the edge and going clockwise: ccw, solved, solved, cw (hence CO = 2001 in the comments of the example solve by assigning numbers modulo 3 to the respective corner orientation).
So this comes down to the quest of getting this pattern in 6 moves ATM. For now I'm not able to figure it out fast but I suspect we can find some tricks, orders and standard cases similar to how LSE is solved in Roux nowadays.

My example solve has almost the worst case for the first two steps and yet only 17 moves as opposed to the projected movecount of 23 for the original method.

This looks great! I applaud you! It could even push the method to sub-40 ATM consistently (which would be insane cause no blocks)- we just need to get it fast. It may also mean that the method only needs 4 algs though I think we may be better of working out better ways to do CPBL for various LEE- I have a few thoughts: getting opposite colour bars and stuff like that. I'll elaborate in the morning. Crafto's EZD isn't bad either so perhaps we could seperate edges while doing CP or something like that.

 

[wir3sandfir3s]

I propose a slight variation of SSC. Let's have a look at this example solve:

F' L D' B' D2 F //EO + line
R2 //separate RB/RG
U' R2 // RB pair
U2 D R // RG pair
D' R // CO = 2001 on U/D
U' D2 R // CO + belt
D' R2 D R2 // CTL
U' R2 D' R2 D R2 D' R2 D' R2 D2 R2 //CPBL
D2 //DLDR
M2 U2 M2 U //ULUR
D2 M' D2 M //M perm

39 ATM

Steps 3 and 4 are the same as in the original method although I'm pretty sure we can make use of a wider variety of algs for CPBL to influence LEE during that step.
Instead of EO + 3 edges followed by orienting 3 corners on D, setup and do WV/SLS I propose doing only the EOline (on left) as the first step (which is easier to see in inspection and has less moves) and then solve the remaining 2 edges of the belt and CO of both layers in one intuitive step. I believe the latter can always be done in 10 moves ATM but will take significantly less moves in most of the cases. With the first step being a bit simpler it's maybe possible to predict the location of the unsolved belt edges as well as CO on D so recognition would be fine (only required to look at U).
I'm not yet sure if that step can be performed fast using the intuitive way. I basically observed that by using SLS we always break up the 3rd belt edge anyway and that the algs finish in the same manner all the time so we can save a lot of moves by not putting it in the first place.
As for how to perform that intuitive step, we would first of all separate the two edges into the U and D layers and then manipulate the CO (in relation to those edges) in both layers by setup <U,D> followed by <R>. Directly before solving belt and CO there is always a pretty easily recognizable pattern, that is, by starting from the edge and going clockwise: ccw, solved, solved, cw (hence CO = 2001 in the comments of the example solve by assigning numbers modulo 3 to the respective corner orientation).
So this comes down to the quest of getting this pattern in 6 moves ATM. For now I'm not able to figure it out fast but I suspect we can find some tricks, orders and standard cases similar to how LSE is solved in Roux nowadays.

My example solve has almost the worst case for the first two steps and yet only 17 moves as opposed to the projected movecount of 23 for the original method.

Please make this a good method, I need a new main method badly... Looks good so far

 

[sqAree]

l U' R D2 R' U l' //CPBL
U' R2 D R2 D' R2 D R2 D' R2 //different CPBL

Dunno after that. I initially wrote this method off, but some variation of it might have potential with some more tricks like different CPBL algs which affect E-slice. I think permuting some of E layer at the end (which means as long as belt edges are in the belt and oriented you don't need to permute them at first) along with PBL tricks could be worth looking into, would need some work though.

Interesting idea. There is this alg R2 U2 R2 U2 R2 that repairs the E-slice in your second case and preserves ULUR. 5 additional moves so not sure it's worth in the end.

This looks great! I applaud you! It could even push the method to sub-40 ATM consistently (which would be insane cause no blocks)- we just need to get it fast. It may also mean that the method only needs 4 algs though I think we may be better of working out better ways to do CPBL for various LEE- I have a few thoughts: getting opposite colour bars and stuff like that. I'll elaborate in the morning. Crafto's EZD isn't bad either so perhaps we could seperate edges while doing CP or something like that.

Nice to hear that from the inventor of SSC himself!
I was actually working on it a few weeks ago but had to stop due to an internship and now that I lost focus a bit I decided to post what I had so far. I started analyzing the patterns that occur one step before the 2001 but didn't find anything fast to visually recognize, although there are not many cases. If someone is successful here it would be a major breakthrough as the solver would only need 2 or 4 moves ATM to reach that pattern which I believe is possible with just a bit of training.
Then there was the approach to build pairs as I did in my example solve (pairs like you use to setup for WV in the original variant) and extend them to triples (2 and 1 of CO) but I encountered some cases where that strategy didn't work out (or needed more than 8 moves).
My last attempt was to reduce the cube state in that step to a simplified mathematical system just consisting of 8 numbers (modulo 3) describing the CO state and analyze what certain "allowed" operations (R2, R, R') do (because <U,D> stuff doesn't affect it) but it was too complicated to do in a speedsolve I guess.

That's why I'm not yet entirely convinced of my idea. The best I could think of would be a tree diagram for certain classes of cases indicating how to reach the next step (as in Roux LSE EO step) which would be great but a huge task to create in the first place.

As for CPBL tricks let's examine those two solves of the same scramble (L2 F2 D' F2 U' F2 D2 R2 B2 L2 U R F2 D R U' B' R' B L' F):

y' U' R F B L B' //EO + line (separation of GO/GR skips)
U D' R2 U R U' D2 R D R' //CO + belt
D2 R2 U' D R2 //CTL
U D R2 U' R2 U' D' R2 U' R2 //CPBL + DLDR edges opposite
U' D' M2 D' //DLDR
U M U2 M' U M2 U //ULUR
U2 M2 U2 M2 //M perm

y' U' R F B L B' //EO + line (separation of GO/GR skips)
U D' R2 U R U' D2 R D R' //CO + belt
D2 R2 U' D R2 //CTL
U D R2 U' R2 U D R2 D' R2 //CPBL
M2 U' M' U2 M' D M2 D' //DLDR
M2 U M' U2 M' U' M2 U' //ULUR
M U2 M U2 M2 //M perm

The first solve has 38 moves ATM using an alternative CPBL alg that sets up DL/DR edges as opposite.
The second solve uses my standard CPBL alg and has 46 moves ATM (I suspect it's mostly due to me not being good at LSE/LEE generally but I think the difference is noticeable especially as both CPBL algs have the same length). I don't think a lot of alternatives for each alg are required for this simple improvement. Even if there were 5 for each it would still only be 40 algs but I suspect 2 per case are sufficient.

Maybe I will generate those algs tomorrow and try to find something for recognition. Also looking forward to what you will find.

Oh, and I forgot to mention that the two missing belt edges are totally interchangeable for my step 2 (as the last move is never R2 we can decide whether to finish with R or R' depending on which edge is where) which is nice because the solver doesn't have to keep track of that information in inspection / until the very end of CO+belt.

 

[shadowslice e]

Just to summarise the current form of SSC-O which seems to be quite nice:

1) EoEdge (7/7)
2) orient corners and place L2E (8/15)
3) CTL (5/20)
4) CPETL (10/30)
5) EPBL (7/37)

Total alg count: 60 or 72 though many could be considered intuitive. Especially CPETL and about half will be inverses which remain pretty ergonomic so maybe around 40 need to be learned.
(48 for CPETL I believe though many will be intuitive and inverse/mirrors.)
Potentially more for CTL though if you want to learn that. You could also learn CETL and have about the same number of algs in total I believe.
12 or 24 for EPBL depending on whether you want to do M2 U2 M2 or not though this could quite easily cancel as well.

Comments: lookahead could be very good after EoLine as all the steps can be predicted during the last few steps of the previous method. Ergonomically, it is quite nice as the moves are mostly <R,U,D> with the occasional <R,U,F> alg and an <M,U,D> last step. In addition, moves may be more easily cancelled in a lot of the steps (especially 2->3 and 3->4 due to the movesets being similar and often both starting with R moves).


To do list:
Create a way of doing step 2c nice and easily
Gen full set of algs for CPETL
Sub-15

 

[sqAree]

Just to summarise the current form of SSC-O which seems to be quite nice:

1) EoEdge (7/7)
2) orient corners and place L2E (8/15)
3) CTL (5/20)
4) CPETL (10/30)
5) EPBL (7/37)

Total alg count:
(48 for CPETL I believe though many will be intuitive and inverse/mirrors.)
Potentially more for CTL though if you want to learn that. You could also learn CETL and have about the same number of algs in total I believe.
12 or 24 for EPBL depending on whether you want to do M2 U2 M2 or not though this could quite easily cancel as well.

Comments: lookahead could be very good after EoLine as all the steps can be predicted during the last few steps of the previous method. Ergonomically, it is quite nice as the moves are mostly <R,U,D> with the occasional <R,U,F> alg and an <M,U,D> last step. In addition, moves may be more easily cancelled in a lot of the steps (especially 2->3 and 3->4 due to the movesets being similar and often both starting with R moves).


To do list:
Create a way of doing step 2c nice and easily
Gen full set of algs for CPETL
Sub-15

So CPETL solves CP on both layers and brings the edges into their respective layer at the same time?

 

[shadowslice e]

So CPETL solves CP on both layers and brings the edges into their respective layer at the same time?

Yep: basically turns the <R2, L2, F2, B2, U, D, M2, S2, y> group into <M2, U, D>

 

[bren077s]

So quick question. If I want to learn this method, would you suggest that I start by learning the variation that is in the first post, or one of the later variations that you talked about in the thread? If a later variant, which one?

 

[shadowslice e]

For now, I would recommend doing it like this:
1) EoEdge+1
2) Orient 3 D corners
3) create and insert final pair
4) OCLL
5) CTL
6) CPBL
7) LEE however you want

Then learn WV or SLS
Try to orient corners while doing EO

By the time you've mastered that, we may have finished the modified step 2 and genned the CPETL+ EPBL algs

 

[bren077s]

What does CTL and CPBL stand for? Also, do you have a link to those algorithms?

 

[sqAree]

What does CTL and CPBL stand for? Also, do you have a link to those algorithms?

CTL = corners to layers
You basically put the according corners to the top resp. bot face of the cube. You don't really need algs for that, at least I strongly recommend solving this intuitively using U, D and R2 moves.

CPBL = corner permutation of both layers
Now that the corners are in their respective layer you permute the corners top and bot in one step.
I use these algs:
adj top right: R U R' F' R U R' U' R' F R2 U' R' U'
adj bottom right: R' U R' F' R U R' U' R' F R2 U' R' U' R2
adj top front&bot front: R2 U' R2 (U D) R2 D' R2
diag top: F R U' R' U' R U R' F' R U R' U' R' F R F'
diag bottom: [R2 (U D) R2 D]*3
double diag: R2 (U D') R2 (U D) R2 U2 R2
top diag, bot right adj: (R2 U' R2 U) (R2 U' R2 U') R2 U2 R2
top right adj, bot diag: (R2 D' R2 D) (R2 D' R2 D') R2 D2 R2
adj = adjacent, diag = diagonal, just refers to the unsolved corners.

 

[TyeDye]

2) Orient corners:
. 2a) Orient 3 corners on the d-face (any 3: it doesn't matter)
. 5 moves, 0 algs
. 2b*) Pair the remaining e slice edge on the U layer- basically, set up a pair that would occur in WV but the colours do not have to match as they would in WV (this is probably the hardest thing to get your head around).
. 3 moves, 0 algs
. 2c*) Use the Winter Variation Last Slot algs to insert the remaining pair.
. 8 moves, 27 algs

List of Algs (so far checked and generated)

Spoiler

For images of the CO, see cubewhiz's winter variation page http://absolutemind.pagesperso-orange.fr/f2ll-angl.htm here. However, there are 5 additional cases that are not listed: All corners oriented (0), right facing one corner oriented (1a), almost anti-sune (2a), passing bikes mirrored (4a), anti-sune (8a), and anti-spiral (20a). Each are numbered to the WV case they are a mirror of.

Spoiler

I can't find the images of the Corner Orientation for the SLS algorithms. Can you find another site that has the images on them?

 

[sqAree]

Spoiler

I can't find the images of the Corner Orientation for the SLS algorithms. Can you find another site that has the images on them?

Did you use the correct link?
http://absolutemind.pagesperso-orange.fr/f2ll-angl.htm

This one works for me. But in any case some images are missing and they are not entirely ordered iirc, so here we go with full SLS:

http://imgur.com/fAT4gh1 (I hope the images are self-explanatory)

 

[bren077s]

CTL = corners to layers
You basically put the according corners to the top resp. bot face of the cube. You don't really need algs for that, at least I strongly recommend solving this intuitively using U, D and R2 moves.

CPBL = corner permutation of both layers
Now that the corners are in their respective layer you permute the corners top and bot in one step.
I use these algs:
adj top right: R U R' F' R U R' U' R' F R2 U' R' U'
adj bottom right: R' U R' F' R U R' U' R' F R2 U' R' U' R2
adj top front&bot front: R2 U' R2 (U D) R2 D' R2
diag top: F R U' R' U' R U R' F' R U R' U' R' F R F'
diag bottom: [R2 (U D) R2 D]*3
double diag: R2 (U D') R2 (U D) R2 U2 R2
top diag, bot right adj: (R2 U' R2 U) (R2 U' R2 U') R2 U2 R2
top right adj, bot diag: (R2 D' R2 D) (R2 D' R2 D') R2 D2 R2
adj = adjacent, diag = diagonal, just refers to the unsolved corners.

Thanks for all of the information. This method looks pretty cool and pretty unique.

 

[bren077s]

So, I know this method is new and still needs developing, but I was wondering what the best times you guys have gotten so far with this method? It would be nice if you also included what your pb is and what method you got your pb with.

 

[shadowslice e]

So, I know this method is new and still needs developing, but I was wondering what the best times you guys have gotten so far with this method? It would be nice if you also included what your pb is and what method you got your pb with.

My pb is 11.xx with an average of around 18 with not much practise tbh using the vanilla SSC-O (EoEdge+1+3 corners, SLS/WV, CPBL, LEE) I also don't know full WV/SLS so sometimes I just insert and so OCLL.

I still think that the SSC variant I did above is better though.

 

[shadowslice e]

CO step:
1) separate edges (or at least make sure they aren't adjacent)
2) form a 1x1x3 "column" for the D edge and put in DL
3) 1x1x2 for U
4) OL5C (orient last 5 corners: can learn algs if you want- I think there are 16 cases but you only need about 10 algs. They're about 4 or 5 moves on average I think, basically the same as 2-gen guimond algs)
5) finish CO by stuff

 

[sqAree]

CO step:
1) separate edges (or at least make sure they aren't adjacent)
2) form a 1x1x3 "column" for the D edge and put in DL
3) 1x1x2 for U
4) OL5C (orient last 5 corners: can learn algs if you want- I think there are 16 cases but you only need about 10 algs. They're about 4 or 5 moves on average I think, basically the same as 2-gen guimond algs)
5) finish CO by stuff

Can you give 1-2 example solves for this approach? I think it's close to what I had on my mind but not the most efficient that is possible.