The New Method / Substep / Concept Idea Thread

[mDiPalma]

slow method:.

i used to experiment with a method I called something like EODB (you can probably find it if you look hard enough). basically you solve EO and the DB edge in the first step (easier than EOLine)

then u solve F2L with R U L F2 moves (more efficient because you can use F2 sometimes) (but it sucks because sometimes the pieces you need get stuck in DF) (but u can get around that by solving a U-layer pieces to the DF slot during the first step. #nbd)

then u solve COLL like usual. the movecount might be less because you have some random lil piece in DF.

then u solve EP5 with Athefre's alg set (which has since disappeared off the face of the interwebs, but if u search hard enough u can probably find it because I posted it like 10 times to this forum. i think he used photobucket or something)

tl;dr it's the same thing as urs with more blockbuilding and less omg-i-cant-see-whats-going-on and fewer moves.

 

[TDM]

Introducing a new

No, it definitely is not.

extremely fun

Nope.

yet very slow method

Yep.

---

then u solve EP5 with Athefre's alg set (which has since disappeared off the face of the interwebs, but if u search hard enough u can probably find it because I posted it like 10 times to this forum. i think he used photobucket or something)

http://i140.photobucket.com/albums/r25/Athefre/EP5Final.png

 

[mDiPalma]

http://i140.photobucket.com/albums/r25/Athefre/EP5Final.png

aw yiss

 

[IAmAPerson]

No, it definitely is not.


Nope.


Yep.

New-ish

Fun imo

I agree with you on this one

 

[2180161]

Further movecount reducing stuff (but not necessarily better for speed):

1) form 3 2x2x1 blocks with one colour face in common (like heise but without the centre-edge-centre 1x2x2 heise block) (well, I can always do it in <16 STM but someone who is really good at heise blocks can probably do it in <12 STM, maybe average ~10 STM)
2) move to form an F2L-2 (+orient remaining edges) (~5 STM)
3) TCLL (preserving orientation) (~9 HTM)
4) L5E (just permuting) (6.7 STM)

Total: <~30? (STM as well as about the same in SQTM, perhaps even lower in ATM or AQTM) (apparently about the same movecount as Snyder3 but is actually published and the algs are actually known)

Only issue is that the 2 slots (for the edge and the corner) are not always adjacent so it's probably the problem when it comes to speedsolving.

Also, this uses a new(?) LL/double LS variant: L5E, L5C (or TCLL)

Pros
Very low movecount
3-look method
1LLL (if we use 5 cycles for corners)
Still about the same number of algs as CFOP and far less than (for example) ZBLL
Few rotations (potentially none if done correctly)
Can be transfered directly to 2x2x2 where it averages <14 moves (not even CN) and is a 2-look method (although you could lookahead like in actual TCLL)

Cons:
Very heavy reliance on inspection
Orienting edges while doing the blocks could be difficult for a beginner
Still a fair few algs to learn (~84 for TCLL, do the L5E intuitively

Summary: I don't really know what to say here: this could be fast but it would require very good block building (but no worse than that of roux or petrus), the LS/LS style of LL can easily be combined with many other methods as well (notably ZZ where it could give a 1LLL)

Variations with easier F2L-2
CFOP: partial cross (~4 moves), 3 pairs- you could use a ZZ style approach (12 moves), orient all edges (6 moves), TCLL (9 moves), L5E (7 moves)- ~38 moves
ZZ: Eo (6 optimal, 7 realistic) FB+1 pair+1 edges (13 moves), TCLL (9 moves), L5E (7 moves)- ~35 moves
Petrus: 2x2x3 (10 moves), pair (3 moves), orient edges+place the last 2nd layer edge (7 moves), TCLL (9 moves), L5E (7 moves): 35 moves

Average: 35.666... Moves

For future reference I'll probably call this method MCELL (Missing Corner Edge Last Layer solution)

What does everyone else think?

Here is how I understand the method.
alg.cubing.net

I didn't do Heise blocks cuz I dont know how
INEFFICIENCY! YAY!

 

[shadowslice e]

Extreme Roux

Big cube roux style solution/ generalisation without reduction.
1) NxN-1x1 block on left
1b) Same on right
1) (Usually done as 2 centres, 6 edges, blockbuild)
2) CMLL
3) Orient and pair edges in the second from outside layer (Match up misoriented edges and orient as with LSE, pair, solve)
4) Using the M and E slices as well as the U layer, permute the centre pieces in the same layer as the edge
4a) corner centres
4b) edge centres
5) repeat with next layer in.

Meh, mostly a mental challenge to see how to generalise Roux for big cubes without resorting to reducing. Not much use until you get to the big Big Cubes (like 7x7x7 or 8x8x8+) where the centres will become somewhat easier than in standard reduction (as you only have to deal with less at a time) One thing that's worth pointing out though->Big Block Corners!!!!!<-

EDIT: Alternative (probably better) variant)

1) F2B (1x(Number of layer)xN blocks on both sides)
2) CMLL
3) Orient edges and pair the ones that will go in F, B, RU and LU with a matching corner centre from F, B R or L (Blue with Blue, Green with green etc in WCA orientation with standard western colour scheme)
4) insert pieces below the U corners so you have 1x1x2 blocks
5) LSE with Uw moves
6) repeat but for step 3) form 1x1x(number of layer) blocks including centres in it.

Well, slightly more practical but still only best on the bigger cubes. Just a bit of fun really. Try it and see how it goes for you.

 

[shadowslice e]

2x2x2 method: OCP
1) Orient all pieces (53 algs- you will pretty much be able to always find an orientation with no parity. If not then orient 4 pieces (~3 moves) then continue or correct the parity (~1 move) and continue)
2) Corners to layers. (Intuitive or 4 algs)
3) PBL (5 algs)

 

[StachuK1992]

You mean Ortega?

 

[Berkmann18]

2x2x2 method: OCP
1) Orient all pieces (53 algs- you will pretty much be able to always find an orientation with no parity. If not then orient 4 pieces (~3 moves) then continue or correct the parity (~1 move) and continue)
2) Corners to layers. (Intuitive or 4 algs)
3) PBL (5 algs)

This method already exist and is called Guimond.

 

[shadowslice e]

You mean Ortega?

No, with Ortega you solve a correct face. This method is a face of opposite colours.

This method already exist and is called Guimond.

Similar but not quite. Guimond is 2 step ortientation, this is one has more algs and combines steps 1 and 2. It's essentially a 3 look instead of 4 look. The variants if you can't find a OBL set is fairly similar though. You're probably better off learning something like SOAP or EG if you want to get good at 2x2x2 although the algs for this are the same as in SOAP.

EDIT: Or TCLL. I like TCLL

EDIT 2: I'm also not proposing this as a speedsolving method. The movecount/lookahead to time put in learning algs ratio is not very good (see above)

 

[Berkmann18]

No, with Ortega you solve a correct face. This method is a face of opposite colours.



Similar but not quite. Guimond is 2 step ortientation, this is one has more algs and combines steps 1 and 2. It's essentially a 3 look instead of 4 look. The variants if you can't find a OBL set is fairly similar though. You're probably better off learning something like SOAP or EG if you want to get good at 2x2x2 although the algs for this are the same as in SOAP.

EDIT: Or TCLL. I like TCLL

EDIT 2: I'm also not proposing this as a speedsolving method. The movecount/lookahead to time put in learning algs ratio is not very good (see above)

Don't confuse Orientation and Splitting.

 

[shadowslice e]

Don't confuse Orientation and Splitting.

well, if we're going
0) 3 oriented corners on D
1) Orient remaining corners
2) Separate pieces
3) PBL,
then the method above goes in the same way but merges steps 0 and 1 with 53 algs (I did it as N+1 for the step number) So, if we are going with the above, Guimond is still 2 step orientation whereas the above is 1 step but the bit after that is exactly the same.

 

[Berkmann18]

well, if we're going
0) 3 oriented corners on D
1) Orient remaining corners
2) Separate pieces
3) PBL,
then the method above goes in the same way but merges steps 0 and 1 with 53 algs (I did it as N+1 for the step number) So, if we are going with the above, Guimond is still 2 step orientation whereas the above is 1 step but the bit after that is exactly the same.

That's beginner's Guimond ! On full Guimond, you 1-look the Orientation step rather than setting up (or not) a case where you would have 0-2 corners to orient.

 

[shadowslice e]

That's beginner's Guimond ! On full Guimond, you 1-look the Orientation step rather than setting up (or not) a case where you would have 0-2 corners to orient.

With Guimond you always start with 3 oriented cubies from matching faces- a step which can be skipped fairly often for CN solvers (hence why it is step 0). You then orient all the remaining 5 cubies in one alg in a way similar to WV/SLS/CLS etc. This will often have algs which are for a secondary or ternary subset of full CO which is found in orientation of both layers that cannot be solved using CLL as it has a corner parity (I'm using the term very loosely) occuring 2/3 of the time. The method I described is always 1 look orientation and does not need to have the V to work. Thus, the aformentioned methid also is fairly difficult to do predicted separation with (which is what advanced Guimond users will do).

 

[Berkmann18]

With Guimond you always start with 3 oriented cubies from matching faces- a step which can be skipped fairly often for CN solvers (hence why it is step 0). You then orient all the remaining 5 cubies in one alg in a way similar to WV/SLS/CLS etc. This will often have algs which are for a secondary or ternary subset of full CO which is found in orientation of both layers that cannot be solved using CLL as it has a corner parity (I'm using the term very loosely) occuring 2/3 of the time. The method I described is always 1 look orientation and does not need to have the V to work. Thus, the aformentioned methid also is fairly difficult to do predicted separation with (which is what advanced Guimond users will do).

You are right, but it did already existed as Advanced Guimond/2-step Guimond or something like that.

 

[shadowslice e]

You are right, but it did already existed as Advanced Guimond/2-step Guimond or something like that.

Nope, 2-step Guimond still needs a V. It carries out orientation and separation simultaneously to reduce the step count. Essentially Guimond only really refers to mismatched V methods which the method I described is not as it has no V needing to be formed anywhere in the solve. If you really want a method it is somewhat similar to, look up RoFL, a method which Stefan did some number crunching for a few months ago which also eventually led to TCLL bring created

 

[Berkmann18]

Nope, 2-step Guimond still needs a V. It carries out orientation and separation simultaneously to reduce the step count. Essentially Guimond only really refers to mismatched V methods which the method I described is not as it has no V needing to be formed anywhere in the solve. If you really want a method it is somewhat similar to, look up RoFL, a method which Stefan did some number crunching for a few months ago which also eventually led to TCLL bring created

Anyway, have you generated the algs or not yet ?
Did you heard about Sortega ? That might help for some algs.

 

[shadowslice e]

Anyway, have you generated the algs or not yet ?
Did you heard about Sortega ? That might help for some algs.

https://docs.google.com/spreadsheets/d/1ZGTBDeU8vs-vc61sKA_6GPboZSodYuLVCFjjJmwlxmQ/pub?output=html these are the algs for the step in SOAP although shorter ones are definately possible as this one preserves the separation of the pieces. TBH, I don't really think it too necessary to generate algs for the method as I feel the amount of time spent learning the algs will nit be justified by the times achievable as it is a fixed 3 look method.

The only real application of the orientation algs is in the RC 3x3x3 method maybe to be categorised as RC-c

 

[Berkmann18]

https://docs.google.com/spreadsheets/d/1ZGTBDeU8vs-vc61sKA_6GPboZSodYuLVCFjjJmwlxmQ/pub?output=html these are the algs for the step in SOAP although shorter ones are definately possible as this one preserves the separation of the pieces. TBH, I don't really think it too necessary to generate algs for the method as I feel the amount of time spent learning the algs will nit be justified by the times achievable as it is a fixed 3 look method.

The only real application of the orientation algs is in the RC 3x3x3 method maybe to be categorised as RC-c

Okay.

 

[molarmanful]

2x2x2 method: OCP
1) Orient all pieces (53 algs- you will pretty much be able to always find an orientation with no parity. If not then orient 4 pieces (~3 moves) then continue or correct the parity (~1 move) and continue)
2) Corners to layers. (Intuitive or 4 algs)
3) PBL (5 algs)

Or you could use Guimond

What perceived benefits are there over Guimond (the 3-step one, not the beginner one), anyway? After all, Guimond does have far less algorithms, and it's still relatively easy to predict.