# Columns First - underestimated method?

#### Sajwo

##### Member
Today I learnt Columns First method and I think it at least as fast as CFOP or roux.

This is short summary of this method:
- Solve 4 F2L pairs.
- Orient last layer corners, (CO, 7 algs)
- Permute LL corners, (CP, 2 algs)
- Solve centres together with 2-3 first layer edges.
- Orient the remaining edges (EO, 5 or 9 cases depending on method)
- Permute last edges, one way is to solve the last one(s) to FL and use EPLL, another is to use Roux style and put RU-LU and then permute the M-slice.

So; 4 F2L pairs can be done much faster than in fridrich method, because we can also use M-slice moves, so slots can be done very very quickly. We've got also very often after scrambling the cube at least one slot done. In my opinion 4 F2L pairs can be done in average under 18 moves.
Next step is CO and CP on last layer. Right there we can learn all CLL, so we can done this step using only 1 algorithm. Then we solve the centers with 2-3 FL edges, which is trivial and requires a few moves. The last step is EOLL and EPLL. We can done this of course in 1 algorithm, using ELL. Summarizing: we've got 18 moves for 4 slots, about 8-10 moves for CLL, 8 for centers and Dedges and ~11 for ELL, which gives us under 50 moves in average. This method is very friendly for CFOP and Roux users. So please express an opinion what do you think about my idea.

ps; Such a thought occurred to me when I did my first sub20 solve( and this was after spent 2 hours with this method). I can't do CLL and ELL, I am lame at dedges and I split 4th step into two stages, and I can still do sub20 solve. here you've go my video:

Last edited by a moderator:

#### Aero

##### Member
This method is very interesting, as a CFOP and Roux user I think this might be the perfect method.

#### Anthony

##### Professional Speedcuber
In my opinion 4 F2L pairs can be done in average under 18 moves.
Summarizing: we've got 18 moves for 4 slots, about 8-10 moves for CLL, 8 for centers and Dedges and ~11 for ELL, which gives us under 40 moves in average.
...

#### Sajwo

##### Member
fail, sorry, I'm just tired. But sub50 moves it's still very little

#### 5BLD

##### Member
And is it easy for you to sub-50 consistently? Please show some examples speedsolves.
Also, EO in ColF is awkward to do.

#### Sajwo

##### Member
And is it easy for you to sub-50 consistently? Please show some examples speedsolves.
Also, EO in ColF is awkward to do.
I can't do CLL and ELL, I am lame at dedges and I'm split 4th step into two stages, and I can still do sub20 solve.
please, read with understanding. I am pretty sure, you could get some sub10 solves with this method (if you know CLLs and ELLs), because of your experience with block building.

S MOVES ARE FAST

#### mDiPalma

##### Member
i used to use columns first

for the lse, place the DB edge while making the rest of the edges easy to orient
orient the edges with M' U' M or something else to force a better ep5
then use ep5: http://athefre.110mb.com/Step5.html

i just did a few speedsolves and reconstructed the best one

L' U2 R2 F' D2 R B' D2 L' D B2 L' R' B' R2 U B L D B2 F D' U2 B2 L'

xy (0 pair 1)
u R u' L U' (5 pair 2)
x U M' [i see that my other pair is really bad] R2 y R U2 R' (6 pair 3)
L' U2 L U L' U' L (7 pair 4)
R U2 R' U2 R' F R F' (cll 8)
R2 S R2 (3 centers and one edge)
M' U' M (3 one edge)
U2 M U M' U' M U' M' (8 one edge, orient, make easy ep5)
y M' U2 M (3 lucky ep5)
~43 stm

the only reason why i went back to zz was because there are too many steps in columns first. yes, i know many are only 3 or 4 moves, and you may skip a step very often, but it still seems to take ages to solve a cube when you have 9+ seemingly independent steps.

@5BLD, yeah it's impossible to consistently have <50 moves at any decent pace. 50-55 is probably the average. some lucky solves do come around, and it is fun to pretend you skipped a step on purpose.

Last edited:

#### applemobile

##### Member
I switched to Columns for a while as I really like it. But to me, it seems apparent that starting with a cross a la CFOP really does make things easier. Having edge pieces that you need, hidden in the bottom layer makes looking ahead so much harder. You have a lot more freedom with the cube, but I think if you would have a hard time sub 8'ing with it.

Just to add, To make your inspection time as efficient as you can be with roux or cfop, you would need to be able to plan your first 3 pairs. I'm no expert, but I can't see that being easy.

Last edited by a moderator:

#### emolover

##### Member
I tried switching to it and was able to get in the 17's, but I do not think it is sub tenable. Doing four pairs is inefficient because of the hidden edges and placing the 3 edges.

A better method I think would be this.

1) 2x2 or 1x2x3
2) 2x2x3
3) Solve the last two F2L pairs, maybe use WV and SV
4) COLL or CP if WV or SV is used
5) Two look the last 5 edges

#### Ollie

##### Member
I love this, though once I'd solved the F2L slots and the upper corners the temptation was there to use 3-style BLD edgecycles to solve everything else. I'm averaging ~30-35s in the first hour of practicing solving it that way

#### 5BLD

##### Member
Just to add, To make your inspection time as efficient as you can be with roux or cfop, you would need to be able to plan your first 3 pairs. I'm no expert, but I can't see that being easy.
Actually this isn't too hard..

#### Endgame

##### Member
Before we can polish these methods, we should put more effort in building a stable foundation and a better documentation for these methods, but that's just my opinion.

#### Athefre

##### Member
Patch PLL into any method and suddenly it appeals to more people.

#### qqwref

##### Member
It's not just "oh, it's PLL, I'll use this method", though - I honestly do think that when PLL is a choice you need a pretty solid reason to choose something else. PLL has had a ton of experimentation go into every single case, has pretty short and fast algs on average, and has easy recognition (permutation only, so you can literally look at only one sticker from each piece, which also means good lookahead). It's certainly worth it to try to improve on it, but honestly I think the vast majority of ideas end up slower.

#### Athefre

##### Member
If it's fast, it's fast. And PLL is fast. I'm saying that I think it's possible there are fast alternatives that better fit the theme of a method or make the method more its own thing. Or at least wish more people were spending time searching for these alternatives.

#### Ickathu

##### Member
I was actually just thinking about a similar method
4 F2Ls
CMSLL (Corners [ignoring M and S] Last Layer)
F and B centers + DF, DB
cube rotation <<<
LSE

I was able to get sub40 with it after just a few minutes (normally 26)

Last edited:

#### HelpCube

##### Member
I have a 13.30 avg5 and 14.17 avg12 with this columns first method.
I'm getting a bad request error.

This seems semi-decent though, very easy to sub-40 with 5 minutes of practice. Might have to try it some more later.