# Expansion of sandwich centers

#### rachmaninovian

##### Member
As of now, I know all 12 base algs to solve the last 2 centers, be it opposite or adjacent. My current strategy goes like this:
1. first center in 1-3 algs (3 algs is rare)
2. 2nd center in 1-2 algs
3. last 2 centers in 1 look. (can be opposite or adjacent)
However, this is horridly inconsistent; I often get ugly cases after the first center is solved.
Statistics that I have done a while back have shown that this system produces a 4.5 look average (or so).

A few days ago I thought of a new expansion which could possible solve this problem, and perhaps produce a 3.5 look average, and with center control maybe it can be <3.5 looks on average.

The new strategy would be as the following:
1. first center in 1-3 algs (again, 3 algs is rare.)
2. 2nd center (on U face) in one look; solving from pieces in F and D.
3. last 2 (adjacent) centers in 1 look.

These are the algs I've generated so far (or listed out, whatever) using Clement Gallet's optimal444 solver. Move count is decent, and the algs feel square-1ish. These algs allow U and D faces to be adjusted. Adjusting the F face would be dumb and bad to do because it's easy to mess up while undoing setup moves. All my last 2 centers algs only require AUF to setup, which means I had to learn from 4 angles (or 2, depends =P). If I allow only AUF to setup for these new set of algs, there would be a horrid amount of algs to learn, so I decided against it.

I love clement :3
EDIT: this is for 4x4 only. and I will avoid this system IF the F center can be solved in 1 look.

Last edited:

#### Hyprul 9-ty2

##### Member
Notice how nobody posted here because they're all afraid

#### iSpinz

##### Member
I just don't want to read it

#### Cride5

I've updated the sandwitch page on the Wiki a little, but lots of info is still missing.

@rachmaninovian if you have any more details (especially for the info box on the top-left) it could certainly do with updating..

Also, do you use this as your main 3x3 method? If so, what are the steps?

#### rachmaninovian

##### Member
10 more algs to generate...
Cride: I'm the proposer of this cage variation, I presume? after adopting mizzle's midges method, and mixing ideas from others as well as my own =P
avg number of moves...not too sure, ranges from 120-155.
and yea, I'd be using this method for 3x3 as a competition method though it sucks. me is better at fridrich. just a note, sandwich 3x3 = just a corners first method, nothing special.

Last edited:

#### somerandomkidmike

##### Member
You should switch to corners first for 3x3... Oh, and as for the improvements, these are some good ideas you're coming up with. Using some of your ideas (combined with my blockbuilding variation of this method), I managed to pull off a 1 minute 9 second solve with my new rubik's. If improvements continue to be made on this method, I wouldn't be surprised to see a sub-50 average.

You should try experimenting with blockbuilding. I solve 1 layer (minus 1 edge), and then solve the corners.

Last edited:

#### rachmaninovian

##### Member
I've been experimenting a lot these few days, and with this set of algs I realize that an average of 3.2 looks (just an approximation by gut feel) is possible, with some center control here and there. Very close to my dream of having a 3 look system that k4 ELL has =)

yea I'm gonna use corners first for 3x3 because its cool.
On the other hand, I've tried to experiment with blockbuilding the FL on 4x4 but I simply cannot do it fast enough. Right now I'm mainly using spamming FL corners + CLL which is at this point of time around the same speed as doing ortega (but hey, I used ortega for a long time, and CLL is new to me).
My rationale in doing purely corners first is because the flow does not get broken, and the flow between algs/steps is extremely important in direct solving methods. Move count wise I'd lose out, but the edge I get in recognition cancels out and even triumphs in my opinion.

#### rachmaninovian

##### Member
bump: finished generating all the algs needed DDDDDD

#### Am1n-

##### Member
I use a roux-variation of sandwich: F2B and CMLL to start
After that, I finish the L and R face while building the D-center.
Then, inserting the Down-edges while putting as much F and B centerpieces in place and finnish L2D (with center control if I created a nice case in the previous step).
That way, the most of the time, I only have to do 2 (or 3) center-cycles to finnish up.
I have no idea if it's any good (cfr movecount), but I don't care, I'm not that fast anyway

mvg

#### rachmaninovian

##### Member
hehe thats a variant of mizzle's rouxbyfour.

he does the following steps i think:
1. first 2 centers
2. 1x3x4 block on L
3. 1x3x4 block on R
4. CMLL
5. "as much of D center and DF and and DB dedges"
6. last 2 dedges
7. finish centers.

I pretty much use his middle edges method except I do without the D center and I do use forms of center control either using double layer moves or different algs (I know multiple algs for almost every case)

#### somerandomkidmike

##### Member
I've been experimenting a lot these few days, and with this set of algs I realize that an average of 3.2 looks (just an approximation by gut feel) is possible, with some center control here and there. Very close to my dream of having a 3 look system that k4 ELL has =)

yea I'm gonna use corners first for 3x3 because its cool.
On the other hand, I've tried to experiment with blockbuilding the FL on 4x4 but I simply cannot do it fast enough. Right now I'm mainly using spamming FL corners + CLL which is at this point of time around the same speed as doing ortega (but hey, I used ortega for a long time, and CLL is new to me).
My rationale in doing purely corners first is because the flow does not get broken, and the flow between algs/steps is extremely important in direct solving methods. Move count wise I'd lose out, but the edge I get in recognition cancels out and even triumphs in my opinion.
I suppose I've probably been doing blockbuilding for a much longer time anyway. My first speedcubing method was petrus, so I've been doing it long enough to recognize the cases quickly enough. I've just managed to pull off a 1:05 solve. Maybe you could try Waterman's method for the first layer (the same approach, but on the 4x4). Anyway... You're insanely fast.

#### rachmaninovian

##### Member
somerandomkidmike:
nah, i deproved, am around 57s now.
I need to be high to sub 55.

I only came up with this system because sometimes its irritating to see bad cases for the last 3 centers.

I believe with your efficiency you can be my speed sooner or later. I average about 140 moves, I can imagine you average about 120 or less.

#### somerandomkidmike

##### Member
somerandomkidmike:
nah, i deproved, am around 57s now.
I need to be high to sub 55.

I only came up with this system because sometimes its irritating to see bad cases for the last 3 centers.

I believe with your efficiency you can be my speed sooner or later. I average about 140 moves, I can imagine you average about 120 or less.

I won't average what you do any time soon. I still have my rubik's, and it's terrible. I've had solves where I've gotten to the centers by 50 seconds, but because I couldn't do the fingertricks well enough, the entire solve ended up taking 1 minute 30 seconds. I'll take an average for moves sometime, and I'll post again.

#### rachmaninovian

##### Member
somerandomkidmike:
nah, i deproved, am around 57s now.
I need to be high to sub 55.

I only came up with this system because sometimes its irritating to see bad cases for the last 3 centers.

I believe with your efficiency you can be my speed sooner or later. I average about 140 moves, I can imagine you average about 120 or less.

I won't average what you do any time soon. I still have my rubik's, and it's terrible. I've had solves where I've gotten to the centers by 50 seconds, but because I couldn't do the fingertricks well enough, the entire solve ended up taking 1 minute 30 seconds. I'll take an average for moves sometime, and I'll post again.
I need to get a rubik's and time myself :O then you can really see where you are...

#### somerandomkidmike

##### Member
I took an average for turns a couple of times. I spread it out over the last 2 days. I didn't write it down, but I averaged between 115 and 125 turns. It did vary by the day for some reason.

#### rachmaninovian

##### Member
wow, you are being EXTREMELY efficient for a direct solving method. probably the block building part is saving a lot of moves for you.

my solves range from 120-150 moves, but the average is just slightly below 140 moves (I believe). I did my move counts as what I would do during a solve, so I didn't bother about trying to find the shortest way to do stuff; that's how I believe one should go about doing move counts.

Are you using any form of center control? would be interesting if you knew as much center control stuff as me =P or perhaps more =P I do center control in both stages of midges, either by doing double layer moves (U2 becomes Uu2, etc), or using different algorithms (which is mostly for the last 2 dedges). I do know a trick or 2 here and there so I'm not too rigid in my center controlling...
however, sometimes it does no benefit, or is impossible to do center control though =(

Last edited:

#### somerandomkidmike

##### Member
Yes, I use center control at times. I also just got so close to getting sub-1 for the first time. I had 2 turns left with 56 seconds, and my cube exploded.

#### rachmaninovian

##### Member
wow, with a rubik's? I can't even sub 1:30 on the new rubik's because they lock up like crazy >_>

#### somerandomkidmike

##### Member
My rubik's is fairly broken in. Also I'm 100% sure I wouldn't do as well as you with a meffert's or a QJ because I always look for shorter solutions. Sometimes it's bad for my times. I also can't do fingertricks as well as you, so I won't be able to finish the last step like you can.