ECE - New 3x3 Solving Method

[Shiv3r]

I'm not sure I entirely agree with this... You can reach almost as high tps in intuitve without the pause needed to recognise an alg case. There may not be anything wrong with learning large alg sets per se though I'm not sure it offers a significant advantage over less alg heavy methods. It just depends how good the lookahead for that part of the solve is. For example, people can sometimes reach 8+ tps in F2L for CFOP and that is almost as fast as the alg tps.

To be more succinct, larger alg sets may be nice as they solve more of the cube though at the price of recognition being worse. The only time I would say it is good to have a large alg set is at the very end of the solve unless we can easily predict what will happen next (such as CMLL-LSE in Roux).

My thoughts are that methods will move towards a lot of small steps with excellent lookahead where only a few easily identifiable will need to be tracked (I think LSE is likely the best example of this) and the steps will almost blend into each other ( I tried to do this with SSC and will explain how that works if I ever get around to the video but I think F2L and LSE are good examples of this)

I absolutely agree.
Also, this method has been independently created by 3 People, and who knows how many more. So this method has potential. I think what we need to kick this method off to starhood is someone using ECE as their main method, which means we probably have to find a beginner-friendly version of the method to teach to other people, and have a trail to the advanced method, like CFOP does, so this method takes off.
I believe in this method, and the fact that this method has been pushed to sub-10 proves it.

 

[efattah]

To be more succinct, larger alg sets may be nice as they solve more of the cube though at the price of recognition being worse.

This is true, unless of course you design your method so the huge alg set happens right off the bat and the cuber can use the 15 second inspection time to identify and 'fetch' the algorithm from memory. That's the essence of what I'm working on now. Huge algorithm sets like ZBLL fetched in 'real time' during the solve are a lot more challenging. Theoretically recognition time could drop to near zero but we haven't really seen anyone accomplish this yet for ZBLL (or have we?) The best analogy is EG for the 2x2. Either the cuber can 1-look the whole solve (and fetch 1 of 126 algs from memory during the inspection), or else he cannot see the solve and his time suffers dramatically due to the recognition time required mid-solve. I think CFOP OLL is probably an exception as recognition is so easy (Friedrich designed it specifically for easy recognition, she knew what she was doing...)

Right now ECE/SSC is my #2 favorite method and if my current method doesn't pan out over the next few months I will switch to ECE/SSC as my main method. My only concern with ECE/SSC is recognition on the double-permute steps.

With so many people coming up with 'new' methods, I really think any method needs at least one person to land a single sub-10 solve (even if slightly lucky) to show the potential. Crafto22 already did that.

Also I think we need to agree upon one name for this method. I think probably EPBL is a good option (E-slice, permute both layers). Thoughts on unifying the name?

Eric Fattah
Vancouver, BC

 

[Shiv3r]

This is true, unless of course you design your method so the huge alg set happens right off the bat and the cuber can use the 15 second inspection time to identify and 'fetch' the algorithm from memory. That's the essence of what I'm working on now. Huge algorithm sets like ZBLL fetched in 'real time' during the solve are a lot more challenging. Theoretically recognition time could drop to near zero but we haven't really seen anyone accomplish this yet for ZBLL (or have we?) The best analogy is EG for the 2x2. Either the cuber can 1-look the whole solve (and fetch 1 of 126 algs from memory during the inspection), or else he cannot see the solve and his time suffers dramatically due to the recognition time required mid-solve. I think CFOP OLL is probably an exception as recognition is so easy (Friedrich designed it specifically for easy recognition, she knew what she was doing...)

Right now ECE/SSC is my #2 favorite method and if my current method doesn't pan out over the next few months I will switch to ECE/SSC as my main method. My only concern with ECE/SSC is recognition on the double-permute steps.

With so many people coming up with 'new' methods, I really think any method needs at least one person to land a single sub-10 solve (even if slightly lucky) to show the potential. Crafto22 already did that.

Also I think we need to agree upon one name for this method. I think probably EPBL is a good option (E-slice, permute both layers). Thoughts on unifying the name?

Eric Fattah
Vancouver, BC

I think the ECE name is best, due to the fact that it is the real steps: e-slice edges corners edges

 

[shadowslice e]

I think the ECE name is best, due to the fact that it is the real steps: e-slice edges corners edges

Well, it can basically be like Fridrich/CFOP. Either is fine and they can be interchangable (although ECE is more specific and corresponds to SSC-M rather than SSC-O which is where I would draw the distinction).

 

[Shiv3r]

Well, it can basically be like Fridrich/CFOP. Either is fine and they can be interchangable (although ECE is more specific and corresponds to SSC-M rather than SSC-O which is where I would draw the distinction).

If we could find out crafto22's last name, we could call it the Crafto-Briggs method, or whatever part of Crafto's name he's willing to tell us(kinda like the ZB method, or VH method)

Weve got enough figured out for the method figured out I suggest making a Wiki page. There would already be one except there seems no way to create a new page on the wiki.

Also, The first time a method like this was derived was in the 80's. (look at 10 second cube's post on the new method/substep/concept ideas thread)

 

[shadowslice e]

If we could find out crafto22's last name, we could call it the Crafto-Briggs method, or whatever part of Crafto's name he's willing to tell us(kinda like the ZB method, or VH method)

Weve got enough figured out for the method figured out I suggest making a Wiki page. There would already be one except there seems no way to create a new page on the wiki.

Also, The first time a method like this was derived was in the 80's. (look at 10 second cube's post on the new method/substep/concept ideas thread)

I'm up for renaming it.

There is also a wiki page ([wiki=SSC (Shadowslice Snow Columns)]this one[/wiki]). I've already run it past crafto and he seems ok with all the stuff.

Regarding the 80s stuff, the earliest form of this method I can find is the Thistlethwaite method (though neither crafto or I took inspiration from it). I'm not really sure about the 10 second cube thing as I can't find anything on it other than post (I tried when I first saw it) though I'll admit it does bear a resemblance to it.

 

[efattah]

I think we need a single page that describes the approach in detail, with a list (preferably with graphics of the cases) for each algorithm. Also I think the page should focus on the best variant. Crafto has so many variants I think this is not only confusing, but to make progress here we should (in general) be focusing on the same method. Only with an 'easy to follow' page will we get other people working on the method together, which is what we need to make real progress and develop it.

Eric Fattah
Vancouver, BC

 

[shadowslice e]

I think we need a single page that describes the approach in detail, with a list (preferably with graphics of the cases) for each algorithm. Also I think the page should focus on the best variant. Crafto has so many variants I think this is not only confusing, but to make progress here we should (in general) be focusing on the same method. Only with an 'easy to follow' page will we get other people working on the method together, which is what we need to make real progress and develop it.

Eric Fattah
Vancouver, BC

Basic structure:
1) 3/4 e-slice and orient 3 corners on D
2) Place final e-slice edge and orient remaining corners
3) permute corners
4) finish edges.

I find the best ways to do this for speed are EZD and SSC-O
SSC-Domino is the best for FMC
The one thing I feel need to be improved in any of the methods is the final phase after domino reduction. Atm, the best we have are LEE->LSE and EZD. Both are fairly good but I feel we can do much better.

Other thoughts I put on the wiki page.

I think these are the most useful methods/variants atm. I also have a beginners variant which could go like this:
1) 3/4 e-slice edges (coukd eo or not. Depends)
2) orient 3 corners by inserting corners in the slot like in keyhole
3) pseudopair and insert
4) OCLL
5) separate corners
6) permute corners
7) edges however you want
This needs 7 OCLL+ 4 CP and can be sub-20- I've tried

 

[efattah]

I suppose my question is in the 'main' variant of this method are we forcing corners to the same face color?
Therefore when we permute the corners we are doing Ortega PBL style algs or do we allow the much harder situation where all the corners are oriented but may be on opposite/mixed sides? In that case recognition is really hard.
Same question on the edges.

Eric Fattah
Vancouver, BC

 

[Shiv3r]

I suppose my question is in the 'main' variant of this method are we forcing corners to the same face color?
Therefore when we permute the corners we are doing Ortega PBL style algs or do we allow the much harder situation where all the corners are oriented but may be on opposite/mixed sides? In that case recognition is really hard.
Same question on the edges.

Eric Fattah
Vancouver, BC

no, we use square-1 algorithms to first seperate into their respective layers, then permute all the corners(like 10 algs altogether.)

 

[Shiv3r]

I'm up for renaming it.
Regarding the 80s stuff, the earliest form of this method I can find is the Thistlethwaite method (though neither crafto or I took inspiration from it). I'm not really sure about the 10 second cube thing as I can't find anything on it other than post (I tried when I first saw it) though I'll admit it does bear a resemblance to it.

remember the 10 second cube guy? I know you saw it, you replied to it he went into detail about how he got sub-20 with a method similar to this one.

 

[shadowslice e]

remember the 10 second cube guy? I know you saw it, you replied to it he went into detail about how he got sub-20 with a method similar to this one.

Yeah I know. I actually said it looked like a belt method sort of thing/thistlethwaiteish. However, I meant I could not find any intonation on it at all no matter how hard I looked. I realise this was before the internet but I would've though that it would at least have a small mention somewhere (I would've been happier if he went into more detail on how each of the steps did the stuff he said it did but it seems he's not planning on coming back to speedsolving) Also, I think thistlethwaite's algorithm may have preceded it and that bears a more similar resembalance to SSC-O at least.

On another note, SSC could stand for ShadowSlice-Crafto

(But actual names would be better I admit)

 

[Rouxkie]

I'm also interested in this method. As the only methods I ever looked closer at are CFOP (a bit) and Roux (mainly, avg. ~35s), many concepts are hard to grasp for me, but I think I basically got the idea. Only when I reach LEE I feel quite at home
Here are some thoughts/questions:

• Is the idea to plan step 1 completely during inspection? Getting the 3 edges in place is easy of course, but controlling corner orientation while doing it is much harder. But there are tons of freedom, so maybe it's just a matter of wrapping your head around it, just as with blockbuilding.
• Instead of using WV I just slot the pair in and orient the top layer with CMLL I already know. This costs me an additional 3 moves. Is learning WV or SLS really worth it, or does it have other aspects?
• I'm most concerned with step 3. Separating the corners to the layers takes an insane amount of moves for me. An example would be DU2R2D2R2 == 9 moves QTM, and there are even harder cases. In the SSC-thread it is mentioned that there are many simple cases that only take 1-2 moves. I don't see how that should be possible. Can someone maybe give me an example of such a simple case?
• The following algs for corner perm are also all very long (>15 moves QTM). This is in contrast to the 11 moves (HTM?) the OP mentioned. Maybe he does this in one step?

All in all, does this method really give a lower move count than Roux?

Sorry for those beginner-level questions...

 

[shadowslice e]

I'm also interested in this method. As the only methods I ever looked closer at are CFOP (a bit) and Roux (mainly, avg. ~35s), many concepts are hard to grasp for me, but I think I basically got the idea. Only when I reach LEE I feel quite at home
Here are some thoughts/questions:
[*]Is the idea to plan step 1 completely during inspection? Getting the 3 edges in place is easy of course, but controlling corner orientation while doing it is much harder. But there are tons of freedom, so maybe it's just a matter of wrapping your head around it, just as with blockbuilding.

This one is basically like making a load of pseudopairs at the most basic level. It gets harder one you do Eo as well but it isn't too hard once you get used to it

[*]Instead of using WV I just slot the pair in and orient the top layer with CMLL I already know. This costs me an additional 3 moves. Is learning WV or SLS really worth it, or does it have other aspects?

That's the beginner's variant I described in an eariler post. If you really want to get fast, you should probably learn the others (though in the SSC thread an interesting alternative has been created)

[*]I'm most concerned with step 3. Separating the corners to the layers takes an insane amount of moves for me. An example would be DU2R2D2R2 == 9 moves QTM, and there are even harder cases. In the SSC-thread it is mentioned that there are many simple cases that only take 1-2 moves. I don't see how that should be possible. Can someone maybe give me an example of such a simple case?

Here we are measuring things in ATM rather than QTM as (U D) only takes 1 motion (though most use HTM/STM, that is because they don't have free motion in the D layer)

[*]The following algs for corner perm are also all very long (>15 moves QTM). This is in contrast to the 11 moves (HTM?) the OP mentioned. Maybe he does this in one step?

Yes, again we are using ATM.

All in all, does this method really give a lower move count than Roux?

Potentially, and quite possibly yes with the modifications though it depends how good you are at blockbuilding and how used to this method you are. I believe it may be a fair bit lower once all the developments are complete and better algs created for certian steps.

 

[Shiv3r]

He

I'm also interested in this method. As the only methods I ever looked closer at are CFOP (a bit) and Roux (mainly, avg. ~35s), many concepts are hard to grasp for me, but I think I basically got the idea. Only when I reach LEE I feel quite at home
Here are some thoughts/questions:

• Is the idea to plan step 1 completely during inspection? Getting the 3 edges in place is easy of course, but controlling corner orientation while doing it is much harder. But there are tons of freedom, so maybe it's just a matter of wrapping your head around it, just as with blockbuilding.
• Instead of using WV I just slot the pair in and orient the top layer with CMLL I already know. This costs me an additional 3 moves. Is learning WV or SLS really worth it, or does it have other aspects?
• I'm most concerned with step 3. Separating the corners to the layers takes an insane amount of moves for me. An example would be DU2R2D2R2 == 9 moves QTM, and there are even harder cases. In the SSC-thread it is mentioned that there are many simple cases that only take 1-2 moves. I don't see how that should be possible. Can someone maybe give me an example of such a simple case?
• The following algs for corner perm are also all very long (>15 moves QTM). This is in contrast to the 11 moves (HTM?) the OP mentioned. Maybe he does this in one step?

All in all, does this method really give a lower move count than Roux?

Sorry for those beginner-level questions...

Hey, you are pretty much exactly like me! I average a crappy time(35 secs w roux) and I didnt get this method at first... But just keep trying, and this method will become your favorite!
also, for the first step the easiest way to learn it is actually solve the e-slice edges then work about corner orientation


And also, crafto22, What's your name, so we can actually give this method a real name!

 

[Shiv3r]

UPDATE:
I just generated some algorithms using cube explorer instead of rubiks-cube-solver.com and I got some decently short algs, I made another alg list here.

Note: it does involve s-slice moves, but I suggest doing a Y' move then doing the S moves as M moves.

 

[wir3sandfir3s]

Correct me if I'm wrong, but isn't this just belt with several variations?

 

[sqAree]

Correct me if I'm wrong, but isn't this just belt with several variations?

Belt is more a general idea how to solve the cube than a specific method ; in this respect you are right and wrong at the same time. So it's a belt method but belt methods are meant to have several variations etc.

 

[Shiv3r]

Belt is more a general idea how to solve the cube than a specific method ; in this respect you are right and wrong at the same time. So it's a belt method but belt methods are meant to have several variations etc.

In the same facet, it it way more of a columns first/Human thistlewaite algorithm/orient first method than a belt method. also, you think you could really push a belt method to sub-10 like this method?

 

[wir3sandfir3s]

In the same facet, it it way more of a columns first/Human thistlewaite algorithm/orient first method than a belt method. also, you think you could really push a belt method to sub-10 like this method?

So it's kind of a mix? 3/4 belt then columns?