• Welcome to the Speedsolving.com, home of the web's largest puzzle community!
    You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.

    Registration is fast, simple and absolutely free so please, join our community of 40,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

The WaterRoux 3x3 method thread

efattah

Member
Joined
Feb 14, 2016
Messages
610
I am creating a specific thread to compile all related research & development of the WaterRoux method which was originally discussed in the New Method thread.

MAJOR EDIT April 13 2017: Updated this original post to reflect the current state of WaterRoux.

After extensive discussion and real world experimentation, the description of WaterRoux is as follows:

1. Solve the first block on the left as you do in normal Roux.
2. Solve the DFR and DBR corners in any permutation/orientation with at least one of them having the correct facelet on the D-slice
3. Solve all six remaining corners in one algorithm with TLEG-1 (which is a mix of CLL, LEG-1, and Twisty LEG-1 which has four sets). There are about 252 algorithms if you learn them all but you don't need anything except CLL or CMLL. You can use classic Roux CMLL, but you can use fewer moves with modified CLL since you don't need to preserve the right block.
4. Now the first block and all the corners are solved. Solve two redges (R layer edges) using ERL algorithms (you can find these in the LMCF document or later in this thread).
5. Now you have seven edges left (UL, UR, all the M-slice, and either FR or DR). Now you can finish the solve in MANY different ways, but the best way is Crafto's L7E method which averages only 18 moves and 91 algorithms.

WaterRoux is a rotationless method with extraordinary ergonomics. The average movecount is usually less than LMCF (which was itself the lowest movecount speed method), but WaterRoux has better ergonomics than LMCF and unlike LMCF, WaterRoux doesn't use rotations. The exact average movecount is somewhat TBD but looks to be around 42 moves. Full WaterRoux would be 252 corner algorithms and 91 L7E algorithms. The ERL 'algorithms' are essentially intuitive once you understand how they work. So the full count would be 343 algorithms which is remarkably low considering the efficiency and ergonomics.

Furthermore, it is quite common to be able to solve the first block AND the DBR and DFR (disoriented) corners in 7 moves or less, making it possible to ONE LOOK the left block, and ALL the corners of the cube. This puts the TLEG-1 252 algorithm recognition time in the inspection phase which is a HUGE advantage over ordinary Roux where you must do CMLL recognition in the middle of the solve.

You can find more detailed documentation & discussion later on in the thread.
 
Last edited:

efattah

Member
Joined
Feb 14, 2016
Messages
610
EDIT: Please disregard this post and use Crafto's L7E method which is found later in the thread!!!


----- OBSOLETE ------------------
Okay so I have generated the last seven edges (L7E) algorithm set for WaterRoux. That took a long time. If you include reflections there are 80 cases to solve UL, UR, FR at the same time while orienting all four M-slice edges. For completeness I have created an L7E document which also includes Waterman L6E which is useful in many of the cases as well. Neuro has finished TMCLL+, I will let him post it.

Last 7 edges (L7E) PDF download:
https://drive.google.com/open?id=0B2QnZ3uD6I8kbnRRM0sxSDhHbkk

Last 7 edges (L7E) DOCX download:
https://drive.google.com/open?id=0B2QnZ3uD6I8kZ2k3RXM5eG1IQzA

I am providing the DOCX. As you see in the document many of the cases I provide six or more algorithms that are of decent speed to let the user decide which one they like. This is still somewhat of a work in progress but I have spent WAY too much time generating algorithms/methods lately so I need to take a step back but please feel free to edit and re-post the DOCX file with any corrections or additions so the community can keep developing WaterRoux.
 
Last edited:

crafto22

Member
Joined
Sep 4, 2015
Messages
303
Location
Canada
WCA
2014ADAM03
This seems amazing! Didn't think L7E was actually an achievable concept... I wonder if we could push it any further?

Sent from my Moto G (4) using Tapatalk
 

Neuro

Member
Joined
Dec 23, 2016
Messages
522
I saw a suggestion earlier today that said that we could use EG algs to do the corners. I'm not entirely on board with it but it does sound feasible. I'm currently making a spreadsheet with the TCMLL+ and normal CMLL algs on it, hopefully I will have it done soon. Will add TCMLL- when it's completed.
 

Miro

Member
Joined
Dec 29, 2015
Messages
33
I saw a suggestion earlier today that said that we could use EG algs to do the corners. I'm not entirely on board with it but it does sound feasible.
I think viable is only EG1-style second block (partly second block). First block is fully solved, and second block has oriented, but swapped two corners (DFR, DBR). Recognition for CMLL and EG1-style is the same.
 

Neuro

Member
Joined
Dec 23, 2016
Messages
522
Great to hear on TCMLL, I'll see if I can find them or at minimum contact him to confirm. EG-1 could be good, I can make the algs and come back
 

Mnts

Member
Joined
Dec 26, 2010
Messages
59
Location
Lithuania
WCA
2011SIDA01
Interesting approach, looking forward for a complete set + more examples and a tutorial. From my perspective it seems like extremely hard to master technique although if you put enough practice in it, it might return with benefits. And also how many algs TCMLL and WL7E aproximately have?
 

crafto22

Member
Joined
Sep 4, 2015
Messages
303
Location
Canada
WCA
2014ADAM03
Hey guys would L8E be in any way possible/useful? Just thinking in terms of methods like SSC-M or ECE which can reach L8E in less than 20 moves.

Sent from my Moto G (4) using Tapatalk
 

efattah

Member
Joined
Feb 14, 2016
Messages
610
The EG1 variant is really only plausible if you don't solve any edges in the second block, otherwise the algorithms would be really bad and high move count. Having said that I have already experimented with several of the EG variants even before we created WaterRoux and they were promising. The EG variants take several forms:

Variant 1: Solve first block on the left; solve the two corners in the second block (on the right) in any permutation, execute an EG style algorithm; at this point you are now in a pure Waterman state to finish the cube

Variant 2: Like variant 1, except for the EG1 algorithm you hold the cube with the first block on the back. Using this method about half of the EG1 algorithms work without modification and actually keep the back block undamaged, but you do need to do a y rotation to get into the Waterman finish.

Variant 3: This is the original Waterman-EG, where you solve the first block, put it on the D face at the back, then solve the DFR and DFL corners in any permutation, and execute the EG1 algorithm; again many of the existing EG1 algorithms will work unmodified, but then you need to do a z+x rotation to get into the Waterman finish.

I suck at block building and my TPS is pretty low (3-4), but despite my low TPS, give me a cube that has the first block finished and the corners finished and I can easily finish the cube in 6 seconds. An expert would take 3. It should be possible for good block builders to build the first block and finish the EG algorithm in 2.5 seconds, giving a 5.5 second average.

One of the problems of Roux is the CMLL recognition has to happen in the middle of the solve. LMCF fixes that by moving the corner alg recognition into the inspection, but with some other drawbacks. In WaterRoux-EG, there is a decent chance that an expert MIGHT be able (once in a while) to see the first block and the EG1 or CMLL algorithm in the inspection. That could lead to an excellent solve.
 

efattah

Member
Joined
Feb 14, 2016
Messages
610
Interesting approach, looking forward for a complete set + more examples and a tutorial. From my perspective it seems like extremely hard to master technique although if you put enough practice in it, it might return with benefits. And also how many algs TCMLL and WL7E aproximately have?

CMLL, TCMLL+ and TMCLL- are around 40 algs each so about 120 total, the same as full EG. L7E varies dramatically. Keep in mind advanced Roux LSE already has 120+ algorithms if you use EOLR and memorize all the reflections. The 2-look Waterman L6E also has around 120 algorithms, and the pure L7E is 80 with reflections. L7E can be finished with intuitive solving then doing Roux LSE; it is the bad cases that are the problem and advanced L6E/L7E are designed to deal with those bad cases. The more cases you learn, the more you convert 'bad' cases into 'amazing' cases and actually want them to happen. You can learn subsets of L6E/L7E to deal with the worst cases and gradually learn more cases to deal with the 'medium-bad' cases and so on.

Neuro and I will generate more example solves soon.
 

Neuro

Member
Joined
Dec 23, 2016
Messages
522
I have all of TCMLL+ generated and added all of the CMLL's from alg.db. Will add TCMLL- as soon as I'm done making the algs. Leave a comment asking to help and I'll add you as a contributor. Mean movecount is basically the same between CMLL and TCMLL+ and I'd assume TCMLL- as well, being between 10 and 11 moves. Here's the link:

https://docs.google.com/spreadsheets/d/1alIgdOKLVbHvooQKmdTcrj-Ew1jb-9TzRYNTvpzOgRw/edit?usp=sharing

Also, there'd be 128 corner algs to learn, 42 CMLL's, and 43 TCMLL+ cases as well as 43 TCMLL- cases.
 

Neuro

Member
Joined
Dec 23, 2016
Messages
522
Well it's not quite as easy, but it's still decent and could probably be as fast as Roux LSE. Setup can take a while to get used to, but it's pretty fast. In setup, you can eventually learn to predict your set and whether the LE is in DF or DB. And from there all you need to do is check the orientation of the top edges.
 

Neuro

Member
Joined
Dec 23, 2016
Messages
522
I finished all of the TCMLL+/- algs and put in the CMLL algs for completion sake. Since I was using CubeExplorer, some of the algs are pretty nasty. If you have a better alg, please leave a comment and feel free to ask to become a contributor as well. Here's the link:

https://docs.google.com/spreadsheet...GfvmTDta1tF5-u3VIM58yE5Gug/edit#gid=670959495

**the movecount formula used only works with OBTM, so algs using M will need to have the movecount manually entered
 

obelisk477

Member
Joined
Aug 26, 2010
Messages
1,144
Location
Raleigh, NC
WCA
2009BATT01
YouTube
Visit Channel
I finished all of the TCMLL+/- algs and put in the CMLL algs for completion sake. Since I was using CubeExplorer, some of the algs are pretty nasty. If you have a better alg, please leave a comment and feel free to ask to become a contributor as well. Here's the link:

https://docs.google.com/spreadsheet...GfvmTDta1tF5-u3VIM58yE5Gug/edit#gid=670959495

**the movecount formula used only works with OBTM, so algs using M will need to have the movecount manually entered

link isn't public
 

efattah

Member
Joined
Feb 14, 2016
Messages
610
For those wanting to experiment with the WaterRoux-EG variant:

1. Solve FB on the left (roux style)
2. Solve DFR and BDR corners in a random permutation
3. Execute an LEG-1 or CMLL algorithm to solve the top corners and permute the bottom right corners
(LEG-1: http://www.cyotheking.com/leg1/ )
4. Solve two R-layer edges at once in a pair algorithm (LMCF style) or if possible solve a triplet (UL, and two redges)
5. Finish with L7E or Waterman L6E

The vast majority of the 2x2 LEG-1 algorithms maintain the Roux first block undamaged on 3x3. The few that damage it would need to be regenerated with Cube Explorer.
 
Top