The New Method / Substep / Concept Idea Thread

[TDM]

that is not how this works. it says

it can reach this achievement, but i did not put in enough time to find it yet. it will be updated to sub 40 soon. if i admit i cant find it, i will come up with another excuse of why it should not be banned.

They aren't serious: Matt's post was a joke. It's not going to get banned, don't worry.

 

[Abram Lookadoo]

They aren't serious: Matt's post was a joke. It's not going to get banned, don't worry.

well, i don't catch onto jokes of this style quickly.

 

[Pyjam]

Hasn't Matt been banned for having provided a Petrus solution generated by a genetically modified alien artificial intelligence?

 

[Neuro]

I think that we can make LSE more efficient but it will require a lot of algs. More info to follow. Also, playing with a half-turn redux method; may be good for FMC.

LSE idea isn't panning out too well. HTR is decent right now but it needs a lot of work. Right now I have two schools of thought: corners-first variant or EO2x2x3 into HTR. Seems like CF would be a good approach but I need to do research into cube theory/thistlethwaite before I go any further.

 

[Teoidus]

Maybe look into SSC's EOBelt type starts, but with the addition of CP?

 

[Neuro]

Maybe look into SSC's EOBelt type starts, but with the addition of CP?

Redux to a 2-color state isn't very difficult, but I don't really understand how the CP works in thistlewaite. Any special rules or anything that I need to know?

 

[Teoidus]

Basically the "key swap" has to be in a restricted subset of all possible key swaps. If you track CP through random halfturngroup move sequences, you'll start to get a feel for which key swaps are allowed. (I haven't verified this, but I think all key swaps have to be either in the Point or Singularity quadruples)

Reduction will work similarly to 2GR--you'd just have to make use of quadruplets to manipulate the key swap to one of those allowed key swaps.

 

[Hazel]

I actually think this one might be ok:
1) standard Roux block. Nothing fancy
2) 2x2x1 in RB and pair up an F2L pair
2.5) insert the F2L pair while simultaneously solving 2 adjacent corners on the LL
3) EOStripe <M/U>
4) a subset of ZBLL, much fewer algs since you take away the L, Pi, H, Sune, and Antisune cases as well as a bunch of others

 

[Neuro]

My HT redux method is proving difficult but I think I understand the CP states now

If you look on the top/bottom, the corners work in adjacent pairs with no diagonal cases (they can either be 1-2 moves away from diag swap on top and bottom or solved) or one move away from making pairs (when corners are diagonal on top/bottom)

What this means unfortunately is that the corners absolutely need to be solved a certain way on the top/bottom to truly be solvable with only half-turns so a traditional CP approach will work but it may be difficult to work with.

I think either starting with EO line, 2x2x3, or something similar to Attila's CF approach are the best ways to get to a proper HT state. What's nice is that HT seems to always solve in 12 moves or less and that's w/o allowing quarter turns and the solutions are really easy to see, hence why I wanted to try this

If anyone's confused/interested in this method PM me and maybe we can work together on it

 

[Metallic Silver]

I actually think this one might be ok:
1) standard Roux block. Nothing fancy
2) 2x2x1 in RB and pair up an F2L pair
2.5) insert the F2L pair while simultaneously solving 2 adjacent corners on the LL
3) EOStripe <M/U>
4) a subset of ZBLL, much fewer algs since you take away the L, Pi, H, Sune, and Antisune cases as well as a bunch of others

It sounds like a better method of ZBroux.

 

[Tao Yu]

I actually think this one might be ok:
1) standard Roux block. Nothing fancy
2) 2x2x1 in RB and pair up an F2L pair
2.5) insert the F2L pair while simultaneously solving 2 adjacent corners on the LL
3) EOStripe <M/U>
4) a subset of ZBLL, much fewer algs since you take away the L, Pi, H, Sune, and Antisune cases as well as a bunch of others

The problem I see with this method is that ZBroux has a movecount very close to Roux, and this method has a movecount higher than ZBroux. Hence, this method is definitely worse than Roux.

I think a better idea would be to use EOstripe + ZBLL only when certain CMLLs come up, and do normal Roux otherwise. For each CMLL you intend to do EOstripe +ZBLL for, you only have to learn 12 algs, and you can choose which CMLLs you want to do this for: You can simply do normal roux whenever a sune, pi or H CMLL comes up. This allows you to use ZBroux in your solves without learning too many algs, avoiding the bad cases, and without sacrificing movecount.

Whether or not this is worth it is another matter, but I think it would be better than your method.

 

[Neuro]

The problem I see with this method is that ZBroux has a movecount very close to Roux, and this method has a movecount higher than ZBroux. Hence, this method is definitely worse than Roux.

I think a better idea would be to use EOstripe + ZBLL only when certain CMLLs come up, and do normal Roux otherwise. For each CMLL you intend to do EOstripe +ZBLL for, you only have to learn 12 algs, and you can choose which CMLLs you want to do this for: You can simply do normal roux whenever a sune, pi or H CMLL comes up. This allows you to use ZBroux in your solves without learning too many algs, avoiding the bad cases, and without sacrificing movecount.

Whether or not this is worth it is another matter, but I think it would be better than your method.

This could be a really good proposition especially for OH. If you see a 2G case, all you need to do is EODFDB and end in super-fast 2GLL. I'll do some research on this and see what happens!

 

[efattah]

The problem I see with this method is that ZBroux has a movecount very close to Roux, and this method has a movecount higher than ZBroux. Hence, this method is definitely worse than Roux.

Congratulations on your sub-10 average ZBRoux video. I suppose that makes you the only (?) person to get a sub 10 average with a 'new' (or significantly 'new') method in this thread? Maybe we could make a table of the more promising 'new' methods with the best known averages and singles. Crafto (I think) had an 8.xx single with ECE and a 12.xx average.

 

[Neuro]

Here's a solve using an FMC-oriented method that reduces to Domino state. While I think that Thistlewaite reduction is possible, domino/Kociemba redux is signifigantly easier to see/do. The solve is a bit tricky after that, wso we'll need to edevise soemthing similar to Attila's pair-building method.

U2 L2 B L2 R2 B2 F' D2 F' L2 F' R' B R2 F2 L B' D F2 D F'

F R' F' B U R' U' L'//EOF2L-1 (8/8)
F2 D R' U R F2//Domino Redux (6/14)
R2 U2 F2 U' B2 U' L2 B2 U' F2 D L2 D//Solve (13/27)

What I realized in trying to create these methods was that you can check every EO "line" from only 6 out of 24 possible center positions by abusing FB symmetry and the fact that you can solve the line on top or bottom.

Right now I check all lines and look for ones with easy F2L blockbuilding. Once that is found, any method can be used to reduce to domino state; although my preferred is a CLS/ELS based approach. And then it's a matter of solving the cube like a domino!

 

[Teoidus]

I have a 9.95 2GR single and a 19.52 best avg5.

 

[JTay]

I have an 8.69 Ribbon single and a 10.11 best avg 5.

 

[Tao Yu]

Congratulations on your sub-10 average ZBRoux video. I suppose that makes you the only (?) person to get a sub 10 average with a 'new' (or significantly 'new') method in this thread? Maybe we could make a table of the more promising 'new' methods with the best known averages and singles. Crafto (I think) had an 8.xx single with ECE and a 12.xx average.

ZZ-CT maybe? I know that sub 10 averages have been done with it. (Can't really say it's a promising method though, since ZZ-a is objectively better).

I think a list like that would be cool. I imagine times will probably reflect how much practise people have had with their methods, and how similar they are to their main methods more than anything, but it would still be interesting.

 

[Neuro]

Are there any (good) speedsolving methods that reduce the cube into a completely 2-gen state where R and U are scrambled?

2GR, modified mDiPalma's CPLC variant, Briggs v1 (CPFB, EO Line), and Porkyv1/2. These are all the one's I can think of.

Briggs v1 could be really good as CPFB could potentially be done fully in inspection and EO Line can be made algorithmic.

Modded mDP would take a while to get recognition down but it could be incredibly fast. Same with Porky's variants

2GR has great potential, but just go to Teoidus's website for it; he can explain it much better than I can

 

[shadowslice e]

Congratulations on your sub-10 average ZBRoux video. I suppose that makes you the only (?) person to get a sub 10 average with a 'new' (or significantly 'new') method in this thread? Maybe we could make a table of the more promising 'new' methods with the best known averages and singles. Crafto (I think) had an 8.xx single with ECE and a 12.xx average.

I have a sub-10 Ao5 with a "half and half" 42 since I don't know all the required cases just yet.

Here's a solve using an FMC-oriented method that reduces to Domino state. While I think that Thistlewaite reduction is possible, domino/Kociemba redux is signifigantly easier to see/do. The solve is a bit tricky after that, wso we'll need to edevise soemthing similar to Attila's pair-building method.

U2 L2 B L2 R2 B2 F' D2 F' L2 F' R' B R2 F2 L B' D F2 D F'

F R' F' B U R' U' L'//EOF2L-1 (8/8)
F2 D R' U R F2//Domino Redux (6/14)
R2 U2 F2 U' B2 U' L2 B2 U' F2 D L2 D//Solve (13/27)

What I realized in trying to create these methods was that you can check every EO "line" from only 6 out of 24 possible center positions by abusing FB symmetry and the fact that you can solve the line on top or bottom.

Right now I check all lines and look for ones with easy F2L blockbuilding. Once that is found, any method can be used to reduce to domino state; although my preferred is a CLS/ELS based approach. And then it's a matter of solving the cube like a domino!

Isn't this SSC?

 

[Neuro]

I have a sub-10 Ao5 with a "half and half" 42 since I don't know all the required cases just yet.

Isn't this SSC?

Nope I tend to use EO Line, blockbuilding to F2L-1, advanced TSLE to orient the rest of the puzzle, and solve as a domino. This isn't an exact method much like Petrus FMC, so I'm sure you could use SSC to solve it, but the overarching theme is to do EO, blockbuilding or some other way to get to nearly domino, solve into and then fully solve domino. But I can see the confusion between this and SSC as they are fairly similar methods.