The New Method / Substep / Concept Idea Thread

[Twisted101]

I've been thinking for a long time about how to make KALL into a useful LS/LL approach.

PapaSmurf is correct that it is no better than LS > 2-look OLL > PLL (which has only 31 cases to boot).
It is also hard to justify targeting EPLL as a final step when Z-PLLs are generally some of the slowest.

Here is my proposal:

1. Solve "sideways" Nautilus block-- unsolved is dfR.

Option #1 (4 looks)
2. Arrange dFR into a premade pair.
3. Insert dFR using WVLS.

Option #2 (4 looks)
2. Solve and insert dFR.
3. OCLL

4. "OLLCP/B"
5. L5EP-DR

Now, what is "OLLCP/B"?
It is my unimaginative name for a step very similar to OLLCP/A.
Assuming corners are oriented, OLLCP/A resolves CPLL and EOLL with a single alg.

Assuming corners are oriented, "OLLCP/B" (or "OLLCP/DR" or...) will resolve CPLL and L5EO-DR with a single alg.

This leaves L5EP-DR (almost entirely <RUS>) as a reasonably fast last step.

Are the algs good like OLLCP/A?
I'm sifting through Batch Solver now to find that out.

Does the 5th edge villify this LS/LL approach?

Tim.

Idk if the algs gonna be rly that good for “OLLCP-B”

 

[Averageof12]

1. Solve a 2x2 in BL
2.. Solve a 2x2 in FRU.
3. Solve the rest using your preferred method.
Some examples:
-Corners -> Edges
-Edges -> Corners
-EO -> F2L -> Restricted LL

 

[edward_9x]

Is there a name for algorithmically forcing no dot EO cases during last slot? (ie, at most 2 LL edge flips. Basically ZBLS but without necessarily having a cross on top)

 

[precompsubwr]

Is there a name for algorithmically forcing no dot EO cases during last slot? (ie, at most 2 LL edge flips. Basically ZBLS but without necessarily having a cross on top)

Partial edge control

 

[edward_9x]

Partial edge control

I meant, something like a systemised approach where you recognize the pEO "case" and execute the corresponding alg, without caring if the pair is built, split, in the D layer, etc. You just look at the last pair in whatever state it currently is while checking if a flip needs to be done and where, then execute an alg (not necessarily intuitive) that does the 2-flip and pair simultaneously.

 

[precompsubwr]

I meant, something like a systemised approach where you recognize the pEO "case" and execute the corresponding alg, without caring if the pair is built, split, in the D layer, etc. You just look at the last pair in whatever state it currently is while checking if a flip needs to be done and where, then execute an alg (not necessarily intuitive) that does the 2-flip and pair simultaneously.

Just do it intuitively, there's no use in learning partial edge control for every case if it's not gonna be efficient, just go for either intuitive pec or full zbls

 

[LukasCubes]

Improvement (kind of) on CLL in Waterman

This is weird. So as many of you guys know, I main Waterman. A lot of people know about CLL in Waterman. I just learned all of EG-1 on 2x2 and all of those algs preserve a 1x2x3 block in the back of the bottom layer, so there's that I guess. I want to start learning TCLL soon (not now) and I have a feeling they will preserve the 1x2x3 block as well. Maybe in Waterman, to help with efficiency (but adding hundreds more algs), we can just make a 1x2x3 block and insert the corners however we want.

Variants with algs already genned
CLL, EG-1, TCLL+, TCLL-, TEG-1 A+, TEG-1 A-. TEG-1 C+, TEG-1 C-
43 Algs each set (42 with normal CLL) = 343 Algs (43x7+42=343)

Variants with algs not genned
CLL-T (Corners solved but twisted and it looks like a T case if you do x2)
CLL-U (Corners solved but twisted and it looks like a U case if you do x2)
CLL-T Parity (Same as CLL-T but corners are swapped, not solved)
CLL-U Parity (Same as CLL-U but corners are swapped, not solved)
CLL Gun Right (Corners are solved and twisted, but it looks like a gun case from TCLL facing right)
CLL Gun Left (Corners are solved and twisted, but it looks like a gun case from TCLL facing left)
CLL Gun Right Parity (Same as CLL Gun Right but the corners are swapped, not solved)
CLL Gun Left Parity (Same as CLL Gun Left but the corners are swapped, not solved)
43 Algs each set = 344 Algs (43x8=344)

If there is already Algs genned for these cases, please let me know.

L5C (486 algs + 486 mirrors = 972 Algs) can be used if you just want to do a 1x2x3 block + 1 corner and not two.

Someone (I forget who) texted me on discord wondering if EG-2 Was possible in Waterman. EG-2 on Waterman is a 1x2x2 block on DBL (BL portion of D Layer), and the DR Edge and DFR Corner solved. The swapped corners are DFL and BDR.
There is also a mirror set where the swapped corners are DFR and BDL with a 1x2x2 block on DBR and DL Edge solved.
I guess I can label the top set EG-2 A (43 algs) and the other set EG-2 B (43 Algs) which would total up to 86 algs.

Let's screw this up even more. What if we did an EG version of L5C? 972 more algs to learn, One corner is placed but in the wrong spot.

343+344+972+86+972=2717 Algs in total.

The funny thing is that all of this is still less algorithms than 1LLL.

On another note, I am currently generating Algs for L7E Algs where the L-Layer is solved and 3 of the redges are swapped and flipped in some order on the R-Layer; no redges are in the M-Layer (80 algs). I also finished L7E Algs for if UL is solved, but flipped, and two redges are swapped in some order in the R-Layer; no redges are in the M-Layer. I'm just passed halfway done with the L-Layer solved stuff, however, I might pause genning L7E Algs for the CLL Variants I listed earlier.

Thank you.

Edit: Now that I think more of EG-2 waterman, maybe TEG-2 is possible
TEG-2 A+
TEG-2 A-
TEG 2 B+
TEG-2 B-

43 x 4 = 172
172 + 2717 = 2889

Another Edit: Some people I guess like the twisted corner to be in the back. TEG-2 A will Have the DBR corner twisted and TEG-2 B will have the DBL Corner Twisted

TEG-2 A+ Back (43 Algs)
TEG-2 B+ Back (43 Algs)

43x2 = 86
86+2889=2975

So all this CLL crap is 2975 Algs, STILL LESS THAN 1LLL ON 3x3

 

[Averageof12]

1. EOLine
2. Belt
3. OBL
4. Solve F2L using <R2, L2 U>
5. PLL

Possibly have an algset that solves CLL during Step 4.

Movecount:

1. 8 moves max.
2. 8 moves max.
3. 14-16 moves max.
4. 18 moves max.
5. 14 moves max.

 

[Caleb Valenzuela]

Improvement (kind of) on CLL in Waterman

This is weird. So as many of you guys know, I main Waterman. A lot of people know about CLL in Waterman. I just learned all of EG-1 on 2x2 and all of those algs preserve a 1x2x3 block in the back of the bottom layer, so there's that I guess. I want to start learning TCLL soon (not now) and I have a feeling they will preserve the 1x2x3 block as well. Maybe in Waterman, to help with efficiency (but adding hundreds more algs), we can just make a 1x2x3 block and insert the corners however we want.

Variants with algs already genned
CLL, EG-1, TCLL+, TCLL-, TEG-1 A+, TEG-1 A-. TEG-1 C+, TEG-1 C-
43 Algs each set (42 with normal CLL) = 343 Algs (43x7+42=343)

Variants with algs not genned
CLL-T (Corners solved but twisted and it looks like a T case if you do x2)
CLL-U (Corners solved but twisted and it looks like a U case if you do x2)
CLL-T Parity (Same as CLL-T but corners are swapped, not solved)
CLL-U Parity (Same as CLL-U but corners are swapped, not solved)
CLL Gun Right (Corners are solved and twisted, but it looks like a gun case from TCLL facing right)
CLL Gun Left (Corners are solved and twisted, but it looks like a gun case from TCLL facing left)
CLL Gun Right Parity (Same as CLL Gun Right but the corners are swapped, not solved)
CLL Gun Left Parity (Same as CLL Gun Left but the corners are swapped, not solved)
43 Algs each set = 344 Algs (43x8=344)

If there is already Algs genned for these cases, please let me know.

L5C (486 algs + 486 mirrors = 972 Algs) can be used if you just want to do a 1x2x3 block + 1 corner and not two.

Someone (I forget who) texted me on discord wondering if EG-2 Was possible in Waterman. EG-2 on Waterman is a 1x2x2 block on DBL (BL portion of D Layer), and the DR Edge and DFR Corner solved. The swapped corners are DFL and BDR.
There is also a mirror set where the swapped corners are DFR and BDL with a 1x2x2 block on DBR and DL Edge solved.
I guess I can label the top set EG-2 A (43 algs) and the other set EG-2 B (43 Algs) which would total up to 86 algs.

Let's screw this up even more. What if we did an EG version of L5C? 972 more algs to learn, One corner is placed but in the wrong spot.

343+344+972+86+972=2717 Algs in total.

The funny thing is that all of this is still less algorithms than 1LLL.

On another note, I am currently generating Algs for L7E Algs where the L-Layer is solved and 3 of the redges are swapped and flipped in some order on the R-Layer; no redges are in the M-Layer (80 algs). I also finished L7E Algs for if UL is solved, but flipped, and two redges are swapped in some order in the R-Layer; no redges are in the M-Layer. I'm just passed halfway done with the L-Layer solved stuff, however, I might pause genning L7E Algs for the CLL Variants I listed earlier.

Thank you.

Edit: Now that I think more of EG-2 waterman, maybe TEG-2 is possible
TEG-2 A+
TEG-2 A-
TEG 2 B+
TEG-2 B-

43 x 4 = 172
172 + 2717 = 2889

Another Edit: Some people I guess like the twisted corner to be in the back. TEG-2 A will Have the DBR corner twisted and TEG-2 B will have the DBL Corner Twisted

TEG-2 A+ Back (43 Algs)
TEG-2 B+ Back (43 Algs)

43x2 = 86
86+2889=2975

So all this CLL crap is 2975 Algs, STILL LESS THAN 1LLL ON 3x3

So basically you used part of the eg method and used it to modify waterman?

 

[LukasCubes]

So basically you used part of the eg method and used it to modify waterman?

pretty much but I also have way more variants that have nothing to do with EG at all

 

[Kami]

Ya’ll im really tired and thinkin of weird stuff but theoretically, could a top 2x2 solver that can one look every solve just do ortega or eg-1 for 3x3 bld corners and then just any normal bld edges method afterwards

 

[Kami]

I meant, something like a systemised approach where you recognize the pEO "case" and execute the corresponding alg, without caring if the pair is built, split, in the D layer, etc. You just look at the last pair in whatever state it currently is while checking if a flip needs to be done and where, then execute an alg (not necessarily intuitive) that does the 2-flip and pair simultaneously.

If you see that youre gonna get a dot case then just sledge the last pair in so that you get an L case

 

[Twisted101]

Ya’ll im really tired and thinkin of weird stuff but theoretically, could a top 2x2 solver that can one look every solve just do ortega or eg-1 for 3x3 bld corners and then just any normal bld edges method afterwards

They would have to predict where exactly the edges would be for their bld algs but yes you could do that

 

[Isaiah Scott]

Ya’ll im really tired and thinkin of weird stuff but theoretically, could a top 2x2 solver that can one look every solve just do ortega or eg-1 for 3x3 bld corners and then just any normal bld edges method afterwards

That sounds cool but you’d have to have extreme luck/skill to know where all of the edges would end up after doing the 2x2 algs

If anyone remembers Matyas Kuti, an accomplished blind solver who cheated by looking under the blindfold ~ 15 years ago before sight blockers were required and he literally did 2x2 OLL algs when a site blocker was in place which just messed up all of the edges and got dozens of DNFs in a row when he had very very few official DNFs due to peeking under the blindfold before sight blockers were required. Kind of a run on sentence but I think you get the idea

 

[edward_9x]

Silly LLOB variant
1. CPFB
2. SB
3. DFDB
4. <RrU>-gen 1LLL

 

[FMCubing55]

Method for FMC, invented by me:

1. EO
2. Solving edges (you can leave 1 3-cycle edges; try to solve some corners with edges)
3. Solving corners with insertions

 

[PapaSmurf]

Method for FMC, invented by me:

1. EO
2. Solving edges (you can leave 1 3-cycle edges; try to solve some corners with edges)
3. Solving corners with insertions

This is a method for FMC, but it's not a good one, and it's not invented by you.

Silly LLOB variant
1. CPFB
2. SB
3. DFDB
4. <RrU>-gen 1LLL

This also isn't new. It's a combination of Briggs-2/CPRoux and the RouxFOP method that gets proposed about once per month. Check the first post of this thread for more details.

 

[edward_9x]

This also isn't new. It's a combination of Briggs-2/CPRoux and the RouxFOP method that gets proposed about once per month. Check the first post of this thread for more details.

I mean, as far as I know, this specific combination hasn't been proposed yet. The main idea was just the 3-gen 1LLL subset that doesn't have EO solved

 

[PapaSmurf]

I mean, as far as I know, this specific combination hasn't been proposed yet. The main idea was just the 3-gen 1LLL subset that doesn't have EO solved

Maybe not proposed on this thread, but I've seen it thrown around and dismissed quite quickly as a speedsolving idea a few times now. I've thought about it before, as have many others.

Now on to more abstract musings. It is very easy to step bash. I'll do it right now: solve the DR corners, and 2x2x2 in ubl. Insert the DL edge using <RFES>, then using that sovle LB. Finish the first layer, then solve the rest in 2 algs.
These steps do solve the cube, but are obvoiusly bad. There are lots of steps that solve a cube, but the vast, vast majority of them are bad. Yes, if you're new, you maybe don't understand these things, but it's not a super nuanced concept once you've been cubing for a year or two. What we want is critical thought about either the steps you have bashed together, or the the meta method you have in mind. This is the way we can have breakthroughs. But, as I explain here, they will probably take a long time to actually become relevant. With one big caveat - they actually have to be good. So RouxFOP isn't actually good, in the same way combining two tasty foods probably won't be good (chicken and porridge/oatmeal for example). However, Roux and CFOP are both really good for many reasons.

As this regards this thread, we have 420+ pages of mostly step bashing, a lot of repetition, but a few gems. This is because there have been some people who have thought properly about what it means to make a method good and haven't just dumped thoughts on the internet (always a dangerous activity). You could argue that I'm doing it right now, and maybe that's true, but I have thought about this and commented about this a couple of other times. Let's aim for quality and curiosity.

 

[Averageof12]

Maybe not proposed on this thread, but I've seen it thrown around and dismissed quite quickly as a speedsolving idea a few times now. I've thought about it before, as have many others.

Now on to more abstract musings. It is very easy to step bash. I'll do it right now: solve the DR corners, and 2x2x2 in ubl. Insert the DL edge using <RFES>, then using that sovle LB. Finish the first layer, then solve the rest in 2 algs.
These steps do solve the cube, but are obvoiusly bad. There are lots of steps that solve a cube, but the vast, vast majority of them are bad. Yes, if you're new, you maybe don't understand these things, but it's not a super nuanced concept once you've been cubing for a year or two. What we want is critical thought about either the steps you have bashed together, or the the meta method you have in mind. This is the way we can have breakthroughs. But, as I explain here, they will probably take a long time to actually become relevant. With one big caveat - they actually have to be good. So RouxFOP isn't actually good, in the same way combining two tasty foods probably won't be good (chicken and porridge/oatmeal for example). However, Roux and CFOP are both really good for many reasons.

As this regards this thread, we have 420+ pages of mostly step bashing, a lot of repetition, but a few gems. This is because there have been some people who have thought properly about what it means to make a method good and haven't just dumped thoughts on the internet (always a dangerous activity). You could argue that I'm doing it right now, and maybe that's true, but I have thought about this and commented about this a couple of other times. Let's aim for quality and curiosity.

You can stretch this out as much as you want, but it just says one EXTREMELY flawed statement (And a bunch of self-promotion , which is even worse). Most of this is just "I'm one of a few people actually competent enough to create a good method." You may not like the ideas on this thread, and you are allowed to, but that doesn't mean you're better than everyone else just because you think you are some sort of complete expert. Now, regarding this 'method': Lookahead is completely inefficient. E moves are not as fingertrickable as you might think. Quite frankly, algs for this set in the last step would not be worth learning. Lastly:

I know what you're thinking. Why would I, of all people, be allowed to say anything? Yes, back when I joined, I spammed new method threads. I also spammed this thread. But as you can see, I don't do that anymore. Sorry for the long rant, but I don't regret saying any of this.