Noah's CP Block Method 2.0

[TheNextFeliks]

New step 1 approach:

1a: Solve DBL and DFL while permuting the corners, place DL with a slice move when convenient.
1b: Place BL and FL along with the left center.

I have very little 2x2 experience, but I imagine that 15 second inspection is enough time figure out which two corners need to be swapped to solve CP. Does that sound reasonable?



I'll add those. Thanks!

Placing BL and FL is ugly. Besides, not bad.

 

[Noahaha]

Pros and cons of 2gll:

Pros:
-1LLL
-most cases are only R and U
-lots of cases are sune+anti sune combos

Cons:
-crappy recog on some cases
-most cases are only R and U (can be crappy for 2H)
-algs are either long, or super long and RU or okay length and super crappy

judging by my experience with them

If 2GLL is not a worthy goal to work towards, then I guess we'd have to consider what the method would look like without the CP step.

1: Left Roux block
2: EO + Finish Petrus block
3: Right block
4: LL (OCLL/PLL or COLL/EPLL)

I guess the question to ask is whether that's better than Petrus.

Step 1: Roux Block vs 2x2x2
Step 2: Roux-style EO + Two edges vs 2x2x3 + Petrus EO

I'm probably biased, but to me it looks equal at the very least and likely better because of no rotations and much less awkward.

 

[TheNextFeliks]

If 2GLL is not a worthy goal to work towards, then I guess we'd have to consider what the method would look like without the CP step.

1: Left Roux block
2: EO + Finish Petrus block
3: Right block
4: LL (OCLL/PLL or COLL/EPLL)

I guess the question to ask is whether that's better than Petrus.

Step 1: Roux Block vs 2x2x2
Step 2: Roux-style EO + Two edges vs 2x2x3 + Petrus EO

I'm probably biased, but to me it looks equal at the very least and likely better because of no rotations and much less awkward.

2GLL seems great in theory. But for example all L cases take 15 moves. All U and T require at least 13. The idea without cp might be good. It is just recog is bad (why cpls was shot down).

 

[BaMiao]

If 2GLL is not a worthy goal to work towards, then I guess we'd have to consider what the method would look like without the CP step.

1: Left Roux block
2: EO + Finish Petrus block
3: Right block
4: LL (OCLL/PLL or COLL/EPLL)

I guess the question to ask is whether that's better than Petrus.

Step 1: Roux Block vs 2x2x2
Step 2: Roux-style EO + Two edges vs 2x2x3 + Petrus EO

I'm probably biased, but to me it looks equal at the very least and likely better because of no rotations and much less awkward.

Swapping 1 and 2 basically gets you ZZ. I'm definitely biased, but I would prefer doing the EO at the beginning of the solve, and keep the flexibility in the f2l that ZZ has. This is from a ZZ solver, so feel free to ignore.

 

[TheNextFeliks]

I meant cp is hard to recognize, not 2gll.

 

[Kirjava]

Tried looking for something nice for CP detection last night. Nothing good enough to be useful.

1a: Solve DBL and DFL while permuting the corners, place DL with a slice move when convenient.
1b: Place BL and FL along with the left center.

Solving three edges after placing two centres makes the first step way uglier. Does it make CP that much easier? Not sure if worth it.

If 2GLL is not a worthy goal to work towards, then I guess we'd have to consider what the method would look like without the CP step.

Could be a minor improvement to Petrus. At least, an alternative 2x2x3 strat.

2GLL seems great in theory. But for example all L cases take 15 moves. All U and T require at least 13.

2GLL *is* great. Do you know it all? 2GLL is pretty much better than any two look system.

 

[Escher]

Hm, this is a good idea, but I've been thinking about how to improve the CP step here. How does permuting but not orienting any two adjacent corners sound? Recognition of the slightly less simple cases could be done easily with hyper-orientation (or whatever it's called). I believe with enough practise this would eventually develop into an intuitive understanding of how to permute 6 corners (obviously quite suboptimally). Recog of UFL/UBL, and UFR/UBR would be trivial anyway. Once the user roughly understands how to do 6 corners intuitively then the transition from the Roux block to a completed 2x2x3+CP would feel much less artificial, though I'm not sure it's the right thing to make a case-by-case system to learn.

I'm also not sure I like the 2-gen finish of f2l, so I wonder if any compromises can be made to mix the two stages. Even the fastest people rarely get sub 4 on average with 2-gen + 2gll - I would rather be able to do last 2 slots + LL in about 3, if we're using the power of CP -> 2gll properly.

 

[Kirjava]

can't we just work out how to do hyperorientation on 6 corners somehow

 

[antoineccantin]

2GLL *is* great. Do you know it all? 2GLL is pretty much better than any two look system.

Well, I know about 60% of it, and there is a reason I don't know it all. Most of the algs I haven't learned is because the alg really isn't great.
Take the H no-permute case for example. The shortest 2-gen alg is R U R2 U' R2 U' R U2 R U2 R U' R2 U' R2 U R. If you find it too long to be useful, you could also learn this shorter alg; R U2 F2 R' U' F R' U2 F' U2 R2 F U R' F which isn't any better. For that case, and others, I don't find it worth it to do a proper 2gll, and just do OLL/PLL (R U2 R' U' R U R' U' R U' R' z U2 R U R U' R' U' R' U' R U').

PS: I talking about it for OH, and it's even less good for 2H.

 

[Kirjava]

Well, I know about 60% of it, and there is a reason I don't know it all. Most of the algs I haven't learned is because the alg really isn't great.
Take the H no-permute case for example. The shortest 2-gen alg is R U R2 U' R2 U' R U2 R U2 R U' R2 U' R2 U R. If you find it too long to be useful, you could also learn this shorter alg; R U2 F2 R' U' F R' U2 F' U2 R2 F U R' F which isn't any better. For that case, and others, I don't find it worth it to do a proper 2gll, and just do OLL/PLL (R U2 R' U' R U R' U' R U' R' z U2 R U R U' R' U' R' U' R U').

So the worst 2GLL case ends up being just a good OLL/PLL case that you execute like it's one 2GLL alg anyway.

Fail to see the issue?

 

[Brest]

Well, I know about 60% of it, and there is a reason I don't know it all. Most of the algs I haven't learned is because the alg really isn't great.
Take the H no-permute case for example. The shortest 2-gen alg is R U R2 U' R2 U' R U2 R U2 R U' R2 U' R2 U R. If you find it too long to be useful, you could also learn this shorter alg; R U2 F2 R' U' F R' U2 F' U2 R2 F U R' F which isn't any better. For that case, and others, I don't find it worth it to do a proper 2gll, and just do OLL/PLL (R U2 R' U' R U R' U' R U' R' z U2 R U R U' R' U' R' U' R U').

PS: I talking about it for OH, and it's even less good for 2H.

R U' R U R U R' U' R' U' R2 U' R U R' U' R U' R' (20q)
R' U R' U' R' U' R U R U R2 U R' U' R U R' U R (20q)
R U R' U R U' R' U R2 U R U R U' R' U' R' U R' (20q)
R' U' R U' R' U R U' R2 U' R' U' R' U R U R U' R (20q)

 

[Renslay]

I don't know if it is mentioned earlier, but I have to:

1b - PLACE, but do NOT orient the FDR and BDR corners. This should take 4 moves at most.

You can also put the FDR corner to BDR and the BDR corner to FDR. (So the two corners is a swap from each other.) In that case, you have to solve CP for the top corners in a way that there has to be an opposite swap.

 

[elrog]

I proposed a method that solved CP early on in the solve weeks ago also...

I was thinking you could solve CP during inspection to save on recognition time. The only way I have found to do this is doing something similar to the BRASS 2x2 method.
https://sites.google.com/site/devastatingspeed/2x2x2/brass

My method was to solve a 1x1x3 block and corner permutation during inspection, then expand it to a 1x2x3 and build another on the other side. Then place the DF and DB edges while orienting the top edges and finish with 2GLL.

 

[Noahaha]

I proposed a method that solved CP early on in the solve weeks ago also...

I was thinking you could solve CP during inspection to save on recognition time. The only way I have found to do this is doing something similar to the BRASS 2x2 method.
https://sites.google.com/site/devastatingspeed/2x2x2/brass

My method was to solve a 1x1x3 block and corner permutation during inspection, then expand it to a 1x2x3 and build another on the other side. Then place the DF and DB edges while orienting the top edges and finish with 2GLL.

The problem with that it is just Roux with extra steps.

Roux vs. Your method
First block vs. First block + CP = Roux wins by a lot
Second block vs. Second block = same
LSE vs. Orient + place DF, DB = You win by a little bit
CMLL vs. 2GLL = Roux wins by a lot.

The problem is that you are making your CP block and then trying to do Roux. Roux already contains a CP step, however, so there's really no reason to make a CP block. That's why my method transitions to a Petrus-style approach, because Petrus does not already come with a CP step.

I'm not saying my method's any good, just saying that adding a CP step to an existing method and changing the ending a little bit to result in a solve with more and slower steps does not count as coming up with a new method. This was the same problem that my Petrus-style CP-block "method" from a year ago suffered from. It added a step to Petrus that was significantly longer than the LL step saved.

Also, the new thing about my method is not the solving of CP early but the use of Roux-style EO to preserve CP on the way to a Petrus block with EO.

Either way, the challenge is to come up with a good way to do CP from inspection.

 

[elrog]

I get your point. I never expected my method to be quicker than Roux, but I think it can come close.

You know, when I first read CP blocks I was thinking about doing psuedoblocks like in Heise, but instead of solving edge orientation while you fix them, solve CP. Just an interesting idea to throw out there.

 

[Veerexx]

This has been an interesting read About the method itself, I'm not fast enough to have too much insight and I'm only just learning Roux, so I can't really compare it to that. CP from inspection would be great to figure out I think.

 

[ottozing]

I know a lot of what people are talking about is a good way to do CP, but I just feel like saying that 1x2x3 > 2x2x3 isn't as bad as I initially thought it was. After reading Noah's explaination of doing step 2c by setting up the 2 edges at UL and UR, I thought to myself "Hey, doing 1x2x3 > 2x2x3 using mostly that approach could actually be a good idea". Here's some examples using qqtimer's <R,r,U> scramble option.

1 - R' r U R U2 R2 U' r U' R U2 R r' U' R r2 U2 r2 R' U2 R2 U2 R U' R'

R U M' U' R U r2'

Not bad. Not too great either, and kinda hard to see in a speedsolve.

2 - R U r U r' U' R2 U r' U2 R' U' R2 r U R2 U2 r U' r' U' r R2 U' r

r' U M2' U' r

Pretty efficient again, but still kinda complex. Didn't use the UL UR approach because it was sucky for that case.

3 - U2 R' r2 U r2 R' U' R U' r2 U' r2 U R2 r2 U2 r' R' U R' U' r U2 R' U2

M U r U2' r2'

Again, didn't use the UL UR approach. This one was obvious though.

4 - U r' U' r R' U R2 r' U2 r U R U' r R U2 r' R' U R2 U' R2 r' U' r2

U2 r U' R' U r2

Efficient, but complex-ish.

5 - r U' r U' R r2 U' r2 U2 R' U' r U' R2 U' r2 U2 R U R' r' U2 r U r

U r U' M U' r2

This one was pretty obvious.

My thoughts on doing 1x2x3 > 2x2x3 are..... meh. Doing the petrus block this was isn't that inefficient, but the solutions are kinda complicated and would be hard to see in a speedsolve (Assuming you only plan as far as the 1x2x3). With that being said, standard Petrus has the same problem in that doing 2x2x2 > 2x2x3 isn't "inefficient", but the efficient solutions are complicated.

Overall, I would say that 1x2x3 > 2x2x3 is just as good as doing 2x2x2 > 2x2x3 because they have pretty much the same pros and cons. I would love to see a fast Roux solver do an ao12 using either Petrus with 1x2x3 > 2x2x3, or FreeFOP always using a 1x2x3 > 2x2x3 start.

 

[Athefre]

Hm, this is a good idea, but I've been thinking about how to improve the CP step here. How does permuting but not orienting any two adjacent corners sound? Recognition of the slightly less simple cases could be done easily with hyper-orientation (or whatever it's called). I believe with enough practise this would eventually develop into an intuitive understanding of how to permute 6 corners (obviously quite suboptimally). Recog of UFL/UBL, and UFR/UBR would be trivial anyway. Once the user roughly understands how to do 6 corners intuitively then the transition from the Roux block to a completed 2x2x3+CP would feel much less artificial, though I'm not sure it's the right thing to make a case-by-case system to learn.

Hyperorientation is a cool CLL recognition alternative, but how would it apply here?

 

[bobthegiraffemonkey]

What about CP -> <L,M,U> 1x2x3 on left, etc. That is, do CP before building the first block. Find or make DLF/DLB or DRF/DRB, then do whatever to solve CP. Should be easier during inspection, and you can probably look ahead a little to the first block then trace pieces during the CP execution, and probably not too many moves, especially if you have some degree of colour neutrality. This might be a terrible idea, I really don't know.

 

[Kirjava]

It's not exactly clear what you mean, but I don't think that would even preserve CP.