Hawaiian Kociemba 3x3x3 TUTORIAL - Method of Michael H.

[wir3sandfir3s]

As already stated, mid 30s are hardly proof for the method's utility in FMC. I've had a couple 30s before that essentially use CFOP, and could get a 41 or less using CFOP on just about any scramble. Still, I'd never argue for CFOP being an effective FMC method.

And yes, normally the goal of FMC is to effectively balance orienting edges with building blocks in an efficient manner. I don't mean to say that this method is taking a potentially poor approach to FMC, as in some cases it could very well be effective. It's just that this one tool is far too narrow to be used in FMC, and it lives in a much larger field of approaches that should be considered when doing FMC. ZZ is another example of a method that is too narrow to be efficiently used in FMC -- although simply adapting EOLine to just EO would be more in line with many top FMC solves now, so it's much closer.

I know the method does not take advantage of edge insertions itself, but since L5E is the last step, using this method as a skeleton for edge insertions is a viable argument for this being a method for FMC. I still don't think it's worthwhile, given the arguments I have already made.

Also, I hardly believe the burden of proof is provided for this being a viable speedsolving method. A (potentially joke/setup) video with a few solves and a list of solutions that were not found in a speedsolve setting is far from it. There's potential for it to be a competitive, but move efficiency does not always translate into speed. Otherwise, Petrus or Heise would be far more viable as speedsolving methods than they actually are.

Now I believe I see what you mean with the restrictions. I plan on adding a FMC variant soon with a much less restricted F2L -1 and a move efficient commutator based LL. Hope it works
So for speed solving, I believe we have already seen proof that this is quite a good method for speed.
1. Alex Maass's video
2. Reconstructions of other people's solves (can be found on the master list)
I would still like to hear any thoughts or arguments you might have, as they might help me in the future and I just like (friendly) arguing with people.

 

[Teoidus]

can i see an example solve of the proposed variant?

https://www.speedsolving.com/forum/showthread.php?48028-The-Roux-quot-Example-Solve-quot-Game!
I think if you're going to solve unoriented edges last and start the solve with pure blockbuilding, it might be better to stick with Roux--LSE is pretty good and FB/SB is (or at least has been so far proven to be) faster.

also i dont think it would make much difference doing either eo of all edges or eo of only f2l edges

Disagree. I think the main advantage of a step like eoline is not so much the rotationless F2L (theoretically, a standard CFOP solve needs only 1 rotation) as it is the options you get on LL with all the edges oriented already. In this case, you can use COLL to reduce the cube to <R,U>, and then finish with a nice edge perm.

 

[G2013]

Let's do an example solve
Random scramble from Cubetimer:
B2 L U' F' R2 L2 U' D2 L2 R2 D B' F2 D' F' U2 D2 R D B' F L R' B2 U2

x2 R2 D F L' F U2 F' B2 y2 //EOArrow
R' U' R' U R' //1st pair
R' U R U2 L U L' //2nd pair
L' U L R' U R //3rd pair
U L' U2 L U2 L' U L //4th pair
l' U' L U R U' L' U x' //HKOLL
d' M' U2 M U2 R' U2 R U2 l' U l U R' U' R' F' R2 U //HKPLL + AUF

63 moves HTM, 61 with cancellations
I used reduced HKPLL

 

[JTWong71]

How about like this:

F D' F' L2 F R U D' l //2x2x3
R2 U R' U' R B U f' U' L //Psuedo-F2L
U' r U R' U R U2 r' //HKOLL(CP)
M2 U' M U2 M' U' R2 //HKEPLL

34 STM

 

[wir3sandfir3s]

Let's do an example solve
Random scramble from Cubetimer:
B2 L U' F' R2 L2 U' D2 L2 R2 D B' F2 D' F' U2 D2 R D B' F L R' B2 U2

x2 R2 D F L' F U2 F' B2 y2 //EOArrow
R' U' R' U R' //1st pair
R' U R U2 L U L' //2nd pair
L' U L R' U R //3rd pair
U L' U2 L U2 L' U L //4th pair
l' U' L U R U' L' U x' //HKOLL
d' M' U2 M U2 R' U2 R U2 l' U l U R' U' R' F' R2 U //HKPLL + AUF

63 moves HTM, 61 with cancellations
I used reduced HKPLL

You used the method to the very least of its efficiency... Still thanks for the solve.

 

[JTWong71]

Although when I am doing most of the reconstructions, I am not doing the "Standard" way by making an EOArrow, but instead a 2x2x3 block (Like Petrus).

 

[wir3sandfir3s]

Although when I am doing most of the reconstructions, I am not doing the "Standard" way by making an EOArrow, but instead a 2x2x3 block (Like Petrus).

Petrus variation it's HK, your fine lol

 

[wir3sandfir3s]

Fixed CP+EO variant, was not working before.

 

[JTWong71]

After using the Petrus Variant of Hawaiian Kociemba, I believe it is still possible to save even more moves past 40 (Average), but will probably take either more algorithms or more intuition. Most times after you solve the EO step, a CE pair is either made or is one move away from being solved, and you might want to use that. Sometimes if you see that easy pair, it will have the last bottom layer edge, and HK doesn't suggest you solve the whole F2L before the Last Layer. What may save some moves is solving the F2L-1 Piece, but that piece doesn't have to be that edge, it could be a corner. Then, where possible algorithms come and/or intuition, you could solve up to the point where there are 3 pieces left to solve, and a commutator can be used to solve the Last 3 Pieces.

Example:
z2 //Inspection
D' F D' B2 U' L D2 F D' //2x2x3
U2 F R' F2 U F U //EO + Block
R D2 R F' D2 R D2 R' F R' //AB3C (Cancelled 'D' Move)
D R' D F2 D' R D R' F2 R //L3C

36 STM

I also have a 32 STM Solve, but really it was easier to just solve the whole F2L and leave 3 Corners to solve, making it technically Petrus with a L3C Variant.

z2 //Inspection
R U D2 B D //2x2x2
R' U F U F' R F' //2x2x3 + EO
U R2 U2 R' U R' U R U2 R2 U' //F2L + LLE + 1C
L2 D L' U2 L D' L' U2 L' //L3C

 

[G2013]

(Replying to who replied to my example solve) I know xD Anyway, it's an example solve.

I'm still trying to understand this new layout :S

 

[adimare]

This method seems to be ZZ without solving one of the cross pieces. What's the point of doing that? You save very little time and moves by not solving one piece and in exchange you get a much more complicated oll/pll phase. Also, in 1 out of every 6 solves the last edge needed to finish the F2L is trapped in place of the cross piece.

 

[JTWong71]

I think the main point of the HK method is to solve F2L-1 Piece. Some of the variants require you to solve full EO. Leaving the 'R' moves to move as much anywhere can help save a few moves. Since I am using the Petrus Variant, I build the 2x2x3 first, then solve EO, but then leave one F2L piece unsolved, and solve the rest with different HK algorithms or using a Heise type way to leave 3 corners left to solve if my missing F2L piece is a corner.

 

[wir3sandfir3s]

After using the Petrus Variant of Hawaiian Kociemba, I believe it is still possible to save even more moves past 40 (Average), but will probably take either more algorithms or more intuition. Most times after you solve the EO step, a CE pair is either made or is one move away from being solved, and you might want to use that. Sometimes if you see that easy pair, it will have the last bottom layer edge, and HK doesn't suggest you solve the whole F2L before the Last Layer. What may save some moves is solving the F2L-1 Piece, but that piece doesn't have to be that edge, it could be a corner. Then, where possible algorithms come and/or intuition, you could solve up to the point where there are 3 pieces left to solve, and a commutator can be used to solve the Last 3 Pieces.

Example:
z2 //Inspection
D' F D' B2 U' L D2 F D' //2x2x3
U2 F R' F2 U F U //EO + Block
R D2 R F' D2 R D2 R' F R' //AB3C (Cancelled 'D' Move)
D R' D F2 D' R D R' F2 R //L3C

36 STM

I also have a 32 STM Solve, but really it was easier to just solve the whole F2L and leave 3 Corners to solve, making it technically Petrus with a L3C Variant.

z2 //Inspection
R U D2 B D //2x2x2
R' U F U F' R F' //2x2x3 + EO
U R2 U2 R' U R' U R U2 R2 U' //F2L + LLE + 1C
L2 D L' U2 L D' L' U2 L' //L3C

All your ideas have been great so far, thanks for posting this this will probably be part of the FMC variant which us coming soon.
I also thing any F2L piece on the R face can be left unsolved to be solved using a commutator.

 

[wir3sandfir3s]

This method seems to be ZZ without solving one of the cross pieces. What's the point of doing that? You save very little time and moves by not solving one piece and in exchange you get a much more complicated oll/pll phase. Also, in 1 out of every 6 solves the last edge needed to finish the F2L is trapped in place of the cross piece.

You are somewhat correct, but there are benefits. It's helps the F2L flow easier and gives more option for LL, most algorithm sets can be used from various methods, even method specific ones. The main point of this method is freedom, which you can see in the example solves (especially JTWong71's).

Also, if you read the beginning of the thread more closely, k oringally posted this thread as a THEORY of what Michael H's method could be, and that's what he appeared to do. Pure HK does not seems to be the most efficient, and that's when I made variants which made this method awesome.

 

[adimare]

It's helps the F2L flow

It doesn't. Have you tried speedsolving with this method? Quite often an edge piece you need for f2l is trapped in the position of the cross piece that you inexplicably refuse to solve in the first stages; when that happens your f2l flow is completely destroyed.

easier and gives more option for LL

Having to solve 9 pieces instead of 8 during the LL just increases the amount of algs you need to memorize from dozens to hundreds, makes recognition way harder, and greatly reduces the chances of skipping any of the steps.

The main point of this method is freedom, which you can see in the example solves (especially JTWong71's).

Have you paid attention to his example solves? First of all, let's be clear about the fact that they're FMC attempts, not speed solve reconstructions. Also, they have nothing to do with this method... I'm pretty sure he's just trolling you by presenting mediocre FMC attempts and calling them HK variants. He's just doing standard AB3C skeletons and solving the last 3 corners with a commutator.

Pure HK does not seems to be the most efficient, and that's when I made variants which made this method awesome.

Please name a variant that is "awesome" for speedsolving and provide at least one example solve of such variant.

 

[JTWong71]

Have you paid attention to his example solves? First of all, let's be clear about the fact that they're FMC attempts, not speed solve reconstructions. Also, they have nothing to do with this method... I'm pretty sure he's just trolling you by presenting mediocre FMC attempts and calling them HK variants. He's just doing standard AB3C skeletons and solving the last 3 corners with a commutator.

When I do reconstructions, I don't spend an hour doing an FMC solve (Around 5-10 Minutes). What I do is different than in a Speedsolve, because I don't have to keep of pieces as quickly. Also, for the L3C-Setup, I will mainly use Cube Explorer to solve till that point. I have been testing blockbuilding for a bit, and I can usually solve a 2x2x3 in around 10-13 moves, depending on my inspection. What I am trying to do is mainly see how far the method can go. The step before L3C is to see how the move count is affected by doing a different type of step, other than HKOLL, HKPLL. Right now Speedsolving methods average around 50-60 moves. I have never done FMC before or any cubing competition. I am just testing out ways to solve the cube that could possibly be used in a Speedsolve. I got the LLE+1C, L3C Step, mainly based off on looking at the Snyder method and Heise method, not FMC techniques.

Edit: Also, if Speedsolves were to lower movecount to around 35, what techniques do you think we would use? Some FMC techniques are already applied to Speedsolves, such as Blockbuilding. SSC/ECE, a method that averages around 45 moves, is a speed-optimized method from the Two-Phase Computer System. It can later be possible to lower the speed-optimized system even lower, by changing some things. I am mainly saying that when a Speedsolve's movecount lowers, they will likely resemble FMC, at least a little more than the usual 50-60 move Speedsolves.

I am not trying to troll.

 

[adimare]

When I do reconstructions, I don't spend an hour doing an FMC solve (Around 5-10 Minutes).

You still shouldn't call them "reconstructions". That word makes it seem as though you did a speed solve, recorded it, and then reconstructed the steps you followed. 15 minutes or an hour, an FMC solve is an FMC solve.

Also, for the L3C-Setup, I will mainly use Cube Explorer to solve till that point.

Sequences generated by Cube Explorer usually don't translate into viable speed solving approaches.

I have been testing blockbuilding for a bit, and I can usually solve a 2x2x3 in around 10-13 moves, depending on my inspection.

That's a solid start for FMC. If you take a couple of hours to look into NISS, pseudo blocks and commutator insertions you'll be coming up with 0 move solutions in no time.

What I am trying to do is mainly see how far the method can go.

Seems to me that in doing so you've completely deviated away from the proposed method and are currently rediscovering FMC techniques while giving false hope to people who have faith in this method.

I have never done FMC before or any cubing competition.

You should, you seem to have a talent for it.

I got the LLE+1C, L3C Step, mainly based off on looking at the Snyder method and Heise method, not FMC techniques.

I'm not too familiar with Snyder, but Heise is pretty much the core method used for FMC, I don't think anyone considers it a speed solving method.

I am mainly saying that when a Speedsolve's movecount lowers, they will likely resemble FMC, at least a little more than the usual 50-60 move Speedsolves.

The problem is that the things that make an FMC method good are not the same things that make a speedsolving method good. Otherwise, everyone would be using Heise for speed solving instead of CFOP. A good approach to FMC solves as many pieces as possible in a few moves; but in speed solving, dealing with more than 2 pieces at a time (sometimes 3 in Roux or ZZ) completely destroys lookahead.

I am not trying to troll.

I believe you now, I apologize for saying that. But I do think you're misleading the proponents of this method by posting solutions with low move counts that you got by using Cube Explorer.

 

[JTWong71]

But I do think you're misleading the proponents of this method by posting solutions with low move counts that you got by using Cube Explorer.

The only time where I will use Cube Explorer is when there is something that isn't always the easiest to solve intuitively with close to optimal moves for the step.

 

[JTWong71]

dealing with more than 2 pieces at a time (sometimes 3 in Roux or ZZ) completely destroys lookahead.

After thinking a little bit, I saw that look ahead would be difficult. When I think about it, only solving 2 pieces at a time on F2L will take at most 6 looks. I think it is possible to tracks 3-4 pieces at a time, with practice and as long as all the places that the pieces can be are in sight without having to slightly turn your head (Such as a 2GF2L). I am mainly trying to figure out a possible way that could bring down the move count and be a fast Speedsolving method. I try to throw out some ideas, and I will look at them later. I should probably then just think about the idea a little more before proposing them, as I believe I try to think too hard on how to lower move count than on lookahead.

Edit: I do think it is possible to get Speedsolving move count around 40 or below on average.

 

[wir3sandfir3s]

It doesn't. Have you tried speedsolving with this method? Quite often an edge piece you need for f2l is trapped in the position of the cross piece that you inexplicably refuse to solve in the first stages; when that happens your f2l flow is completely destroyed.
Having to solve 9 pieces instead of 8 during the LL just increases the amount of algs you need to memorize from dozens to hundreds, makes recognition way harder, and greatly reduces the chances of skipping any of the steps.

Have you paid attention to his example solves? First of all, let's be clear about the fact that they're FMC attempts, not speed solve reconstructions. Also, they have nothing to do with this method... I'm pretty sure he's just trolling you by presenting mediocre FMC attempts and calling them HK variants. He's just doing standard AB3C skeletons and solving the last 3 corners with a commutator.

Please name a variant that is "awesome" for speedsolving and provide at least one example solve of such variant.

Lol you seem to have this thought in your head that you absolutely have to not solve an edge piece for this method to work...
Dozens to hundreds? For pure HK, but there is a 14 alg 2LLL for one variant. Again,more HK isn't practical and I'm aware. That's all you are talking about so far.
The Petrus variant is the best so far. evidence is Alex Maass's video of him Speedsolving with the method And getting times faster than he usually gets with a good average for him, better than his main method, only a few minutes after learning the method.

I am not exactly sure where your argument is, what you have against the method, and why. You are just repeating yourself.