M2 or BH/Freestyle for 3x3 edges?

[Sakarie]

I dont' think I'm alone with having this question, or wondering. There is no definite answer, but I want to know what you think.

M2:
Pros: It's very easy, and very fast "recognition", meaning that you can do the "alg" to solve a piece immideatly when you know which one.

Fast cases, where you always (sometimes after a x') solve a piece using RUM or LUM only.

Cons:
You solve one piece at the time, which usually is slower than solving two pieces at the time. A lot of moves (compared to commutators). Fixed buffer.

BH/Freestyle:
What I'm thinking about is commutators, which can be predefined or made up as-you-go. The reason I'm defining that it is on a 3x3 is because halcslice planes (M U2 M' U2) is very useful.

Pros:
You solve two pieces at the time. You use very few moves, often move-optimal algorithms. You have the possibility to freestyle, like using PLL's, or whatever you want. Chance to change buffer, which may save moves.

Cons: Very many cases (20x22=440), so harder to recognise, or slower to make up the correct commutator.

---

I'm thinking very long long-term, which one would be the best with a year or two of practise.

 

[mazei]

I can't say, but so far with M2 I know of people who sub-1(regardless of the corner part). But if you see the top BLD-ers for now, they're using freestyle. I guess in a long run freestyle might be the way because after plenty of practice it becomes like F2L where you don't think of the algorithm anymore, you just do it.

 

[deadalnix]

BH for corners is really intuitive. Less for edges due to a bigger set of commutator type (at least, it's the case for me).

I would recommand DIADEM, but I'm not sure I'm the more objective people here about that.

 

[Mike Hughey]

BH for corners is really intuitive. Less for edges due to a bigger set of commutator type (at least, it's the case for me).

I would recommand DIAMDEM, but I'm not sure I'm the more objective people here about that.

My recommendation would be this:
BH corners
M2 edges

Then, when you get comfortable and are looking for improvements, go to DIADEM. I use something similar to DIADEM, except I never have bothered to really read and understand deadalnix's exact approach. I understood an early post he made as to the general concept of the technique, and then started "redeveloping" it myself. As a result, I do a number of things differently from him, but the concepts are still mostly the same. The nice thing is that if you learn the concepts behind DIADEM, you can just sort of optimize M2 cases a few at a time, and then slowly improve your M2 on an incremental basis - you don't have to do it all at once. There are still some cases I do where I could improve them with DIADEM, but I haven't bothered because they're not all that bad to do with straight M2. I still hope someday to become less lazy and actually work out speed-optimal approaches for every possible edge pair, but I just haven't found the time for that yet.

Remember the fact that you can always optimize the methods by substituting in better fingertrick-friendly algorithms on this. I think of BH and M2 as a good place to start, and from there you can build, algorithm by algorithm, to an optimal freestyle method.

 

[cmhardw]

BH for corners is really intuitive. Less for edges due to a bigger set of commutator type (at least, it's the case for me).

I would recommand DIAMDEM, but I'm not sure I'm the more objective people here about that.

My recommendation would be this:
BH corners
M2 edges

Then, when you get comfortable and are looking for improvements, go to DIADEM. I use something similar to DIADEM, except I never have bothered to really read and understand deadalnix's exact approach. I understood an early post he made as to the general concept of the technique, and then started "redeveloping" it myself. As a result, I do a number of things differently from him, but the concepts are still mostly the same. The nice thing is that if you learn the concepts behind DIADEM, you can just sort of optimize M2 cases a few at a time, and then slowly improve your M2 on an incremental basis - you don't have to do it all at once. There are still some cases I do where I could improve them with DIADEM, but I haven't bothered because they're not all that bad to do with straight M2. I still hope someday to become less lazy and actually work out speed-optimal approaches for every possible edge pair, but I just haven't found the time for that yet.

Remember the fact that you can always optimize the methods by substituting in better fingertrick-friendly algorithms on this. I think of BH and M2 as a good place to start, and from there you can build, algorithm by algorithm, to an optimal freestyle method.

Hi Mike,

Very well thought out response, and I completely agree with you on M2 and BH being good beginning points, and you can optimize or personalize to your own techniques afterward.

I still feel that an edge method that speed optimizes the 3x3x3 using Half Slice-Planes or Half Slice-Planes with setup turns as much as possible would be a very formidable speed approach to 3x3x3 edges BLD.

Take the following cycle: UB -> DL -> DR
easily executed with a Half Slice-Plane by doing: D M' D2 M D

Now, before you comment I agree that those turns are all very awkward to execute, but there are only 5 of them to execute.

Also cycles like UB -> UF -> FR could be done by R' y U' M' U2 M U' y' R

These algs will still have awkward turns in them, but the advantage of being even shorter than the average length for supercube safe BH, I feel, would add a significant advantage over any BH solver, as well as any M2 solver. I make no claims about this vs. Freestyle still.

Also, I know I am not the fastest BLD cuber out there, but my 1:23.xx solve in the final round at Worlds 2009 was done with pure BH corners and pure BH edges. Sure that's not world class sub-50 or sub-40 by any means, but at least BH edges has the potential for sub-90 second solves in official competition. And that's using supercube pure BH for edges. Imagine what Half Slice-Plane optimized BH edges might be like. I am excited at the idea!

Chris

 

[Mike Hughey]

I still feel that an edge method that speed optimizes the 3x3x3 using Half Slice-Planes or Half Slice-Planes with setup turns as much as possible would be a very formidable speed approach to 3x3x3 edges BLD.

Take the following cycle: UB -> DL -> DR
easily executed with a Half Slice-Plane by doing: D M' D2 M D

Yes, this is very nice, and really not that hard to execute with ring finger moves for D - it's actually a pretty fast algorithm. But if you simply take as your buffer piece DF instead of UB, the move for the equivalent case becomes U M' U2 M U, which is even easier. (And that's exactly what I use for that case.) And I can even use this on big cubes. (Deadalnix says this works on 3x3x3's, but I use it for big cubes too, so I don't see the benefit of his much longer algorithm.) I stick with DF for my buffer piece because it allows me to use M2 algorithms wherever they're beneficial.

Also, I know I am not the fastest BLD cuber out there, but my 1:23.xx solve in the final round at Worlds 2009 was done with pure BH corners and pure BH edges. Sure that's not world class sub-50 or sub-40 by any means, but at least BH edges has the potential for sub-90 second solves in official competition. And that's using supercube pure BH for edges. Imagine what Half Slice-Plane optimized BH edges might be like. I am excited at the idea!

It would be useful if you told everyone how fast your memorization is too. Keep in mind that the sub-40 people all memorize in sub-15; execution time is all we should be measuring here to see how fast you are. Also, I've watched you do some solves about that fast, and you have some very noticeable pauses. Your method can go significantly faster than that, I suspect.

 

[deadalnix]

And I can even use this on big cubes. (Deadalnix says this works on 3x3x3's, but I use it for big cubes too, so I don't see the benefit of his much longer algorithm.)

The longer alg doesn't mess up centers. The short one does. It's not a problem if you solve center before edges, but if you do edges before, you to take care of this.

 

[Mike Hughey]

And I can even use this on big cubes. (Deadalnix says this works on 3x3x3's, but I use it for big cubes too, so I don't see the benefit of his much longer algorithm.)

The longer alg doesn't mess up centers. The short one does. It's not a problem if you solve center before edges, but if you do edges before, you to take care of this.

I really need to sticker a supercube to use for these sorts of things. I had tried the short one with X centers messed up, and noticed it didn't seem to move them. But I see now that it messes up + centers.

Anyway, I always solve centers first, so as you mention, I don't have to worry about it.

 

[cmhardw]

I really need to sticker a supercube to use for these sorts of things.

Yes you should! Not only are they fun to solve, but it is very nice to be able to "look up" an alg's effect on the cube by simply executing it onto a solved supercube and seeing what happens. Much easier than tracing the cycles out on paper to see. Every English major has a dictionary at their desk, and I think so too should every big cube BLDer have a supercube handy

Anyway, I always solve centers first, so as you mention, I don't have to worry about it.

Lucky! I also solve all of the 3x3x3 portion (with the exception of parity) first so that I can memorize them using my auditory memory method. This means, though, that I must solve all 3x3x3 pieces with supercube safe algs.

It would be useful if you told everyone how fast your memorization is too. Keep in mind that the sub-40 people all memorize in sub-15; execution time is all we should be measuring here to see how fast you are.

To be honest I don't know what my memo time was on that solve. I feel like, for me, a very fast memo is around 45 seconds. So I figure close to a 40 second solve for a 1:25? In a youtube video of me solving at Hadley's house before one of the Georgia competitions I had a solve of 1:21.88. I memo from 0:12 to 0:58 or 46 seconds and I am done solving by 1:34. So that would be about a 36 second solving phase. My solve at Worlds 2009 was probably close to this breakdown seeing as how the time was very close to the one in the video.

Chris

 

[Mike Hughey]

I really need to sticker a supercube to use for these sorts of things.

Yes you should! Not only are they fun to solve, but it is very nice to be able to "look up" an alg's effect on the cube by simply executing it onto a solved supercube and seeing what happens. Much easier than tracing the cycles out on paper to see. Every English major has a dictionary at their desk, and I think so too should every big cube BLDer have a supercube handy

I was going to say I would promise to order some supercube stickers with my next Cubesmith purchase, but I can't find them on his website now. I know he used to have them - did he stop carrying them? I guess I can get out some markers and just mark one up.

Anyway, I always solve centers first, so as you mention, I don't have to worry about it.

Lucky! I also solve all of the 3x3x3 portion (with the exception of parity) first so that I can memorize them using my auditory memory method. This means, though, that I must solve all 3x3x3 pieces with supercube safe algs.

That makes some sense, but I don't think I could get away with auditory for that much information. I suspect I have to stick with image memorization for the central edges, so it doesn't really hurt me at all to do them at the end. I do corners at the beginning, though.

 

[Swordsman Kirby]

Eh, currently I do freestyle with UFR and DF start. In my opinion, starting at DF has some very nice cases, but at the same time, it has AWFUL cases. Unfortunately, I'm too lazy to switch to UF.

"very long term" obviously freestyle/BH is better than M2. M2 is pretty bad in a "not-so-long term"

 

[Mike Hughey]

To be honest I don't know what my memo time was on that solve. I feel like, for me, a very fast memo is around 45 seconds. So I figure close to a 40 second solve for a 1:25? In a youtube video of me solving at Hadley's house before one of the Georgia competitions I had a solve of 1:21.88. I memo from 0:12 to 0:58 or 46 seconds and I am done solving by 1:34. So that would be about a 36 second solving phase. My solve at Worlds 2009 was probably close to this breakdown seeing as how the time was very close to the one in the video.

Wow, I was even more correct than I thought. That's actually a pretty fast execution - do that with a Ville memorization phase and you're easily sub-minute! Which confirms my suspicion that pure BH can get you in the top 10 in the world at 3x3x3 BLD. (For now. But by the end of this year, there will probably be 20 to 30 people with sub-minute times.)

And it also means I'm memorizing 3x3x3's faster than you these days, using a memory method that's essentially the same as yours. That frightens me.

 

[cmhardw]

Wow, I was even more correct than I thought. That's actually a pretty fast execution - do that with a Ville memorization phase and you're easily sub-minute! Which confirms my suspicion that pure BH can get you in the top 10 in the world at 3x3x3 BLD.

Although I know I am biased towards the method, for obvious reasons, it makes me feel good to hear someone else say that. I truly believe this too, I just don't think I will be the person to demonstrate this fact I do hope one day that someone using BH, or Half Slice-Plane optimized BH, will get a superfast time just to know that the method really can be as efficient and good as Daniel and I believe it can.

And it also means I'm memorizing 3x3x3's faster than you these days, using a memory method that's essentially the same as yours. That frightens me.

Mike I think you underestimate yourself! Do you remember me telling you, back when you first started, that I was already prepared for the day you would surpass me! Well, I think that day is fast approaching, if it has not already happened! I will certainly try my best to catch up to your memo speed, but no guarantees!

Also, Daniel passed me a LONG time ago in memorization speed. I can still beat him on execution, but only because I practice the algs a bit more often than he does. He can certainly still out-memorize me no question.

Chris

 

[yoruichi]

3 cycle will win

 

[deadalnix]

Certaily not xD

BH is incredibly fast for corners, and 3OP is really slow on edges.

 

[amostay2004]

What yoruichi means about 3 cycle is not 3OP =p He cycles stickers without preorientation eg freestyle

 

[Sakarie]

I'm not familiar with DIADEM more than to the basic idea, but I think I'm using a variant. Right now I'm using what I think is speed-optimized commutators for every case where both pieces are in the E-slice. Also, if one of the pieces is UL or UR, I start with a U/U'-move, if I can't cancel the move when solving the other piece.

Shortly, I'm going more and more freestyle, since that's simply much faster in many cases. The question is how many of the cases are faster freestyle, and ow many are faster with a "braindead" m2? (With "braindead" I mean a method that is "Identify piece, apply alg" over and over again. This isn't something negative timewise, since it's faster if you dont't have to think.)

Another thing you could take in mind is that if you later want to advance to an even more advanced method, like speedoptimized commutators, or use non-commutators in some cases, then you'd probably prefer UF or UR as buffer, and not DF.