Big Cube BLD Discussion

[Mike Hughey]

I tried a 5x5x5 sighted with my BLD method today: 7:38.08.

For 5^3: 7:38.77, about 1/3 of my average 5BLD time.

Wow, that's close!

 

[aronpm]

So to bring up a different subject - how fast are you guys at sighted solving using your BLD method? I'm amazed by how long it takes me. My best time for 4x4x4 so far is about 5:30, which is not much faster than my best time BLD. I tried a 5x5x5 sighted with my BLD method today, and saw a bigger difference: 7:38.08. So that's about twice as fast as my typical BLD solve. (I've long believed that 5x5x5 BLD should be not much slower than 4x4x4 BLD if you're perfect at it, so somehow this does not surprise me.) How does everyone else do with this?

Just trying 1 solve:
4bld: 3:09.66. This is probably what I average for execution BLD.
5bld: 5:33.90. I'm not sure about my 5bld splits because they vary too much but this is normal I guess. (omg sub3 corners+centers)

 

[Zane_C]

Just did a sighted 5BLD.
6:52.17, in a solve it is 9+ minutes.

 

[cmhardw]

I know this changes topic a bit, but I am continuing a post in the 45th weekly competition thread here so that I will not continue to thread hijack that thread.

4x4x4BLD splits:

Solve 1)
Close DNF (off by a 3 cycle). At the end I cleared two double swaps using my original buffer, when it was actually a floating buffer for the first cycle. The pacing felt very natural to the same pace I used on my weekly scrambles.

Memo 3:58.65
Solve corners and centers (no corner parity): 1:16.39
solve wings: 1:16.39 (no joke!)
Total solving time: 2:32.78
Total time: 6:31.43

Memo:solving ratio ---> 1.56:1

Solve 2)
Successful solve

orient cube: 2.08
memo centers: 1:38.78
memo wings: 1:25.74
memo corners: 8.22

solve corners (leaving parity): 17.56
solve centers: 1:04.95
fix corner parity: 9.11
solve wings: 1:04.86

Total memo time: 3:14.82
Total solving time: 2:36.48

Total overall time: 5:51.30

Memo: Solving ---> 1.25:1

Solve 3)
DNF off by 3 wings
Wing memo: (M LF OW)(N GP HU AC IK XB SR DJ)(XT)
[(DNF)]' = (V TX)
[X] = (M LF OW)(N GP HU AC IK TB SR DJ)(VX) = Which means I executed what I memorized, leaving (V TX) to solve.

Shot to x twice. Last cycle should have been (VT)

orient: 6.88
memo centers: 53.64
memo wings: 2:24.11
memo corners: 8.77

solve corners: 21.05
solve centers: 51.51
solve wings: 1:13.50

total memo time: 3:33.40
total solving time: 2:26.06

Total solve time: 5:59.46

Memo : Solving ---> 1.46 : 1

Solve 4
Successful

orient: 2.21
memo centers: 1:09.04
memo wings: 2:10.64
memo corners: 12.30

solve corners: 26.59
solve centers: 1:06.53
solve wings: 1:26.64

total memo time: 3:34.19
total solving time: 2:59.76
total time: 6:33.95

memo : solving ---> 1.19 : 1
---------------------------------------------------------------------

5x5x5 BLD splits:

Solve 1)
Very close DNF, off by 3 edges.

wing memo: (Q CK ON UA LT GJ)(E XH B)(M IW FR SD)
[(DNF)]' = (B HX)
[X] that I actually executed for wings was: (Q CK ON UA LT GJ)(E HX B)(M IW FR SD)

orient cube: 5.12 (no re-orient)
memo t-centers: 1:29.88
memo x-centers: 1:29.08
memo wings: 2:34.09
memo corners (images): 1:28.60
memo centrals (words): 1:11.03

solve centrals setting up parity: 1:11.59
solve corners setting up parity: 33.31
solve t-centers: 1:05.97
solve x-centers: 1:01.76
fix corner/centrals parity: 11.83
solve wings: 2:30.23

Total memo time: 8:17.80
Total solving time: 6:34.69
Total time: 14:52.49

Memo:solving ---> 1.26:1

Solve 2)

Here is an example of a solve with recall delays, and a fairly bad DNF

Centrals memo: (A PO BG EW TS SX CS)
Mis-memorized and shot to S three times!!! This would be one too many, as I broke into a cycle at location S, but still
What I think the memorization should have been (did not save the scramble): (A PO BG EW TS RX CR)
centrals [(DNF)]' = (AS)(CR)
[X] that I actually executed based on "correct" memo = (A PO BG EW TR XR)

After realizing that I shot to S too many times, in my frenzy to figure out what happened apparently I forgot to shoot to S at all lol

Wings memo: (Q HP DW GM)(E RU JV NB)(F XT)(IS)(A CK O)
wings [(DNF)]' = (F AO)
wing [X] = (Q HP DW GM)(E RU JV NB)(F XT O)(IS)(A CK)

For the double swap to solve the two odd cycles I should have done (IS)(AO) but apparently I did (IS)(FO)

orient: 8.44
memo t-centers: 1:14.98
memo x-centers: 1:06.27
memo wings: 2:38.84
memo corners (images): 45.91
memo centrals (words): 52.37
memo center cycle: 3.24

solve center cycle and orient permuted but disoriented pieces: 17.48
solve centrals: 2:17.20
solve corners: 38.51
solve t-centers: 2:55.30
solve x-centers: 1:10.03
solve wings: 1:40.52

Total memo time: 6:50.05
Total solving time: 8:59.04
Total time: 15:49.09

Memo:solving ---> 1 : 1.31

Solve 3:
Successful solve

orient: 4.48 (no re-orientation)
memo t-centers: 1:56.50
memo x-centers: 1:12.68
memo wings: 2:06.81
memo corners (images): 28.51
memo centrals (words): 1:03.56

solve centrals: 45.25
solve corners and setup for parity: 35.48
little more parity setup: 20.94
solve t-centers: 1:33.35
solve x-centers: 1:03.64
fix corner/central parity: 8.47
solve wings: 2:14.83

Total memo time: 6:52.54
Total solving time: 6:41.96
Total time: 13:34.50

Memo : Solving ---> 1.03 : 1
---------------------------------------------------------------------

I'll try to add more as I do more solves over the next day or two.

Chris

 

[Mike Hughey]

See, Chris? Your solving times are very similar to mine, except for wings, where you completely blow me away! I always knew my wings were bad, but I had no idea they were that bad. I've learned the first 8 letter pairs (I know it's different from you, but I still find it easiest to learn BH alphabetically), and I switched to using what you call "drop and slide" for several of them - it's a nice alternate approach! 244 more to go!

 

[cmhardw]

See, Chris? Your solving times are very similar to mine, except for wings, where you completely blow me away! I always knew my wings were bad, but I had no idea they were that bad. I've learned the first 8 letter pairs (I know it's different from you, but I still find it easiest to learn BH alphabetically), and I switched to using what you call "drop and slide" for several of them - it's a nice alternate approach! 244 more to go!

Hey Mike,

Yeah I do agree now that my splits seem to be closer to what you've been describing. That's so strange that it feels like a 2:1 split on most every solve, even when it's obviously not. I will still try to view my solves as having this type of split, because it has led to slightly higher accuracy, and my pb single! I'll just know that it's not the actual split.

Also, I do recommend to learn BH for the wings. I know that general opinion is that "BH is too slow, just learn freestyle." I posted about this in the weekly competition #45 thread, but if only either Daniel or I were faster at memorizing then I think BH would be a more credible method. As it stands now, since both of us memorize at not-world-class speeds, I think people attribute our slower overall time to mean that BH is slower. To be fair to Daniel his memorization is far faster than mine. I remember him memorizing the 5x5x5 sub-6 on a somewhat consistent basis on one of our meetings where he came to NC. If he hadn't de-facto retired, he probably would have quickly surpassed my times from that point forward.

I'm hoping that with my splits, if I can get enough good times, then I can prove that idea wrong.

By the way, I posted a bad DNF solve time break down edited into my first post. I mis-executed the last double swap during the wings, but this means that I still did the same number of turns that a regular solve would contain. My solve time for wings on that solve was 1:40.52. That wing solve felt really good, so I'm thinking that at my fastest I am sub-2, and probably average between 2:00-2:30 for wing execution, using move optimal BH of course.

I am obviously very biased, but yes I think using move optimal BH would help speed up your wing times quite a bit! I know you are already learning it, but the more you learn the better I would say

Chris

 

[riffz]

See, Chris? Your solving times are very similar to mine, except for wings, where you completely blow me away! I always knew my wings were bad, but I had no idea they were that bad. I've learned the first 8 letter pairs (I know it's different from you, but I still find it easiest to learn BH alphabetically), and I switched to using what you call "drop and slide" for several of them - it's a nice alternate approach! 244 more to go!

Were you using r2 previously? And by BH wings do you mean to say that you are using DFr as your only buffer?

I recently relettered my wing edges for 4x4 as I plan to start practicing 4BLD. I chose the scheme in such a way that almost every alg is the same as my 3x3 algs, except for replacing M moves with l or r. So other than cases involving adjacent wings and ones that I use <M,U> and <R,U> algs for, I basically know all the 3-cycles I'll need for 4x4 without really practicing them.

 

[aronpm]

I've learned the first 8 letter pairs (I know it's different from you, but I still find it easiest to learn BH alphabetically)

I learned corners alphabetically, learning each case separately, using Mnemosyne. I plan to do the same with edges and wings.

I am obviously very biased, but yes I think using move optimal BH would help speed up your wing times quite a bit! I know you are already learning it, but the more you learn the better I would say

I think optimal BH for edges/wings would speed up solving quite a lot! I started making lists of edge/wing commutators and I'm amazed at how much quicker it is than M2/r2. I definitely have to switch soon.

Going a bit off of this topic, Chris, are all of your algs optimal? Personally I think using some speed-optimal algs is better for some cases, but often the optimal alg is very fast.

 

[cmhardw]

Going a bit off of this topic, Chris, are all of your algs optimal? Personally I think using some speed-optimal algs is better for some cases, but often the optimal alg is very fast.

Hi Aron,

I'm not brazen enough to say that every single one of my algs are move optimal. After (was it Mats?) showed us the 4x4x4 x-center alg 3U r' 3U' l 3U r 3U' l' I knew that there were possibly some very strange, and very amazing, shorter ways to do certain cases! I do use that style alg in my solving by the way, and I've come to start calling them "Pseudo-slice" algs, or commutators that use "Pseudo-slicing". I can understand the logic behind it, and it uses the concept of building a pseudo slice and then turning that pseudo slice one turn to perform the first part of the cycle. All very neat stuff!

I would say that, based on the work Daniel and I have done over the years to slowly but surely refine cases 1 by 1 to be shorter and more move optimal, that I do at least 90% optimal algs when considering all piece types. I'd like to say even 95%, but since we haven't searched any cases via computer for optimality then I don't know if I haven't yet discovered a new concept similar to pseudo-slicing that is more optimal on some case types.

I'll give you one perfect example of where I do purposefully choose a sub-optimal alg that is faster. For BH corners, no B9 alg is optimal. Every corner B9 case is actually optimally 8 turns with a standard commutator. Take the case:

UBL->UBR->BLD

This is move optimal at 8 turns with something like: y' [L D L', U'] y
However, the B9 alg of: U2 R' D2 R U' R' D2 R U' is much faster for me, and I will use this on a real solve.

Sometimes for the wing case: UBl -> DLb -> FLu I will do the 10 move: b2 [U, R' u' R] b2 rather than the optimal:
y' L' u2 L' D' L u2 L' D L2 y

The 10 mover just has a better flow to me, even though I know that the 9 mover is optimal. To be honest it irks me that for UBl -> DLb -> FLu that I don't think of the 9 mover first on a solve, but instead the 10 mover, even though I know it's an A9 case.

So yes, not all of my algs are optimal by choice during a solve. I would say that I don't often substitute an alg for a faster sub-optimal length one though. Also, I don't know which new cycling and slicing concepts of which I don't know haha. So I may be using a sub-optimal alg for certain cases and not even know it.

Chris

P.S.

Ubl -> Bdl -> Ldb
3U r' 3U' l 3U r 3U' l'
This amazing alg works by the following logic, by the way:

It's really D b' D' l D b D' l'
The idea is that we could move Ldb to Ubl via the turn b', but since we also will need to then move Bdl to Ldb (which is now at Ubl after the b' turn) via the move l then we need to do some clever pseudo slicing. If you just do b' l b l' then you will also effect pieces on D, F, and R. Herein lies our problem.

First we imagine that the b layer is a pseudo-slice. The pieces Ldb, Lub, ULb, Ubl, Ubr, URb, Rub, Rdb all form the backbone of our "pseudo-slice." Now we want to turn the layer b' such as to send Ldb to Ubl. However this messes up the l slice at Ubl and Dbl. We want the effect at Ubl, but not at Dbl. To avoid this, we will pseudo slice the b face. The remaining portion of the b face pseudo slice are the pieces DFr, Dfr, Dbr, DBr. We will first do the turn D so that we align this 4x1 block with our "C" shaped backbone already in place on the b layer. This creates the b face pseudo slice, which we will turn with b'. This b' sends Ldb to Ubl, which is part of our cycle. The only effect on the original l layer is to change the Ubl location. The l layer pieces that were at DFl, Dfl, Dbl, DBl are at DRf, Drf, Dlf, DLf respectively after the D turn. Now we break up our pseudo slice on the b layer, and restore the original l layer with D'.

Next we bring Bdl to Ubl with the turn l (remember that the b slice is offset by a turn, so that what is now at Ubl will end up in Ldb after we undo the b face turn). After the l turn we need to do the turn b to bring Ubl back down to Ldb, but this will destroy the Dbl location on the l slice. To fix this we now reform the pseudo-slice b layer with D again. This also moves the 4x1 portion of the inner l slice from DFl, Dfl, Dbl, DBl to DRf, Drf, Dlf, DLf respectively. Now undo the pseudo-slice b layer turn with b. Reform the correct l slice with the turn D', and lasty restore the l slice with l'.

It maybe sounds worse in description than it is, but this alg is absolutely beautiful in its execution. I personally execute it with the triple layer turns, but I had to write it in regular outer layer turns so that I could come to an understanding of how it works.

 

[Mike Hughey]

Were you using r2 previously? And by BH wings do you mean to say that you are using DFr as your only buffer?

Yes, I was using r2, except that for the bad middle slice edges, I was already using commutators (although they were not necessarily optimal, so I'm still relearning a lot of those). And yes, I'm doing BH wing, using DFr as my only buffer - at least to start with.

I recently relettered my wing edges for 4x4 as I plan to start practicing 4BLD. I chose the scheme in such a way that almost every alg is the same as my 3x3 algs, except for replacing M moves with l or r. So other than cases involving adjacent wings and ones that I use <M,U> and <R,U> algs for, I basically know all the 3-cycles I'll need for 4x4 without really practicing them.

I know I should do this as well, but I haven't gotten around to doing it. Currently, my scheme is very different and I think it will be quite confusing and difficult for me to change. I suppose I could reletter the 4x4x4 now as I learn these, but it seems like it will be so painful to do it.

 

[cmhardw]

I know I should do this as well, but I haven't gotten around to doing it. Currently, my scheme is very different and I think it will be quite confusing and difficult for me to change. I suppose I could reletter the 4x4x4 now as I learn these, but it seems like it will be so painful to do it.

Mike, I would say to just do the re-lettering. If you don't, you may end up making memory errors occasionally where you interpret a piece in the scheme that you would like to have, not in your actual scheme. This is the problem I was facing with my R and inner r layers.

Also, riffz and Mike, Daniel Beyer uses a wing lettering scheme that matches his 3x3x3 edges like you're talking about. For example the UF edge piece for me is yellow/green and the yellow sticker is D, the green sticker is E. For Daniel he would do something along the lines of calling UFr D and UFl E, and he would recognize which piece goes where based on whether the yellow/green edge he just came upon is oriented or not if it were a full 3x3x3 edge group.

If you guys are considering it, I would say it would probably be worth it. Don't underestimate the ability of your subconscious mind to worm it's way into your memorization if you're using a scheme that it doesn't like. At least that's my experience anyway. And lettering changes usually only take me about 2-3 weeks before the new scheme feels completely natural, and I start to actually even "forget" the old scheme. For changes I made over a year ago or more ago, like when I re-lettered my B and D faces, I can't even hardly remember the feeling of memorizing with my old scheme. In my old scheme Bul was N, but now Bul is M. The thought of Bul being N to me now just feels wrong, and there is no confusion with the older scheme.

My $0.02

Chris

 

[Sakarie]

Hi Aron,

I'm not brazen enough to say that every single one of my algs are move optimal. After (was it Mats?) showed us the 4x4x4 x-center alg 3U r' 3U' l 3U r 3U' l' I knew that there were possibly some very strange, and very amazing, shorter ways to do certain cases! I do use that style alg in my solving by the way, and I've come to start calling them "Pseudo-slice" algs, or commutators that use "Pseudo-slicing". I can understand the logic behind it, and it uses the concept of building a pseudo slice and then turning that pseudo slice one turn to perform the first part of the cycle. All very neat stuff!

I would say that, based on the work Daniel and I have done over the years to slowly but surely refine cases 1 by 1 to be shorter and more move optimal, that I do at least 90% optimal algs when considering all piece types. I'd like to say even 95%, but since we haven't searched any cases via computer for optimality then I don't know if I haven't yet discovered a new concept similar to pseudo-slicing that is more optimal on some case types.

I suspect this is what you mean: Turbo for centers on 4x4

But I don't think it's necessary to use computers to get any case optimal. If you just put some time for every case, you'll probably see the optimal solution without greater problems.

 

[cmhardw]

I suspect this is what you mean: Turbo for centers on 4x4

Yes that's it! Also, those cases can be done by the same process as what I described for the Ubl -> Rdb -> Bdl case. It's all about forming and unforming a pseudo slice to do each turn.

Ubl -> Rbu -> Ful: l' Uu r Uu' l Uu r' Uu'

Is really doing the following:
1) Form a pseudo-slice out of the b layer by starting with the turn D'
2) send Ful to Ubl with l' affecting only the Ubl part of the real b layer.
3) unform the l face pseudo slice (which reforms the real b layer) with D.
4) bring Rbu to Ubl with the turn b
5) reform the b layer pseudo-slice with D'
6) set-down the piece at Ubl to Ful with the turn l. This will only affect the Ubl portion of the real b layer.
7) unform the pseudo b layer slice (which reforms the real b layer slice) with D
8) set down Ubl to Rbu with b'

All told the alg is: D' l' D b D' l D b'

But I don't think it's necessary to use computers to get any case optimal. If you just put some time for every case, you'll probably see the optimal solution without greater problems.

Here are a couple of cases that Daniel and I have always wondered about:

Ubl -> Lub -> Bul
and UBl -> ULb -> BLu

Are these both optimal at 10 moves, or at fewer than 10 moves?

Chris

 

[Sakarie]

Here are a couple of cases that Daniel and I have always wondered about:

Ubl -> Lub -> Bul
and UBl -> ULb -> BLu

Are these both optimal at 10 moves, or at fewer than 10 moves?

Chris

I can't find anything that could indicate that there are shorter solution than 10 moves. If any-one else finds it, I'll take back the "we don't need computers"...

 

[cmhardw]

I made an observation today about my BAB'A'B mistakes on commutators sometimes. Recall that B is the interchange move and A is the, usually 3 move, insertion.

I have found that I only really make this mistake when the interchange B is a double turn. When the interchange B is a single turn, I can see that B and B' turn the interchange layer/slice in opposite directions. There's very much a to and fro motion that I pick up on intuitively. When the interchange B is a double turn it seems that I somehow, on some subconscious level, forget that the first B turn was done, and then interpret the rest of the commutator as ABA'B' putting in the unnecessary interchange turn at the end.

I have found it difficult to implement Mike's counting technique in general, as I have developed a sort of flow of my algs when I execute them, and it feels strange to count as well. However, since I seem to often DNF BAB'A' algs that use a double turn B interchange, then I will force myself to count on cases where I notice that 1) I interchange first for the commutator, and 2) The interchange turn is a double turn.

It's surprising how effective post mortems are to your solving improvement! Sometimes it takes a while to notice a pattern, like this one. However, when you do notice the pattern it is very much an epiphany "Ah HA!" moment! This type of thing really leads me to believe that witness testimony and memory is less accurate than we would like to believe. I strongly believe that what I do every solve is correct, but the DNF cube that stares me in the face afterward would probably tend to disagree with me.

 

[Zane_C]

I'm just wondering what methods people here use to memorise centers on a 4x4. I solve them first and have been using a combination of audio and visual and I've found it to be slow. Thanks.

 

[Rubiks560]

I'm getting interested in learning 4x4 BLD, but i'm not sure what method to use....whats the easiest method?

 

[Faz]

I'm just wondering what methods people here use to memorise centers on a 4x4. I solve them first and have been using a combination of audio and visual and I've found it to be slow. Thanks.

I use the auditory loop for centers.

I've always thought of memo in layers. For example, for 4x4BLD, I can make images for wings, then use the auditory loop for centers, and then on top of that, use visual for corners.

 

[Zane_C]

Sounds good, I suppose I just need to work on my auditory memory.

 

[riffz]

Centers would be too much for me to memo using auditory loop, but that's maybe just me. I'm thinking when I start practicing 4BLD I'll use auditory for corners, journey for centers and edges.