• Welcome to the Speedsolving.com, home of the web's largest puzzle community!
    You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.

    Registration is fast, simple and absolutely free so please, join our community of 40,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

Thoughts about the BH method

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
this is still in its baby phase
That's exactly the problem. Maybe you simply posted it too early. You've claimed that full 5-style is viable and that you have a "learning mechanism" that will enable you to master it, but you have not told us what that mechanism is or how you have validated it. You just keep repeating that learning full 5-style is possible, and expecting us to take that at face value. Well, my belief today is that it's not humanly possible to learn 200,000 algs in 4 years, and only a very good explanation will change my belief. The problem is one of quantity, but you are not addressing that. Instead, you are making vague qualitative statements about comm types and setups, etc. Unless you can provide a quantitative explanation of how you have broken down 5-style into some manageable system and amount of work, your claim lacks credibility.

(Besides which, there's the important question of whether it would be worth the effort and faster than 3-style. Even if you could learn it all, you would have worse algs with worse recall and less practice.)
 
Last edited:

kubnintadni

Member
Joined
Jan 4, 2018
Messages
71
Would it be possible to reduce all Corner 5-cycles (or edge 5-cycles, for that matter) down to various groups of cases which are essentially the same, just flipped, mirrored, inverted, conjugated, etc., and intuitively reduce to and then solve those cases instead of memorizing the algs? If the number of such groups of essentially identical cases was less than or equal to the number of cases in ZBLL, then it would absolutely be something which should be tried. (I'm not the one to ask whether or not it would be better, but no doubt it would deserve to be seriously tried by someone at the very least)

Like intuitive 3-style but with 5-cycles?

Also, are there any resources on floating buffers? I saw it mentioned on this thread as a more realistic alternative to 5-style. How many moves are saved by using floating buffers?
 

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
Would it be possible to reduce all Corner 5-cycles (or edge 5-cycles, for that matter) down to various groups of cases which are essentially the same, just flipped, mirrored, inverted, conjugated, etc., and intuitively reduce to and then solve those cases instead of memorizing the algs?
Yeah it's possible, but it remains a problem of quantity. How many cases are there? I don't know, but if someone wishes to claim it's a viable approach to learning full 5-style, that's where they should start :).
 

PapaSmurf

Member
Joined
Jan 4, 2018
Messages
1,103
WCA
2016TUDO02
YouTube
Visit Channel
2 ideas to help reduce the number of cases: do 4-style, and the M slice is off by an M2 and the corners have an edge swap in there somewhere, or do 5-style eka. So set up to UF or LDF and do one of a massively reduced set of algs.
 

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
Some other nice ideas include finding ways to work 4-6 move 5-cycles into 3-style solves.

For example M' U M U' can be set up or mirrored to solve the DF-FU-UB case on the M slice followed by any inner and outer edge on opposite sides. It's short and fast and setting up the other two pieces on opposite faces makes it quite easy to use.
Plus it could be useful in just under a quarter of occurrences of DF-FU-UB-xx-yy with no cycle breaks, so occurs frequently enough to get practised.

Or, you can think ahead into 5-cycles to solve pieces out of order, using shorter, faster algs for different buffers. Someone else posted on FB about this recently. If the next 4 targets are ABCD the usual 3-style way is to solve AB then CD, but you could solve it other ways like BCD then AB where BCD is a fast 4 or 5 mover with B as buffer.

I'm not fast enough to seriously consider using this stuff myself, but I'm sure the top blinders already dabble with it. Of course it's a long way short of the full 5-style that @abunickabhi is advocating, but sometimes the best place to start is with whatever low hanging fruit offers the better return on time invested.
 

jesinspirment

Member
Joined
Aug 6, 2013
Messages
5
WCA
2010BENS01
Hi all,

I have previously used Boomerang/M2 for blindfold solves, and now I am learning BH, starting with corners.

What I'm currently doing, is get all the 378 corner cycles from here: http://www.speedcubing.com/chris/bhcorners.html

From there, I simply apply the reverse algorithm to all cases in order to set up the respective cycles, but I figure out a commutator myself for all cases, as I feel that it helps in my understanding of the cube and that it will be easier to "remember" since I came out with the list. For every commutator I figure out, the reverse of that becomes a solution for another cycle.

However, the issue here is I have too many cube rotations in my commutators, which I'm not sure if it's ok. Here is an example:

Screenshot 2018-12-14 at 14.24.41.png

Are such cube rotations ok? Or should I just get used to doing insertions from other angles? Will appreciate help from all of you who are more experienced.
 

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
Are such cube rotations ok? Or should I just get used to doing insertions from other angles? Will appreciate help from all of you who are more experienced.
If you want to get fast, rotations are best minimised. The top solvers today generally favour additional setup moves so that they can avoid rotations, avoid regrips, and keep in the <R,U,D> moveset as far as possible. It's not unusual for 3style algs to use 3 or 4 setup moves to an 8 move comm, for the sake of ergonomics, even if it could be solved in fewer moves with worse ergonomics.

I'd recommend checking out some modern alg lists to get some ideas on what works well. Graham's list is good: https://docs.google.com/spreadsheet...em3XJ5tBm7Dk1dTRcZ7KcXYbGP4/edit?usp=drivesdk
 

jesinspirment

Member
Joined
Aug 6, 2013
Messages
5
WCA
2010BENS01
If you want to get fast, rotations are best minimised. The top solvers today generally favour additional setup moves so that they can avoid rotations, avoid regrips, and keep in the <R,U,D> moveset as far as possible. It's not unusual for 3style algs to use 3 or 4 setup moves to an 8 move comm, for the sake of ergonomics, even if it could be solved in fewer moves with worse ergonomics.

I'd recommend checking out some modern alg lists to get some ideas on what works well. Graham's list is good: https://docs.google.com/spreadsheet...em3XJ5tBm7Dk1dTRcZ7KcXYbGP4/edit?usp=drivesdk

I re-looked at some of my commutators that have double rotations like x' y etc. I happen to think of an easy way to make it rotationless, but not sure if I'm on the right track. That is, I make a conscious decision to standardise my interchange move to U, U' or U2 (my buffer is UFR). In other words, even for cycles that can be solved in 8 moves, I avoid rotation by intentionally introducing 1 setup move to create a U layer interchange move, apply the commutator, reverse setup.

For instance, supposedly we have this: UFR LFU BDL

The optimal solution I figured out (8 moves) is: x' y' [D', R U' R']

To avoid the rotation, I figured that I can do this by forcing a U layer interchange like this: [L: U2, R' D R]

Another example: UFR URB LUB

Optimal: x' [R' D' R, U']
Rotationless: [L2: R' D R, U] or [R: U, R D R']
 
Last edited:

TheGrayCuber

Member
Joined
Feb 21, 2012
Messages
188
WCA
2012GRAY01
YouTube
Visit Channel
For instance, supposedly we have this: UFR LFU BDL

The optimal solution I figured out (8 moves) is: x' y' [D', R U' R']

To avoid the rotation, I figured that I can do this by forcing a U layer interchange like this: [L: U2, R' D R]

Another example: UFR URB LUB

Optimal: x' [R' D' R, U']
Rotationless: [L2: R' D R, U] or [R: U, R D R']

For the first example, you can do [Lw: U2, R’ D R] instead to cancel an R Move and an Lw, and do a rotation instead.

For the second example the L2 setup seems far worse than the R setup.
 

leeo

Member
Joined
Oct 6, 2014
Messages
75
Location
Spokane, WA
WCA
2016ODEG01
My program is finally built to the point where I can easily verify any list or spreadsheet of BH corners or edges. For example, from the wiki http://www.speedcubing.com/chris/bhcorners.html. I reformatted and rearranged the list into a spreadsheet employing the common Speffz letter system and sorted it here: https://docs.google.com/spreadsheets/d/1NOeLuw74d10lyMy9uWE6mFpPaKzSyBFiS-KXUvwfgn4/edit?usp=sharing
Where a target is on the same moving piece as the buffer or secondary, it is noted in the spreadsheet with the speffz letters attached to the same cubie, i.e. <AER>, <EAER> or <RAER>. I was able to verify the the corners list is 100% correct!

For the edges, from http://www.speedcubing.com/chris/bhedges.html, I also reformatted, but did not trim the extra entries -- however I did add a handful of missing entries. This was evident from following the letter system. I also reformatted the result and arranged into a spreadsheet employing the common Speffz letter system and sorted it here: https://drive.google.com/open?id=18s6vbyALkVc3VMKiidhz8wpjFr4b5z7z
Where a target is on the same moving piece as the buffer or secondary, it is noted in the spreadsheet with the speffz letters attached to the same cubie, i.e. <AQ> or <QA>.

For a link to the source code at this point, see my speedsolving.com post http://www.speedsolving.com/forum/threads/creeping-featurism-in-my-program.71692/ or the bit-bucket git push at https://bitbucket.org/leeo3264/cubetrae/src/master/
 
Last edited:

abunickabhi

Member
Joined
Jan 9, 2014
Messages
6,687
Location
Yo
WCA
2013GHOD01
YouTube
Visit Channel
My program is finally built to the point where I can easily verify any list or spreadsheet of BH corners or edges. For example, from the wiki http://www.speedcubing.com/chris/bhcorners.html. I reformatted and rearranged the list into a spreadsheet employing the common Speffz letter system and sorted it here: https://docs.google.com/spreadsheets/d/1NOeLuw74d10lyMy9uWE6mFpPaKzSyBFiS-KXUvwfgn4/edit?usp=sharing
Where a target is on the same moving piece as the buffer or secondary, it is noted in the spreadsheet with the speffz letters attached to the same cubie, i.e. <AER>, <EAER> or <RAER>. I was able to verify the the corners list is 100% correct!

For the edges, from http://www.speedcubing.com/chris/bhedges.html, I also reformatted, but did not trim the extra entries -- however I did add a handful of missing entries. This was evident from following the letter system. I also reformatted the result and arranged into a spreadsheet employing the common Speffz letter system and sorted it here: https://drive.google.com/open?id=18s6vbyALkVc3VMKiidhz8wpjFr4b5z7z
Where a target is on the same moving piece as the buffer or secondary, it is noted in the spreadsheet with the speffz letters attached to the same cubie, i.e. <AQ> or <QA>.

For a link to the source code at this point, see my speedsolving.com post http://www.speedsolving.com/forum/threads/creeping-featurism-in-my-program.71692/ or the bit-bucket git push at https://bitbucket.org/leeo3264/cubetrae/src/master/

What is going to be the primary use of this application. Will it be for verifying algsheets of different people with different lettering schemes?
 

abunickabhi

Member
Joined
Jan 9, 2014
Messages
6,687
Location
Yo
WCA
2013GHOD01
YouTube
Visit Channel
maybe a reduced 5 style, with a very efficent combination of way less algs would be interesting? maybe?

Exactly my point,
and to accelerate the learning rate, I have developed the Yo notation to memorize the algorithms.
https://docs.google.com/document/d/1bfDsydw6pxBftd8Xwik95FNjILGkdrJMBO5EORbftII/edit?usp=sharing

I will be also documenting my progress on the fingertricks that I get for each case like in this video:
Since I am just starting out drilling, not all algs are sub-1ed by me, but I think they can be sub 1.5ed since most of the algs are no more than 14 moves.

Right now, I am 8000 algs down, and I need to memorize ~118000 algs more.
https://github.com/abunickabhi/5-style/blob/master/5-style-edge.pdf

With the Giiker cube coming out last year, I can drill even more algorithms faster and randomly, and save time.

So, by my goal setting, by the end of 2020, I should have sub-1.5ed every 5-style edge alg and open sourced the fingertricks on my YT channel yo.

Mark Rivers, I hope these arguments have made you a bit less skeptic about this method.
 
Last edited:

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
I need to memorize ~118000 algs more
...
So, by my goal setting, by the end of 2020, I should have sub-1.5ed every 5-style edge alg
...
Mark Rivers, I hope these arguments have made you a bit less skeptic about this method.
If you're claiming you're going to learn 118,000 new algs by the end of next year, AND drill them all to sub-1.5, then no, I'm still sceptical. Sorry.
 

PapaSmurf

Member
Joined
Jan 4, 2018
Messages
1,103
WCA
2016TUDO02
YouTube
Visit Channel
In his defence, he did say that he'll have executed them sub 1.5, not learnt. I know of a dude who has sub 2ed every 1lll alg. He hasn't learnt them. But I am skeptical that without an intuitive way to do 5 cycles this method will be nigh on impossible.
 

mark49152

Premium Member
Joined
Oct 29, 2012
Messages
4,719
Location
UK
WCA
2015RIVE05
YouTube
Visit Channel
In his defence, he did say that he'll have executed them sub 1.5, not learnt. I know of a dude who has sub 2ed every 1lll alg. He hasn't learnt them. But I am skeptical that without an intuitive way to do 5 cycles this method will be nigh on impossible.
He said "I need to memorize ~118000 algs more". Perhaps he did mean he would drill them by end of 2020 then move on to memorizing them afterward, but either way, there are 660 days until end of 2020. That's 179 algs/day, every day.
 
U

Underwatercuber

Guest
He said "I need to memorize ~118000 algs more". Perhaps he did mean he would drill them by end of 2020 then move on to memorizing them afterward, but either way, there are 660 days until end of 2020. That's 179 algs/day, every day.
bUt RoMaN rOoMs

Also I doubt they could be 1.5, I looked at his 5style sheet a while back and there were several 5+gen algs that were super bad lol
 
Last edited by a moderator:
U

Underwatercuber

Guest
also would it be at all possible to "lock" this thread and replace it with a 3style thread? I would hate to see people actually picking up BH over 3style because they think its relevant, most of the time when people ask about BH the first thing done is to just tell them about 3style :p maybe someone can just link a 3style thread as the final comment before closing this thread to help people out too
 

abunickabhi

Member
Joined
Jan 9, 2014
Messages
6,687
Location
Yo
WCA
2013GHOD01
YouTube
Visit Channel
bUt RoMaN rOoMs

Also, I doubt they could be 1.5, I looked at his 5style sheet a while back and there were several 5+gen algs that were super bad lol


I have replaced those algs, now all the algs are rotation-less and without B moves.
(It is tough to make 5 cycle edge algs 2gen or 3-gen with Slice moves. 4-gen + slice moves and optimal move count is the sweet spot that I have found. :)

https://github.com/abunickabhi/5-style (Check out the newer algs here, they all are good, except for the corner algs.)

There is almost two regrips in each alg, and it can be reduced to one maybe after investing more time into fingertricking it.

And yes, I have prepped 75 new roman rooms just for memorizing the algorithms.
I will be storing the Yo notation images of each algorithm at a locus in the mind palace.
https://www.memorypalace.com/palace/2129/mbld

Fingers crossed!
 
Last edited:

dbeyer

Member
Joined
Sep 12, 2006
Messages
396
Location
Harrington, Delaware
Eh, well the point this entire method was to implement Commutators across the entire spectrum of 3x3x3 to 5x5x5. The point of BH is that there are 8 move Commutators in ABA'B' format. Then you expand upon that to recognize cancellations for A9s. Followed by cycles that are 10 move optimal. There is a special Cyclic shift 10 mover. Then your column cases for 11 moves. Finally 12 move cases.

Due to the symmetry of the cases, you can adjust your moves to find your intuitive finger friendly cases.

It was never meant to be a comprehensive list of algorithms. Moreso an overview of an understanding of Commutators to solve the cube.

Low move counts and efficiency lends to speed. But that same concept was meant to expand to x-centers, t-centers, wings, edges, corners, and even further big cube expansions.

It is not about memorizing thousands of cases. Rather you efficiently and accurately solve a case based on the relative positional relationships between the cubies you are solving sequentially.

That time could be better spent developing finger tricks and understanding of cancellations between cycles. Also actually practicing blind solves and sighted solutions with cycles.

Anyway just my take on it.
 

tx789

Member
Joined
Sep 6, 2010
Messages
2,009
Location
New Zealand
WCA
2010HUNT02
YouTube
Visit Channel
Eh, well the point this entire method was to implement Commutators across the entire spectrum of 3x3x3 to 5x5x5. The point of BH is that there are 8 move Commutators in ABA'B' format. Then you expand upon that to recognize cancellations for A9s. Followed by cycles that are 10 move optimal. There is a special Cyclic shift 10 mover. Then your column cases for 11 moves. Finally 12 move cases.
.

Speed optimal comm are used today. [R F' R' U':[R D R',U2]] is an example of a 4 move set up and [U R' F R,D' R D R'] is a strange comm I don't understand. You can find list of these speed optimal comms in spreadsheets bestsiteever.ru has a list of links to them. Those are from Daniel Lin's list. Learning from these lists is fine if you understand comms for all of 3-style. Doing it completely intuitively leads to slow comms.

I feel like move optimal was focused on back when this was first published. Not sure people discussed buffers much either back then UFR/UF are considered best now.
 
Top