• 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!

The New Method / Substep / Concept Idea Thread

E

elrog

Member
Joined
Jan 31, 2013
Messages
518
Location
U.S.A.
YouTube
Visit Channel
Please read before posting

I'm making this thread for all of those ideas you have that are interesting, yet are not fully developed. This is a place to post them. I have come up with many ideas and didn't want to post a new thread for every one of them when most don't get very far. Perhaps if an idea gets very far, it may deserve its own thread, but until then, it should go here.

Be open and understanding
Everyone should be open to new ideas, yet also understanding if others don't think it will work.

Post all kinds of ideas
Feel free to post all kinds of ideas here. It does not necessarily have to be oriented around speedsolving. Ideas could range from fewest moves, blindfolding, speedsolving, OH, cube designs and more.

Be clear
Please try to be clear in your explanations of why something is a bad/good idea and use evidence to support your thoughts. Also be clear as to what idea you are referring to.

Avoid cluttering the thread
To avoid unnecessary clutter, you should edit a post to add more information rather than create a new one, unless you want to bump your idea. You should wait at a days since your last post on a certain topic to bump it. To present a new idea, you should present it in the same post as your last idea or bump if you do it on the same day.

Do Your Research
There are a lot of different methods out there. Please try to make sure your idea is new/original before posting. You should check out the methods pages on the wiki.

Here is a list of commonly suggested methods:
Belt - Anything that solves the cube like this (the belt does not always have to be made first / many times EO is solved with it). This is a broad category and there is a large variety of belt methods already out there and there is a good chance you will be repeating something.

F2L blocks - Solving a 2x2x3 block (as in Petrus) by solving 3 cross edges and the two pairs that go between them, solving two 1x2x3 blocks on opposite sides (as in Roux) by placing two cross edges and solving F2L pairs, or any other kind of block, but the already previously listed are the most common.

Cross variants - Solving the cross (or EO line) in a different manner to affect the solve later in some way. Common examples are only solving 3 cross edges, purposely solving the last cross edge as a different edge, and swapping two edges in the cross or EO line.

Last layer variants - There are for subgroups that the last layer is commonly broken down into. These are corner permutation, corner orientation, edge permutation, and edge orientation. Any combination of these in any order has been thought of before. Also, influencing any one of these in some way while placing the last F2L slot has most likely been thought of.

Corners/edges first - Solving either all or most of the corners or edges before solving much (if any) of the other is frequently suggested. In fact most of the early speedcubing methods were corners first.

PCMS - This technically is included in the corners first category, but it is suggested often enough that attention should be brought to it.

Big cube reduction variants - Reducing a 4x4 to a 2x2 has been suggested many times. Influencing edge orientation or the permutation of certain pieces while reducing the cube has been thought of as well.


If there is anything you think I should add to this list, you are welcome to suggest it.
_______________________________________________________________________________________________________________________

To start off, here's a idea I had that ended up similar to the Roux method. Because I liked how easy it was to make corner edge pairs using them M layer, I started making an F2L minus the M layer. I then proceeded to make 2 corner edge pairs in the top layer similar to what is done in Heise. These pairs should not interfere with the M layer and should contain either both BU or FU corners. In other words, you have both the left and right side complete minus 1 corner in each of them. You could then solve the remaining 2 corners and the M layer with an algorithm.

I thought this could better be adapted to speed solving by solving 2 corners while solving the last F2L pair. Then you could solve the UL and UR edges with M and U moves. In Roux, there are variants to solve the M layer last rather than the top 4 edges.

After realizing the similarities with roux, and realizing the ability to AUF the M slice separate from the corners of the top layer, I came up with this: Solve 2 1x2x3 blocks on opposite sides, AUF until you get to a CLL case, Solve the UL and UR edges without misplacing the U layer (you may temporarily move the U layer), solve the top 4 corners and the M layer with an algorithm that is reduced by the ability to AUF the M layer. It should be noted that an experienced solver wouldn't need to AUF to find a CLL case just to have to move the U layer to solve the UL and UR edges.

The number of cases could drastically be reduced by orienting edges while placing the UL and UR edges. You could also reduce it even further by using partial corner control while placing the last F2L slot finishing the second 1x2x3 block. Because CLL cases are recognized by swapping 2 corners, when you AUF to a CLL case and have the UR and UL edges solved, you leave only permutations with an odd number of swaps in the M layer giving you 4 possibilities compared to the 5 possibilities (including solved) that there would be with an even number of swaps. This is not true of CLL cases with corners correctly permuted. In all, I calculate 92 algorithms to solve the top 4 corners and the M layer using partial corner control and having edges preoriented (I did include CPLL but not cases with all corners solved and I did not include mirrors).
 
Last edited:
elrog said:
Does anyone know of a way to tell if a 2x2 solve is 2 gen after 2 corners adjacent corners are solved?

I'm planning on adapting this to a 3x3 method that permutes corners to a 2 gen state during inspection.
Click to expand...

Although it's certainly not the optimal way, I'd just mentally attach 2 corners that are supposed to be adjacent, if none are already attached, and then evaluate the situation. In other words, if you have two adjacent corners that are supposed to be adjacent and are in the correct permutation with respect to each other, and the rest of the corners (which are reduced to 1 side through R and U) are in a solved permutation, then the corners can be solved 2gen. If the adjacent corners are incorrectly permuted with respect to each other and the other 4 corners are diagonally swapped, then the corners can also be solved 2gen. Porkynator had a post in the ZZ/ZB Homethread a while back about using the algs L' U R U' L and L' U R2 U' L to reduce the corners to 2gen.

If this explanation is as unclear as I think it is, PM me with any specific questions. <3
 
@ somerandomkidmike

How do you suggest I do this? Make more divisions to make paragraphs seem shorter or delete some centences that are not really necessary? Maby merge senteces or divide them?

I'm not completely sure about the average movecount for the algs because I have not generated them. I also don't see them being much more useful than any other last layer substep. Heres an Idea I came up with tohugh.

Use the ZZ method but make the EO line with the UF and UR edges in place of the DB and DF edges. Solve the 2x2x3 blocks on both sides as in regular ZZ. Recognise the CLL case and you cen very easily insert the UR and UL edges into place because they are completed with the EO line. You can then solve the cube with a single algs and much less cases than ZBLL. I do think that the steps between and the recognition of the case make other last layer substeps just as good if not better.

@ mDiPalma

It would be much more adequate for speedsolving if you corrected the corner permutation while solving 2 adjacent corners and did so in optimal/near optimal moves. I think it may be hard to look ahead far enough to see where corners will be after you solve 2 of them, but people have somehow done similar things. EG is one example. It is useful to know the 2 corners swapped diagnaly thing though. Thanks.

For a better Idea of my method:

First step: Corner permutation and a 1x1x3 block at the bottom left edge. Only R Rw U Uw M and maby E moves will be used from here on out. Unkown movecount and don't know how to do exactly.

Second step: Add to the first block making a 1x2x3 block by adding the left center and the LF and LB edges. Usually done by matching center with edges on the right side and doing E2 or Uw2. I can usually do this in about 7 moves. I have gotten a 4 some 5s and many 6 moves solves for this step. I've never had this step take more than 9 moves.

Third step: Make a opposite 1x2x3 block similar to the blocks in Roux. Done usually by aligning the DR edge and matching with corner edge pairs. Corner edge pairs can be created easily with the M slice. Putting a pair into a slot can be difficult at first because you must do Rw U Rw' rather than F U F'. This also requires that the pair must be above the slot it goes in before insertion. I can do this in about 17 moves. I've gotten as low as 12 (I've gotten 13 a few times) for this step and never exceeded 18 though I've got it many times.

Fourth step: use M and U moves to solve the FD and BD edges while orienting the U layer edges.This step can take anywhere from 8 to 13 moves. Many times this can be done in around 8 moves, but occasionaly you'll get a case where you have all bad edges, solved centers,and the DF and DB edges are in place, or something like it. This step would be a breeze for any experienced Roux user.

Fifth step: 2GLL to finish the solve. 2GLL can be done in around 13 moves. I got this statistic from the wiki an I'm no sure weather it meant 2-gen or not (I'm assuming it did). Any of the EPLL cases can be done in fewer moves with slice moves, but many prefer to use the 2-gen algs. Despite the higher move count, 2GLL can be done very quickly. Personally, I find it hard to recognize the case quickly before AUFing (I don't actually know 2GLL), but I'm sure it is possible to become quick at recognising it without AUFing first.

Overall, I think this method could be very quick if I could get the first step sorted out and be able to increase my recognition time. It uses more intuition than most methods using blockbuilding and intuitively orienting edges. I find it is easy to do turns quickly with the reduced moveset. When I add the number I got up, I get 45 moves plus whatever the first step takes.
 
Last edited:
I just got to thinking about a post I made a while ago on cyoubx's forum about his new cube. I posted my idea on his thread because it was related to creating a new home-made cube. Now that I made this thread, I'm posting my ideas here. The only got one comment on my idea saying this: "it wouldn't corner cut". I personally don't see why not.

On the other thread, I talked about making a cube in which the center pieces are not seperate from the core. I mentioned the problem of having to have the center of a side circular to make pieces spin aruond it, but not being able to round pieces around it without inhibiting the turning of other sides. I've come up with a solution to this now.

My idea is to have a peice that is a square with a circle cut out of it that fits onto each of the center pieces. You would then have edges between those and corners between edges.The square piece should be able to click on over the core so that it won't fall off, yet can be taken off. The clicking mechanism should also not touch anything once it actually is on. This would create an all plastic cube and it wouldn't need to be tentioned.

Not having to/being able to tension I view as a good thing, but others may like to have it adjustable. I personally just don't like messing with tensions because it is hard to get them equal and they change over time. To get a loose or tight cube, you could get a different sized square piece.

As before, I would use rubber tiles that are inlayed into the pieces and can be taken out simply by pinching them, yet they wouldn't ever fall out or break.

As for the previous comment about not being able to corner cut, you could do it just as well as any other cube. You could round out the square piece that fits around the centers some and also round the edges and corners.

For an anti-pop mechanism, you could make the corners fit into grooves in the core farther in than the square piece. You cuold also use grooves or overlapping parts with the edges, square pieces, and corners just as in many other speedcubes.

One more Idea I had for a 3x3 was to use the design for a New Rubiks 2x2 with edges held in place between the corners. You could then have the center pieces held in by the edges. This design would be possible, but I don't see it ever being a very good speedcube.

If anyone else has any ideas they think are interesting, feel free to post.
 
I see a potential issue, and I believe this is why the other person said it wouldn't cut. The reason why cubes are designed the way they are is that the spring allows for fast turning and corner cutting, but still maintains a stable shape. Your design, from what I can tell because it's hard to know exactly what you're thinking, is either going to be very loose with no stable integrity, or very tight with difficult turning.
 
I was looking for my thread and couln't find it anywhere! then I realized it got stickied. You had me worried there for a second.

I know that the springs allow it to be loose yet stable. My design could be loosened aswell. I do not think it will lose its stability because it still has the pieces interlocking with eachother and the actuall core as well. If you look at speeddcubes with lots of overlapping parts such as the Fangshi Shuang Ren, they are so stable because the overlapping parts they have. If I really need to, I can make a piece that screws into the core after you slide the square piece on that you could use to tension the cube. I really don't want to use metal screws though...

Also, rather than make the square piece click on, I think it would be better to have it 2 seperate pieces that lock together around the centers.

I also realize that my idea before about making the pieces solid plastic really would raise the price up. I think I may make the pieces lighter by hollowing the out, yet not having them be seperate pieces of plastic. I would make the holowness parts on the inside be connected with the outside through small holes or pockets similar to what the Fangshi Shuang Ren has on its corner pieces.

I was just thinking about a better way to orient all of the pieces on a 3x3. So far, heres what I've come up with.

First - orient edges.

Second - solve the bottom color

Third - use an algorithm to solve the top color

To recognize a case for step 3, you would look at where the top colored edges are in the middle layer, then look at the corner orientation case, then look at where the top edges on the top side are. I'm not sure how many algorithms this would take, but I will calculate this soon. I can see it taking anywhere from 100- 150 algorithms. It would be possible to reduce the alg number by using partial E-layer control while placing the pieces in the bottom layer.

(EDIT: This takes 103 algorithms not including mirrors)

You could also do this with counting opposite colored sides as the same, reducing the length of algs. This could be done counting opposite colors as the same for only corners or only edges aswell.

I went and tried a couple of cases in cube solver using only L U and R moves, and the longest algorithm I found was 11 moves. I think I may adapt this to my method idea that used PBL and solved the E slice at the same time. If it were possible to predict the case through inspection, this would be a very quick way to orient everything. I'm not sure it would be very humanly possible to predict during inspection though, but you may get hints as to what case you will get. This would be easier or more feasible counting opposite colors as the same because it would take less moves to get to the point that you use the algs.
 
Last edited:
wouldn't that just be a zz opening with zbll as the second step, then a sandwich/waterman closure

now to my discoveries:
Over time, I have found that during F2L, you can place pairs that don't match using two simple rules: if the corner and edge share a face, then that face will be on opposite sides in the pair. If the edge and corner share no color, than the complementary colors will share the same sides of the pair. Usually, the cube will give you one of these pairs, and if you use it, the cube will usually give you a normal pair for free. No idea why.

Another is that you can place completed pairs during the cross formation so that you don't have to worry about them later. Essentially forming a Petrus block because it was given to you.
 
No, it isnt quite like that. In ZZ, you orient edges and solve the DB and DF edges. With my version, you solve the whole bottom color, but don't worry about permutation of any peces. ZBLL is a lasy layer only substep that permutes the whole top layer and orients corners. With mine, edges are already oriented and you don't care about the permutation (as long as you seperate E-layer edges and U-layer edes). You just orient top layer corners whle moving the top layer edges to the top. It does have the sandwich closure, but not waterman. Waterman solves a single side, then the top corners. It then solves some edges in the top layer and the middle layer at the same time until it solves.

From your explanation, I am getting this: Solve F2L with the E layer not lined up, then line it up afterwards. I've actually done this before. It sort of reminds me of keyloing.

The thing about solving pairs with the cross is basically just the X-cross. It just seems that, rather than plannig the F2L pair, you just use it when it happens. That won't work well if you already plan out your whole cross during inspection.

You talking about F2L reminded me of an idea I had that I threw away because of practicalities. I may aswell post it though.

I was once searching for a good way to incorporate the effectiveness of EG into 3x3 methods. One of my ideas was to not worry about where you place F2L pairs, and permute the pairs while doing another step of the last layer. One good thing about this is that it won't affect the last layer cases at all because your swapping pairs. It is possible to swap any even number of peices regardless of them being corners or edges. You always have an even number because your using pairs that consist of 2 pieces. Basically, you won't have parity in your top layer. Another good thing is that you can AUF the U layer independently of the pairs.

One bad thing about this is the fact that, rather than having only 3 EG cases, you would have 10 because of having centers placed.

This idea could be used with many different last layer substeps. If you were going to do it with PLL or CLL, you would want to always make sure you have 2 F2L pairs in their proper places, ruducing you back down to the 3 EG cases.

It is nice to finaly see someone else posting something on here. I'm also sorry if I completely mis-understood you.
 
Last edited:
I'm not sure if your referring to my idea for orienting or for th EG F2L. The EG F2L is not practical as I've already stated. The orientation idea I had could be quick, but I'm not sure it would beat any other speedsolving method, so its not worth the amount of algorithms for it. I also stated when I made this thread that it was for ideas that you didn't think deserve their own threads.
 
New method

I have developed a new method for solving--somewhat similar to petrus, as well as an edge-first method. First you solve all the edges. You should be able to figure to figure this out. Then use URU'L'UR'U'L to put two corners next to each other in place, oriented correctly. This will take some experimentation. Your cube should now look like you have finished petrus step three. Now use URU'L'UR'U'L again to move corners around until they are all positioned correctly. now use R'D'RD to orient them. You can also use some PLL algs to move the corners. If anyone can give me some better algorithms, it would be appreciated. Your cube should now look like this. If you notice any holes, please notify me.
 
I'd say it's fairly inefficient seeing as it requires you to scan the entire cube during inspection and then again when you finish the edges. It would make look-ahead very difficult. The advantage of methods such as CFOP and roux is that it moves the parts you've already finished to the part of the cube you can't see as well and allows the unsolved pieces to be most clearly in your line of sight.
 
New method

Sorry to burst your bubble, but you have come up with something that many people have come up with before. It's amazing how few ways there are to solve the cube, so it is very rare for someone to come up with something actually new.

Edges first has been speculated about a lot, and the general consensus is that it is not feasible for fast speedcubing. Even using BH for the corners step, the fastest you could expect to get the corners is about 8 seconds, and when added to the time spent solving edges, the best you can probably do is about 13 seconds. That being said, feel free to try to be creative, because ultimately you will learn more about the cube.

By the way, this method shares nothing in common with Petrus. Petrus is about blockbuilding and edge orientation, neither of which this method contains.
 
One of the steps ends up being solvable with petrus step 4 onwards

You could solve all the edges LBL, which would speed things up.

So if you want to teach someone a sub-30 method that is easy, here it is.

You can also put in one corner while you are doing the edges. That would eliminate a large part of identification of the corners.
 
Last edited by a moderator:
You must log in or register to reply here.
Back
Top