[Help Thread] Waterman Discussion

[elrog]

I never bothered to learn the algorithms for waterman. I've always just done commutators intuitively whenever I do waterman. Also, nowhere in your post did you say that the first layer needs to be on the left side after CLL. Adding that to your post may clear up some confusion about doing the R and M layers while you never say anything about the L layer.

I find found that you can very easily insert edges from the M layer into the R layer if you leave a single edge off of the first layer. I also like doing it more like this:
Build a 1x2x3 on the left side like in Roux, but then just solving the remaining 2 bottom corners rather than the whole R block.
You could then do some sort of CLL and continue with the edges very similarly to waterman.

Doing the corners like this also opens up possibilities like having doing CLL with the corners swapped in the bottom layer (like in EG 1) or having them misoriented (like some sort of ROFL subgroup)

 

[rybaby]

I think 16 is definitely possible for an average first layer. Corners are solvable in 9 average. The rest of the cube is possible to do in less than 30 turns. I usually take 20.

20 turns for the rest of the cube? To me it seems like it would take 25+:
1. Two redges -- 9 moves
2. Two more redges + EO -- 12 moves
3. Midge Permutation -- 4 moves.

Any tips on how to improve this?

 

[somerandomkidmike]

20 turns for the rest of the cube? To me it seems like it would take 25+:
1. Two redges -- 9 moves
2. Two more redges + EO -- 12 moves
3. Midge Permutation -- 4 moves.

Any tips on how to improve this?

I honestly have no idea what I was talking about. It seems reasonable to take 25 turns on average. I mean, I can get 20 turns sometimes, but my average would definitely be higher than that.

As far as your signature goes, learn Roux. It's a better method than Waterman.

 

[Ranzha]

Also, nowhere in your post did you say that the first layer needs to be on the left side after CLL.

Well, you can always keep your first layer on L the whole time

 

[rybaby]

As far as your signature goes, learn Roux. It's a better method than Waterman.

What makes you say that?

 

[somerandomkidmike]

What makes you say that?

1) Cube rotations: It's nearly impossible to avoid cube rotations using Waterman
2) After positioning the middle edges (the last 3-4 turns) you never have to reposition the R-layer in Roux. You have to do that most of the time using Waterman.
3) L6E algorithms are generally faster to execute than the algorithms to solve the Right edges.
4) There are less algorithms you have to learn to become fast, so you'd likely make improvements more quickly with Roux.
5) You can plan out the First block in Roux during inspection. It's much harder to plan out an entire layer.

Yes, I still use Waterman, but sometimes there are other reasons for using a method. I used to think that Waterman could stack up against any of the methods out there. I'm sure you can tell that from some of my 2009 posts.

I think it can stack up to Petrus, and MAYBE ZZ. I don't think anybody is going to be using it to get a sub-8 average though. Just on my own time I've had a few sub-13 3 of 5s, and a sub-12 average 3 of 5. I think that's the most I'll ever be able to do with this method.

Don't get me wrong. It is a really fun method, and it's possible to be pretty fast. You're probably not going to be breaking any records with it.

Edit: But with some tricks it is possible to do the rest of the cube after the corners using waterman in UNDER 25 turns average. Maybe not 20, but not as high as 25. I had a bit more of a look into it last night.

 

[rybaby]

1) Cube rotations: It's nearly impossible to avoid cube rotations using Waterman
2) After positioning the middle edges (the last 3-4 turns) you never have to reposition the R-layer in Roux. You have to do that most of the time using Waterman.
3) L6E algorithms are generally faster to execute than the algorithms to solve the Right edges.
4) There are less algorithms you have to learn to become fast, so you'd likely make improvements more quickly with Roux.
5) You can plan out the First block in Roux during inspection. It's much harder to plan out an entire layer.

Yes, I still use Waterman, but sometimes there are other reasons for using a method. I used to think that Waterman could stack up against any of the methods out there. I'm sure you can tell that from some of my 2009 posts.

I think it can stack up to Petrus, and MAYBE ZZ. I don't think anybody is going to be using it to get a sub-8 average though. Just on my own time I've had a few sub-13 3 of 5s, and a sub-12 average 3 of 5. I think that's the most I'll ever be able to do with this method.

Don't get me wrong. It is a really fun method, and it's possible to be pretty fast. You're probably not going to be breaking any records with it.

Edit: But with some tricks it is possible to do the rest of the cube after the corners using waterman in UNDER 25 turns average. Maybe not 20, but not as high as 25. I had a bit more of a look into it last night.

Ranzha V. Emodrach and I have been generating COLLs on R to make a method like Waterman rotationless: http://www.speedsolving.com/forum/showthread.php?44037-COLL-on-R-(Table). But I see what you mean.
I guess they both have a similar move count average. From my observations it seems as if 4 tps on Waterman could get sub-12. I think what's holding me back from Roux is the block building and LSE. I'm more comfortable with Waterman simply because I use Corners First now and inserting redges and ledges is what I'm used to. But I'm fairly sure it could beat Petrus with practice, considering the fastest (former) Petrus user (EJ) averaged around 12 seconds (correct me if I'm wrong), and there are countless Petrus users -- you say you average around 13 seconds, and there are only a few Waterman users today.

I just feel like it should be able to stack up to other methods -- I mean, Waterman averaged around 18 seconds in the 80s, so he had similar times to J. Fridrich. I still think that some optimizations would help Waterman to have more potential.

How have you progressed over the years with times? Have your times plateaued around 13-14 for a while?

thanks for the advice!

 

[somerandomkidmike]

Ranzha V. Emodrach and I have been generating COLLs on R to make a method like Waterman rotationless: http://www.speedsolving.com/forum/showthread.php?44037-COLL-on-R-(Table). But I see what you mean.
I guess they both have a similar move count average. From my observations it seems as if 4 tps on Waterman could get sub-12. I think what's holding me back from Roux is the block building and LSE. I'm more comfortable with Waterman simply because I use Corners First now and inserting redges and ledges is what I'm used to. But I'm fairly sure it could beat Petrus with practice, considering the fastest (former) Petrus user (EJ) averaged around 12 seconds (correct me if I'm wrong), and there are countless Petrus users -- you say you average around 13 seconds, and there are only a few Waterman users today.

I just feel like it should be able to stack up to other methods -- I mean, Waterman averaged around 18 seconds in the 80s, so he had similar times to J. Fridrich. I still think that some optimizations would help Waterman to have more potential.

How have you progressed over the years with times? Have your times plateaued around 13-14 for a while?

thanks for the advice!

I've basically plateaued. Yeah, I've had spurts of improvement in the last year or so, but I haven't maintained my interest in cubing for long enough to make that permanent. I'm sure I could average sub-12 if I practised a lot and I learned all of the algorithms (rather than just the necessary ones).

Don't get me wrong. Waterman is a great method. I just think Roux has more advantages over it. If you're wanting to get a competitive average, you'll be blockbuilding with the FL in Waterman anyway, so why not just learn how to do Roux as well? I mean, it's not a really hard switch. Even for me, after using Waterman for so long, I can switch to Roux and get sub-20 averages.

 

[rybaby]

I've basically plateaued. Yeah, I've had spurts of improvement in the last year or so, but I haven't maintained my interest in cubing for long enough to make that permanent. I'm sure I could average sub-12 if I practised a lot and I learned all of the algorithms (rather than just the necessary ones).

Don't get me wrong. Waterman is a great method. I just think Roux has more advantages over it. If you're wanting to get a competitive average, you'll be blockbuilding with the FL in Waterman anyway, so why not just learn how to do Roux as well? I mean, it's not a really hard switch. Even for me, after using Waterman for so long, I can switch to Roux and get sub-20 averages.

I see what you mean. Method neutrality between those would be cool. Maybe instead of using the algs for inserting the last redge + EO could be replaced with LSE. And I might as well learn blockbuilding techniques. I have been learning COLLs so those can really carry over to anything (though not as efficient for Waterman as the CLLs). Should I just go straight away with full COLL, or use 2-look? Or something else?

 

[somerandomkidmike]

I see what you mean. Method neutrality between those would be cool. Maybe instead of using the algs for inserting the last redge + EO could be replaced with LSE. And I might as well learn blockbuilding techniques. I have been learning COLLs so those can really carry over to anything (though not as efficient for Waterman as the CLLs). Should I just go straight away with full COLL, or use 2-look? Or something else?

For the most part, you should be learning the Roux CMLLs. They're generally quite good to begin with. However, for when you do Waterman solves, keep in mind that some of the Waterman algorithms will be shorter. (ie. R2 U2 R' U2 R2 vs double sune)

 

[rybaby]

For the most part, you should be learning the Roux CMLLs. They're generally quite good to begin with. However, for when you do Waterman solves, keep in mind that some of the Waterman algorithms will be shorter. (ie. R2 U2 R' U2 R2 vs double sune)

But if I do Waterman, won't CMLLs mess up the M-slice (and part of the already solved layer)?
I know COLLs and 3x3x3 CLLs won't disturb any part of the first 2 layers.

 

[somerandomkidmike]

But if I do Waterman, won't CMLLs mess up the M-slice (and part of the already solved layer)?
I know COLLs and 3x3x3 CLLs won't disturb any part of the first 2 layers.

Most of them won't mess up anything that you'd have to worry about in Waterman. Most of them are essentially CLL algorithms. For the few that affect FL in Waterman, you'll be able to tell right away, and you can just switch it out with a similar CLL algorithm.

 

[Ranzha]

Most of them won't mess up anything that you'd have to worry about in Waterman. Most of them are essentially CLL algorithms. For the few that affect FL in Waterman, you'll be able to tell right away, and you can just switch it out with a similar CLL algorithm.

Or maybe learn COLL for the R layer and take note of the edges in the M-slice for planning redge solving? That's really why I wanted to work to get the COLLs done with Ryan here.

 

[somerandomkidmike]

Or maybe learn COLL for the R layer and take note of the edges in the M-slice for planning redge solving? That's really why I wanted to work to get the COLLs done with Ryan here.

COLL takes quite a few more turns than CLL for a lot of the algorithms. Plus the algorithms are often not as fingertrick-friendly. Sure, CLL on the R-layer would be fine to do, but I don't understand why you'd do COLL.

 

[rybaby]

COLL takes quite a few more turns than CLL for a lot of the algorithms. Plus the algorithms are often not as fingertrick-friendly. Sure, CLL on the R-layer would be fine to do, but I don't understand why you'd do COLL.

With COLL, you could predict the positions/orientations of redges because COLL doesn't affect edges in the F2L and preserves orientation of edges in the Last Layer. Also, these algs could carry over to cross on left solvers.

 

[rybaby]

1)
Edit: But with some tricks it is possible to do the rest of the cube after the corners using waterman in UNDER 25 turns average. Maybe not 20, but not as high as 25. I had a bit more of a look into it last night.

What kinds of "tricks?" I know Waterman himself averaged around 40-45 turns for the cube, which seems difficult. Still, I would be very interested in any tricks for reducing move count, especially during edge solving. And I guess the layer is important, too.

 

[BaMiao]

Are there any resources, with lists of algs, other than that website with all the java applets? It doesn't even load on my computer.

In particular, I'd be very interested in a reduced alg set (comparable to 2-look OLL/PLL), geared toward someone just getting into the method.

Any help is much appreciated.

 

[rybaby]

Are there any resources, with lists of algs, other than that website with all the java applets? It doesn't even load on my computer.

In particular, I'd be very interested in a reduced alg set (comparable to 2-look OLL/PLL), geared toward someone just getting into the method.

Any help is much appreciated.

As far as I know (I have searched quite a bit), rubikscube.info is the only website with Waterman method documentation.

On rubikscube.info there is Waterman's original booklet with many tables of sequences. Table 1 is for corners--you can just pick a formula from every orientation and then use the solved orientation cases to permute them. (9 sequences compared to 42 for full CLL). Leave out an edge and solve 3 redges + 1 ledge intuitively. Then use sequence 1 from table 5b to solve the last redge (this step can later be advanced by learning the whole table). Then use tables 6 and 7 for midge orientation and permutation. Good Luck!

 

[Chris_Cube]

Hey there,
so I wanted to start this thread as a form of motivation and discussion. At the moment I switched from CFOP to Roux to Waterman and I think Waterman is my method. It's totally fascinating and oldschool.
This is some sort of Challenge thread and i am trying to achieve at first a Sub-30 average with Waterman.
Also I want to document how to be fast with this method.

To be sub-30 my ideas are as follows:
1. First Layer around 8 secs. ( This is at the moment my fastest, i hope to fasten especially this up)
2. Waterman CLL (fast and logical) - maybe 2 - 3 secs
3. Inserting the redges maybe takes like 5 secs
4. Permuting midges takes maybe 1 - 2 secs

Doing the math you get 18 secs.
More to come...

Just some questions, how do I speed up Part 1 ?

 

[KW24]

Very interesting choice of method.