The New Method / Substep / Concept Idea Thread

[Shiv3r]

i can go for aroun 4 minutes, but this is probably because I started to take 4x4 seriously only a couple days ago.

I'm pretty confident a world class solver could get sub 40 with this because of how intuitive it is, and block efficency.

the problem I have with this method is the same I have with petrus: ergonomics. Petrus is low movecount, but because of the high amount of rotations involved not very many people use it. Also more structured methods are often faster because of the lookahead advantage.
I only started to take 4x4 seriously in january, so I know what you mean.

 

[Neuro]

40 move NLS solve using some advanced cancellations. Had a 41 move w/ no cancellations and full step but I overwrote the alg.cubing link on accident.
https://alg.cubing.net/?setup=U2_D_...U_R_D_R-_U-_R_D-_R y_B-_U2_B-_D-_F_R2_F-_D_B2

Ok so I did 12 solves with optimization and got a mean of 45.92 moves, so the range using α=0.025 would be between 38.72 and 53.13 moves. I think that 30-40 is possible, but that it'd be rare at best. Some of the cases are pretty nasty in terms of moves and sometimes it's hard to build top block; but depending on combinations you can get top block solved, 3 move LS insert, and 7 move LL.

I will continue to use ZZ-CT as my main variant but use this in nice cases. Still, it would be useful for those wishing to learn ZBLL but don't know where to start, are really lazy with learning algs, and/or want eazy(ish) recog for 1LLL.

 

[Teoidus]

Why would you use -CT if NLS has a lower avg movecount?

 

[crafto22]

Why would you use -CT if NLS has a lower avg movecount?

Because it doesn't. He's optimizing his solves, as he said, and when optimizing CT solves the move count is about the same if not lower. Additionally it seems like CT has better recog.

 

[Teoidus]

I highly doubt that optimized CT solves get 46- movecounts.

15 TTLL + 10 TSLE leaves ~20 moves for EOF2L-1(!!).

Unless by "optimized" you mean an FMC attempt.

@Neuro can you do a speedsolve movecount average?

 

[Neuro]

Because some of the cases get really nasty. Sometimes it takes 7 moves to make the block, 12 moves to insert the LS, and 14 moves when optimized to solve ZBLL. So in real speedsolves, move count with bad cases may get in the 60's if not careful.

Basically if no easy block can be made I'll use CT, and maybe if the NLS case is bad and CT would be more efficient.

A speedsolve average is actually my next task. I have the scrambles now so I can probably have the average out tomorrow

EDIT: the way I optimized my solves was using HARCS to find the best EOline+LB and then using cube explorer to solve F2L-1 and top block seperately. From there I used the algs from the wikia to solve ZBLL and my currently generated algs for NLS

 

[Shiv3r]

A speedsolve average is actually my next task. I have the scrambles now so I can probably have the average out tomorrow

EDIT: the way I optimized my solves was using HARCS to find the best EOline+LB and then using cube explorer to solve F2L-1 and top block seperately. From there I used the algs from the wikia to solve ZBLL and my currently generated algs for NLS

okay Neuro, Ill try to do some solves with it, do you have like an alg sheet(printable hopefully) I can use?
EDIT: does HARCS have an optimal EOline solver?
EDIT OF EDIT: I think the reason why ZZ-CT is a great LS method is because it does not rely on any LS setup moves like WV does, because it only looks for one piece at a time.

 

[Shiv3r]

also Neuro what is your system for naming the cases for NLS?

 

[Neuro]

@Shiv3r, HARCS does have an EOline solver. Set it to solve with ZZ, put in the scramble, and set your style(s) *I usually limit it to UF and DB for simplicity* and from there all you need to do is put in EO line and it'll give you optimal EO lines for both styles.

I really like CT for the same reason. Recog of LS+LL is really easy and skips are relatively frequent. My main problem with it is alg count. It's not a problem for me, but might be for other people; hence why I created NLS. It's specifically tailored to people who want a 1LLL but hate learning algs. [85] is still a little high, but definitely learn-able and some of the LL should be known as well as NLS having many intuitive cases.

My recog is based off of where the edge is [A B C] and where the last d face sticker is [A-L]. So B A would have the edge is in the slot and the last corner would be in U F L and is facing left (insertable using R U' R')

Revised the page slightly, noticed an issue with the B/C sets. Generate in Cube Explorer to see cases. Will make official alg sheet soon, but for time being load and switch from generator to solver:

 

[Shiv3r]

@Shiv3r, HARCS does have an EOline solver. Set it to solve with ZZ, put in the scramble, and set your style(s) *I usually limit it to UF and DB for simplicity* and from there all you need to do is put in EO line and it'll give you optimal EO lines for both styles.

I really like CT for the same reason. Recog of LS+LL is really easy and skips are relatively frequent. My main problem with it is alg count. It's not a problem for me, but might be for other people; hence why I created NLS. It's specifically tailored to people who want a 1LLL but hate learning algs. [85] is still a little high, but definitely learn-able and some of the LL should be known as well as NLS having many intuitive cases.

My recog is based off of where the edge is [A B C] and where the last d face sticker is [A-L]. So B A would have the edge is in the slot and the last corner would be in U F L and is facing left (insertable using R U' R')

Revised the page slightly, noticed an issue with the B/C sets. Generate in Cube Explorer to see cases. Will make official alg sheet soon, but for time being load and switch from generator to solver:

However, The TSLE not really "algs", they are more like LSE EO, squan subeshape, or sarah's intermediate, where you cycle through cases and eventually solve it. Then you learn optimized algs for the annoying cases later. so in reality, the alg count for ZZ-CT(NOT case count b/c all TSLE's are cases but not algs) is 78, which is lower than NLS. also recognition is easier.

 

[Neuro]

However, The TSLE not really "algs", they are more like LSE EO, squan subeshape, or sarah's intermediate, where you cycle through cases and eventually solve it. Then you learn optimized algs for the annoying cases later. so in reality, the alg count for ZZ-CT(NOT case count b/c all TSLE's are cases but not algs) is 78, which is lower than NLS. also recognition is easier.

I agree with the approach you take with TSLE, but learning how to effectively cycle the cases can take beginners a long time to master and use effectively. If we argue about alg vs case count, then NLS also has a few cases that can be effectively cancelled because of how intuitive they are (a standard 3 move insert is one of my "algs".) If we look at mirroring cases, S and AS reduces to only 6 algs, and I'm sure that some of the other ZBLL's mirror. I'd argue that they probably have very similar alg counts and that CT has more cases.

However, CT and NLS are aimed at different people. CT is a 2 look LL+LS, NLS is a 1LLL. CT is for people who have no problem learning algs, want a recog system they are familiar with, and will bargain with LL+LS. NLS is for people who don't like algs, are willing to change their recog style, and see benefits of a true 1LLL. Therefore while they can be compared, it'll ultimately come to personal preference. They both have advantages and disadvantages. Personally, I will use them interchangeably, as both of them have probability of low movecount and ergonomic solves.

 

[Shiv3r]

I agree with the approach you take with TSLE, but learning how to effectively cycle the cases can take beginners a long time to master and use effectively. If we argue about alg vs case count, then NLS also has a few cases that can be effectively cancelled because of how intuitive they are (a standard 3 move insert is one of my "algs".) If we look at mirroring cases, S and AS reduces to only 6 algs, and I'm sure that some of the other ZBLL's mirror. I'd argue that they probably have very similar alg counts and that CT has more cases.

However, CT and NLS are aimed at different people. CT is a 2 look LL+LS, NLS is a 1LLL. CT is for people who have no problem learning algs, want a recog system they are familiar with, and will bargain with LL+LS. NLS is for people who don't like algs, are willing to change their recog style, and see benefits of a true 1LLL. Therefore while they can be compared, it'll ultimately come to personal preference. They both have advantages and disadvantages. Personally, I will use them interchangeably, as both of them have probability of low movecount and ergonomic solves.

I will argue with your Idea that NLS is a 1LLL, as you have more looks than an insert+LL. you build a block on the top then use algs to insert the pair. Also, if youre talking about alg count, ZZ-b has less algs than ZZ-NLS and is a more efficient LS IMO, and also has the 1look LL factor. I mean I think this method should really be called ZZ-tripod.

Now, I know a lot of ZZ solvers know COLL, some ZBLL, WV etc. I think that if you already have a U-layer pair then pairing up the last edge then inserting the pair may be more efficient, so learning some of the cases will be Nice.

Hey, so I was thinking of how to Make ZZ much more efficient, and I think I know how, for about as many algs as there are for COLL-EPLL(if you count L3C, which can very easily be done with commutators). It has already been invented by Kenneth gustavvson,who called it fish-n-chips, if you do it over CFOP+VHLS. Here are the ZZ-Snyder step 1(what Im calling it) algorithms: http://blog.naver.com/dmdrlrndk/90192057323. Also called EP+1

and if youre anything like Tony Snyder in efficiency you will probably get around 45-50 moves on average. Also, this is probably easier to recognize than Speed-heise, with around half the algs for the first step. I will post example solves later if you want.
I think the way to go for better efficiency is to start emulating more FMC tactics for the last steps, with things like L3C.

 

[Neuro]

I will argue with your Idea that NLS is a 1LLL, as you have more looks than an insert+LL. you build a block on the top then use algs to insert the pair. Also, if youre talking about alg count, ZZ-b has less algs than ZZ-NLS and is a more efficient LS IMO, and also has the 1look LL factor. I mean I think this method should really be called ZZ-tripod.

Now, I know a lot of ZZ solvers know COLL, some ZBLL, WV etc. I think that if you already have a U-layer pair then pairing up the last edge then inserting the pair may be more efficient, so learning some of the cases will be Nice.

Yeah I've been debating as to if it's a 1LLL or not, but I'm simply not sure what it truly is. I will call it ZZ-T but I will continue to call the algset NLS as tripod LS is pretty much a thing already and traditional tripod doesn't have edges preoriented. Also, doesn't ZZLL still have over 100 cases?

 

[Shiv3r]

Yeah I've been debating as to if it's a 1LLL or not, but I'm simply not sure what it truly is. I will call it ZZ-T but I will continue to call the algset NLS as tripod LS is pretty much a thing already and traditional tripod doesn't have edges preoriented. Also, doesn't ZZLL still have over 100 cases?

over 100 cases, yes, but only 71 algorithms. so the alg count for ZZ-CT and ZZ-B is about the same, except that ZZ-CT is a highly advanced variation of ZZ-C as far as I can tell and ZZLL is a variant of ZZ-a. depends where you want to go with ZZ, there are a few ways.

 

[Neuro]

I agree, many people will take different directions with LL. ZZ-Snyder looks interesting, I'll play around with it

 

[Shiv3r]

I am not even Y axis neutral, and because I dont main ZZ my efficiency is crap, but here is a 53 HTM example solve with ZZ-snyder
I got a really bad case for the setup. this is maybe the worst possible case to get, moves-wise. if not, this could have easily been much more efficient.
but if I had done COLL+EPLL it would have been 55 HTM, heres another example solve where I did what I would normally do , COLL-EPLL, and got a 55 HTM solution. here it is.

So I got a nasty ZZ-snyder setup case but I still got a lower movecount than COLL-EPLL would have been.

If you want me to do more example solves then I will, when I have time. I will also probably tell you the movecount of the "normal" solve and then compare it.

 

[Teoidus]

Isn't zzll better?

This approach doesn't look effective unless you memorize algs from EOF2L-1 to L3C (aka the Heise step).

 

[crafto22]

Isn't zzll better?

This approach doesn't look effective unless you memorize algs from EOF2L-1 to L3C (aka the Heise step).

What's the average movecount for ZZLL?

 

[Neuro]

I'm working on speedsolving stats for NLS and got a 41 move solve with a pretty nice TB cancellation
https://alg.cubing.net/?setup=R2_D_...2_R-_U_R-_U-_R2_U_R-_U_R_ R-_U_L_U-_R_U_L-_U2

 

[Teoidus]

What's the average movecount for ZZLL?

~14