3x3 Method - KwS

I came up with quite fast idea:

-Do an E-layer (aka belt) (initutive)
-Orient both layer edges in one step without putting them in correct layers (MU) (19 algs without z2 correcting, but it's less actually)
-Orient Corners (+ simple parity)
--Top layer (CO)
--z2
--If You had the parity, then do one of 2 algs ( R U2 R U R' R U2 R2 (here put the corner at RFD) and mirror alg with corner at LFD)
--Top layer (CO)
-Put Egdes in correct layers (square-like) (R2UD) (nitutinve)
Put corners in correct layers (R2UD) (less initutive, but still fast)
PLL D) (Fridrich style + 2 Parities) (M2 U2 M2) (R2 U2)x2 R2
after 50 solves i did 2 sub 20s

Algs needed: 57 (you should know 3/4 of them if You use Fridrich)

If you know the OLL on one side be4 the egdes, do it. (very gr8 advantage of knowin full fridrich )

Named after my Nickname - KwS

My further ideas:
-develop algs for Top CO layer WITH PARITY (that would give much more algs I think)
-control edge orientation while solving last belt piece

I did 2 sub 20s and average of 12 25.90 after 50 solves.

Any comments?

It seems to have a lot of steps, and recognition for step 2 seems like quite a challenge...

nope it's fast for me of course

What's the movecount like?

Interesting idea, doubt it'll be a WR beater tho

The ending is similar to the final stages of this method.

Would it possibly be faster to separate and permute corners before doing edges using the square-1 corners-first method. Edges can then be separated with M/U moves and permuted using EPLL both sides, or you could complete the edges using a ROUX-style finish..


EDIT: With the corner orientation phase, I'm not sure about the alg you've got down there, but to save on inspection, look at either the U or D layer. If its not a recognisable OCLL then do R U2 R U R' U R' and it will remove parity. Then just do your OCLL algs on each side as normal.

If you're aiming for fewer moves its possible to remove parity simply by doing R2 with the top and bottom layers aligned correctly. This will create recognisable OCLL cases on both sides, which you solve and then finish with a final R2 to correct the E layer.

EDIT2: Has anyone actually seen a formal spec for the 'belt' method? Is there one, or is it just a sub-step like cross? I'd imagine if starting with a belt, then the approach outlined here eg. (orient -> separate -> permute) would probably be the most sensible way to finish..

DavidWoner said:

What's the movecount like?

Click to expand...

I think any belt-style method is probably going to be a move-count fail. Here's a conservative estimate for the number of moves:

Belt: ~ 6 moves
EO: ~7 moves
OCLL x2 (inc parity): ~20 moves
C+E Separation (intuitive): ~10 moves
PLL x2 (inc parity): ~25 moves
TOTAL: 68 moves

Doing EO+OCLL simultaneously will require an infeasibly large number of algs, so must be done in two steps. Separation can be done in one step by intuition, but doing them simultaneously using algs again creates a huge number of algs: see here. Finishing with 2x PLL (including parity) is again quite a large number of moves. Parity also creates further problems with recognition.

As well as move count issues, belt is always going to suffer from lookahead and ergonomic problems, because the U/D layers need to be solved simultaneously, meaning lots of horrible x2 rotations

hmm - nice idea about corners 1st in separation I'll try that after I return home

your ideas ok, i would recommend only doing one PLL tho, because you could do the F2l afer orientation with simple R2U2 then one PLL, thats how i do it. if thats a bad idea let me know...