# Roux-breaker? The YruRU method

#### Devagio

##### Member
14.97 OH single with YruRU!!

D' R B2 D R' B L' F' D2 F2 L U2 L2 U2 B2 R F2 L2 D'

U2' S' // CPLine
U2' u' R r U' R u2' // FB
U2' R' r U' r' r' R U R' U' r // EO
U' R U' r2 // Stripe
U2' R' U R // Square
U2' R U R' U2' R U' R' // F2L
U' R2 U R U R' U' R' U' R' U R' U2' // 2GLL

49 moves/14.97s = 3.27 TPS

Been practicing a lot lately but it looks like I might have been practicing the wrong method rip. I probably didn't even overinspect this time considering how easy the CPLine was.
That is quite exceptional, especially given you've been practicing this for such a ridiculously short time! It wasn't a very lucky solve though, you'd find such 2-move CP-lines once every dozen solves or so. Imagine what you could achieve on lucky scrambles!

#### PapaSmurf

##### Member
This method is literally Briggs and Briggs is LEOR but you solve CP. For TH, this is pointless. ZBLL is good enough and doing CP for alg reduction is dumb. You also restrict how you can solve EO. For OH, this is also dumb for exactly the same reasons as for TH. The movecount of the method is good, the lookahead is not super great, especially as you're not gonna be inspecting EO in inspection (which is necessary for good LEOR). This means that you're gonna be doing a difficult step followed by a step with bad recog, before 2 nice steps. It's nice to see CP being explored at least, but there are definitely better ways to improve speedsolving (conjugation maybe, or writing an AI that finds good methods). If this were the Roux killer, Roux would've died 4 years ago, yet, as we can see from Sean, Kian and others, that isn't the case. From a cube theory/maybe fancy niche FMC, understanding CP is cool and I'm glad that I learnt, but it's not speedsolving useful (except maybe maybe maybe ZZ OH last slot using full CPLS [solving the pair and CP], but even then just OH ZBLL).
TL;DR solving CP early doesn't benefit anything and this isn't new (it is fun though), but keep up searching for new methods.

#### CuberStache

##### Member
you're not gonna be inspecting EO in inspection (which is necessary for good LEOR).
Wait, so when you do LEOR solves, you plan the whole first block and know where bad edges will be after that?

#### mukerflap

##### Member
Wait, so when you do LEOR solves, you plan the whole first block and know where bad edges will be after that?
thats the only way its viable

#### dudefaceguy

##### Member
(it is fun though), but keep up searching for new methods.
This is the most important part.

#### Devagio

##### Member
This method is literally Briggs and Briggs is LEOR but you solve CP. For TH, this is pointless. ZBLL is good enough and doing CP for alg reduction is dumb. You also restrict how you can solve EO. For OH, this is also dumb for exactly the same reasons as for TH. The movecount of the method is good, the lookahead is not super great, especially as you're not gonna be inspecting EO in inspection (which is necessary for good LEOR). This means that you're gonna be doing a difficult step followed by a step with bad recog, before 2 nice steps. It's nice to see CP being explored at least, but there are definitely better ways to improve speedsolving (conjugation maybe, or writing an AI that finds good methods). If this were the Roux killer, Roux would've died 4 years ago, yet, as we can see from Sean, Kian and others, that isn't the case. From a cube theory/maybe fancy niche FMC, understanding CP is cool and I'm glad that I learnt, but it's not speedsolving useful (except maybe maybe maybe ZZ OH last slot using full CPLS [solving the pair and CP], but even then just OH ZBLL).
TL;DR solving CP early doesn't benefit anything and this isn't new (it is fun though), but keep up searching for new methods.
There are multiple flaws in this argument; but rather than dissecting, I’ll suggest we wait for some time and let the results by people who’re trying it out speak for themselves.

A way to optimise EO is to standardise the way it is solved; this will not only improve efficiency, TPS and look ahead, but also vastly help recognition.
[The way I suggest to recognise EO is perhaps an unscientific one, but it seems to work for me more and more with practice.
You’ll have either 0, 2, 4, 6 or 8 edges flipped. That’s a pretty discrete set; so rather than looking for flipped edges, simply give the cube a glance and “intuit” how many edges are flipped; with familiarity it becomes automatic. The accuracy will improve with time.
If this seems like crap, it probably is; just ignore this bracket and read on.]

0 bad edges (1/256): EO skip

a. The good edge is in DB or DF:
(r U’ r) U R U (r U’ r) / mirror from back
b. The good edge is in R or U layer:
Bring the good edge to DR and do
(r U’ r’) U’ (r’ U’ r)

Place them at UF, UR and do
r U R’ U’ r
b. Opposite bad edges in U layer:
Place them at UF, UB and do
(r U r’) U’ R’ (r U’ r)
c. Both Bad edges in R layer:
Convert to case a.
d. One bad edge in U layer and one bad edge in R layer:
Place the U edge in UF and the R edge anywhere except FR, and do a trigger similar to case a.
e. One Bad edge in DF and other not in DB
Place the second edge in UR (or BR or DR) and do
(r U’ r’) R’ (r U’ r)
f. Bad edges in DF and DB:
(r’ U r’) U’ R’ (r U’ r)
[in each case, you can also perform a mirror from back]

a. Bad edge in DF, not in DB, 3 in U layer
Create a Roux arrow and do R’ r U’ r
b. Bad edge in DF, not in DB, 2 bad edges opposite in U layer:
Convert to case a.
c. Bad edge in DF, not in DB, rest of the cases:
Make a staircase pattern and do r U’ r
d. Bad edges in DF and DB, opposite bad edges in U layer:
Place the U layer edges in UR, UL and do
(r U’ r’) R U (r’ U’ r)
e. Bad edges in DF and DB, rest of the cases:
Place one bad edge in UB, none in UF, and do r2 to get an arrow in the next move.
f. No bad edges in DF, DB:
Either do an r2 to get cases a, b or c; or place bad edges in UF, UR, UB, FR and do
(r U’ r’) U2 (r U’ r)

The terrible 6 Bad edges case (~30%):
Solution 1:
Spam TPS to somehow make a Full or partial staircase and convert to a case of 2 or 4 edges.
Solution 2:
Using an r2, ensure DF and DB are bad, place bad edges in FR, BR, UL, UB and do
r U’ r2 U’ r

Note that the final move in each of these algorithms can be either r or r’; both will always solve EO. In most cases, you should pick the move that brings the D centre to the top.
Note that the second last move in each of these algorithms can either be U or U’; both will always solve EO. In most cases, you should pick the move the attached a D edge to the D centre and/or puts a D edge in UL.

[This categorisation is quite effective to use, especially if you “intuit” the number of misoriented edges; but suppose you made an error and got a wrong intuition; then in the RU gen part of the solve, you would more often than not have 2 misoriented edges. You can fix this by placing the bad edges in UF and UR, and doing r U R’ U’ r’; which almost preserves F2L as well.]

This post should certainly remove the notion that EO of YruRU is “hard” and “inefficient”; this almost algorithmises the step and structurally solves it efficiently. This is best looked at like F2L; with patience it has huge potential.

#### 1cubealot

##### Member
Scramble 3: D2 R2 D2 R2 F’ R2 B’ D2 B F’ D F D’ U2 B D’ U B’ L R

Solution 3: x’ // inspection
R F // 113
U2 F’ U F // CP
U r u R’ U2 R’ u2 // 123
r U2 r’ U’ R r U’ r’ // EO
R’ U’ r2 // 2-gen reduction
After doing the example solves for this method its never solved

#### PapaSmurf

##### Member
There are multiple flaws in this argument; but rather than dissecting, I’ll suggest we wait for some time and let the results by people who’re trying it out speak for themselves.
If you think that this method is better than doing straight up LEOR for TH, go ahead, but from experience, this isn't. You also ignore the simple EOs of F R* F' because they mess up CP. If you want to see all of LEOR's EOs btw, here's a sheet that me and a few other people made.
Anyway, I'm pretty certain that you'll never be able to be faster than the big three with this method, even if CP is super cool, but if you're doing it for fun, carry on.
Wait, so when you do LEOR solves, you plan the whole first block and know where bad edges will be after that?
I wish, but as said, it's the only way it's viable.

#### CuberStache

##### Member
After doing the example solves for this method its never solved
Devagio just cuts off the solve at the point where you'd finish the same as CFOP

#### Devagio

##### Member
Here's some ideas on how to tackle the final stage of YruRU, the RU-gen part.

Stuff that's great:

a. F2L freedom
: Suppose you end up here -
Scramble: R2 U’ B2 R2 D’ F2 D B2 R2 D2 B F2 U L2 R U L’ R’ F2 D
Solution: z // inspection
R S R2 F’ U’ F // CP-line
u’ R u’ R U R u2 // 123
U2 R2 U R’ r U’ r // EO
R2 U' r U2 r' U2 r2 // 2-gen reduction
Of course you could go ahead and put down the white-orange an continue with a nice F2L; but here's another option:
U R U R U'
R' U' R' U R U' R' U
z
And now we have 2GLL, but on orange instead of yellow. If you have been CN on CFOP, this is going to massively help you every 3-4 solves.

b. 2GLL Recognition:
The point of having 2GLL goes far beyond alg reduction and ergonomic turning. Now that you know for certain that corners are going to be permuted, the case recognition is massively simplified. Here's a fact to make you appreciate that - once you know which of the 7 corner orientations it is, you can identify the 2GLL by looking only at the relative colours UFR corner, the UR edge and the UF edge; and that is in fact easier to do than PLL recognition. This should be the way you do 2GLL recognition for most cases as long as you use YruRU. Of course, this won't work with CFOP or ZZ because first you will have to check whether CP is solved to be able to do 2GLL.

c. F2L is extremely easy:
Lookahead is a piece of cake during F2L, and the number of F2L cases is really restricted; most of them are great. This doesn't need more explanation.

Stuff that may not be great but I think it might:

a. Psuedo-block:
Suppose you end up here -
Scramble : D2 F2 R B2 D2 B2 R’ F2 L’ B2 F2 U F D L’ F’ D B’ U L2
Solution: z y // inspection
F U’ f’ U F // 113 + CP
U r U r’ R’ U’ R’ u2 // 123
r U’ R2 U r U’ r // EO
U R’ U’ r2 // 2-gen reduction

You could now continue as follows:
U' R2 U R U' R' U' R U2 R2 U R U R' U R //
U2 R' U R U' R U2 R' // WV
R U' R U R U R U' R' U' R2 // EPLL

This was clearly not one of the best opportunity to do this, but it may so happen that you find a pre-made square; then no matter what square it is, you will be able to utilize it this way if you know WV. Also, You may simply do 2GLL instead of WV+EPLL if you are good at recognition of 2GLL from multiple angles (or you could recognize by flipping colours).

Stuff that may be great but I do not personally recommend:

a. Phasing:
You could do this and reduce the number of 2GLL from 84+4 to 28+2, simply by adding 1-2 moves during your last slot 2/3 of the time. This also increases the chances of an LL skip from 1/324 to 1/108. Also, case recognition gets easier. However, the cases that remain after phasing are kinda bad, the average movecount of this subset is 2 moves more than 2GLL as a whole; also the R U R' insert has to be dealt with slightly painfully is you want to phase it. It may be worth it, especially to begin with; its each person's call.

b. Un-phasing: You could do this and reduce the number of 2GLL from 84+4 to 56+2, simply by adding 1-2 moves during your last slot 1/3 of the time. The subset that remains is better than the average 2GLL algs by over 1 move. It is questionable, but recognition may improve slightly. However, this reduces the chances of an LL skip to 0, and again, the R U R' insert has to be dealt with slightly painfully is you want to un-phase it. It may be worth it, its again each person's call.

Stuff that's not worth it:

a. 1-look-LSLL:
While it may seem practical to do 1-look lsll since it has only ~300 algs, all 2gen; it requires a built pair and the recognition isn't amazing. Doing 2GLL is so much more worth it that this should definitely not be done even if you somehow know all 300 algs.

b. Winter Variation: Unless its a really simple case like R U R' U' R U' R', or if you know for certain that LL is going to skip, this is not very helpful; because you will nonetheless have to do one EPLL alg, which is not much better than 2GLLs expect in recognition.

c. Weird variations like snake, etc.: This is a TPS based step, keep it as simple as possible unless theres a huge advantage.

#### Stryder73

##### Member
What you’re talking about is a method very similar to Petrus, but this one looks a bit better.

#### TheRouxGuy

##### Member
Do you have a pdf explaining the method? Probably including the algs?

#### Devagio

##### Member
What you’re talking about is a method very similar to Petrus, but this one looks a bit better.
Not really...
All method are bound to be similar to some extent.
This is like petrus because there’s a 223; it’s also like Roux because there’s an FB; it’s like CFOP because there’s F2L followed by LL and like ZZ because there’s EO.

Do you have a pdf explaining the method? Probably including the algs?
I do not. Maybe I’ll make one. If someone posts a tutorial of any kind they can put the link or pdf here.
Though, this thread serves the purpose of a tutorial pretty well I guess.

##### Member
I want to adress a thing, about EO at middle of the solve. It is not easy to see which of the 8 edges are oriented and to solve along with forming line. Also, doing u moves is not that easy as you will hold the cube at a very small area.

#### ZB2op

##### Member
I think I'll just stick to ZB

#### 1cubealot

##### Member
Devagio just cuts off the solve at the point where you'd finish the same as CFOP
Oh ok

This method seems good, I may try learning it even though I do CFOP
Now I don't because of CP

#### antonio19kamh3

##### Member
I want to adress a thing, about EO at middle of the solve. It is not easy to see which of the 8 edges are oriented and to solve along with forming line. Also, doing u moves is not that easy as you will hold the cube at a very small area.
Agree with this - it's most likely one of the reasons why Petrus is not very popular

#### ZB2op

##### Member
0 bad edges (1/256): EO skip
So your arguments are mostly based on skips which is luck. When if using just something like regular CFOP it is consistent. Also it has much better look ahead which is the main difference between people like Feliks Zemdegs and people who average around 10 seconds. Look ahead is the reason this method will never be better than roux and CFOP and if you're bored with CFOP just learn full ZB which will keep you occupied for moths at least. Also I practiced this method for about a month and I do better with the beginners method and roux (which I gave up on in about a minute). As someone else said just move on.

Edit:
Didn't mean to sound as harsh as I did..

#### dudefaceguy

##### Member
Didn't mean to sound as harsh as I did..
Yeah I've had a lot of fun investigating CP because of this method. I have concluded that if I'm going to do CP, I will do it in the last layer because that's the only time that recognition is reasonable, and it only saves like 2 moves doing it earlier. But, it was still fun to experiment and learn.