• 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 35,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

Random thoughts on improving/evolving Roux for 3x3 2H that may or may not suck

ottozing

Platinum Member
Joined
Aug 13, 2011
Messages
3,206
Location
Canberra, Australia
WCA
2012MCNE01
YouTube
ottozing
Over the past year or so I've posted a lot of content to my YT channel & Newsletter about ways to improve many of the steps of CFOP. However, I've also spent at least a small % of my time thinking about Roux (no plans to switch or even try to get fast as a side project) after seeing certain holes and gaps in knowledge that some Roux solvers seem to have (for example not knowing certain solutions like R' F R F' R' U' R which are useful for SB, even though they break the "rule" of only doing RrMU for SB, which is almost definitely a stupid rule)

In my opinion, it's pretty hard to improve Roux simply because CMLL+LSE as it is (using EO+ULUR/UFUB and also ULUR/UFUB+L4E tricks) seems to be optimal as far as speed is concerned due to how close the movecounts are to optimal solutions. I know Kirjava in the past has said that in order to improve Roux it would have to be something related to making the first two blocks better in some way, and for the most part I agree

I don't know how easy or feasible this is, but I think if Roux solvers were able to see left 1x2x3 + right 1x2x2 in inspection every time, CFOP would have an incredibly hard time keeping up long term! I also can't help but wonder if doing 1x2x2 on left & right and then finish both 1x2x3's could be used more than we currently see, or any other ways of ordering F2B for that matter

One idea that I've had for FreeFOP is doing your almost cross+blocks solution, but also making sure that the edge(s) that are occupying the spaces where the cross edges *should* be are filled with LL edges, so that you can do F2L like normal without getting stuck & generally have a better chance of combining an F2L pair with a cross edge. Maybe applying this to Roux but with L6E edges during "weird" F2B scenarios would be nice, especially since you have more edges to choose from, only have to "fill" FD and BD, and also you don't have to place them around any center (you also don't have to "fix" these placed edges later on, like you would with my FreeFOP idea)

Other than Roux solvers inspecting more than they do currently & making F2B more flexible using the above tactics, I don't think there's much else that can be done that isn't already kind of obvious (like doing "PSF2L" with niche 2nd block cases such as D' R U R' U R U' R' D), so let's get into some CMLL+L6E ideas that I have that may or may not suck

Alex Lau aka 5BLD aka the first person to put Roux on the map in a big way (and generally speaking the most based Roux solver to ever do it), towards the end of his career mentioned a method called "Pinky Pie" that pretty much entailed doing F2B + placing ULUR in DFDB oriented & around the U or D center, doing CMLL, and then a really easy L6E case

The biggest downside to this I think is the fact that OLLCP requires learning like 300 algs, and Roux solvers tend to choose Roux because they hate learning algs, so this will be a hurdle for many who attempt this

Still, I think Pinky Pie might be more promising than we once thought (especially if OLLCP gets improved, which I think it can greatly)

I also think you don't necessarily have to limit yourself to placing ULUR in DFDB, and could do UFUB or even DFDB and still get great L6E cases using OLLCP instead of CMLL (though DFDB might be iffy, idk enough about L6E cases with DFDB placed at the bottom but I feel like cases where they're placed around the D center might be better with OLL & PLL lol)

Even if it's limited to ULUR or UFUB in DFDB, I don't think it would be that difficult to place both edges alongside a 2nd block solution. Right now the best Roux solvers seem to inspect FB+DR edge, and I think throwing at least one (if not both) U layer edge(s) around U/D center and placing in D wouldn't be much harder. It might be best to do these edges after the right 1x2x2 or even 1x2x3 in some cases as far as move count is concerned, but I also worry about the recognition at that point (again, someone who actually uses Roux will have to explore this and tell me if my ideas suck, which they might)

Random ugly example solve with Pinky Pie as I've just described it

L U R' F2 R2 F' U' R D L2 F L2 U2 B' D2 R2 U2 F L2 F' R2

x2 y'
R2' U' R D r' D (FB)
r U' r R U' r' U r (DR+"Pinky Pie stuff")
U R U' R' U' R U2' R' U R' U2' R U' R' U R (SB with yikes level efficiency)
L F' L' U' L' U L2 F L2' U L U' L' U' L (OLLCP mirrored to preserve back block thing)
U' M2' U2' (L6E)

I think placing the Pinky Pie edges too early might lead to really inefficient SB cases, but at the same time I don't think it's great to do them during SB either. I think doing them after SB might be ok, so let's try that on the same scramble

L U R' F2 R2 F' U' R D L2 F L2 U2 B' D2 R2 U2 F L2 F' R2

x2 y'
R2' U' R D r' D (FB)
U' R2' U r U' r' R' U' R' U2' R U R' U' R (SB)
U' M' U2 M (Pinky Pie)
U' R U2' R' U' R U' x' U2 L' U' L U L' U' L U' l' (yeah idk this case, or it needs genning)
M2' U M U2 M (L6E)
 

GuRoux

Member
Joined
May 6, 2013
Messages
1,714
Location
San Diego, California
WCA
2014TANG03
YouTube
franktangtartharakul
I never really liked the idea of pinkie pie. eolr is like 8 moves? (I don't really know). the ul/ur set up for pinkie along with harder recognition and worse corners alg makes it seem like a small improvement at best (not worth the 300 algs).

I definitely almost never break the second block "rules" and think that's a place many rouxers can get better at because I've seen many cfop videos with really nice solutions for solving one or two pairs at once.
 

OreKehStrah

Member
Joined
May 24, 2019
Messages
125
I never really liked the idea of pinkie pie. eolr is like 8 moves? (I don't really know). the ul/ur set up for pinkie along with harder recognition and worse corners alg makes it seem like a small improvement at best (not worth the 300 algs).

I definitely almost never break the second block "rules" and think that's a place many rouxers can get better at because I've seen many cfop videos with really nice solutions for solving one or two pairs at once.
I’m not super experienced with Roux so I want to ask your opinion on a variation of Pinkie Pie. Do you think there’s any potential to solving the front pair of the SB last, doing a WV alg to place the pair in and just orient corners, and then do an OLLCP alg for one of i think four different OLL shapes. That way the horrible issue of learning ~300 algs for marginal benefit is negated? I’m just a cuber super interested in methodology
 

ottozing

Platinum Member
Joined
Aug 13, 2011
Messages
3,206
Location
Canberra, Australia
WCA
2012MCNE01
YouTube
ottozing
I never really liked the idea of pinkie pie. eolr is like 8 moves? (I don't really know). the ul/ur set up for pinkie along with harder recognition and worse corners alg makes it seem like a small improvement at best (not worth the 300 algs).

I definitely almost never break the second block "rules" and think that's a place many rouxers can get better at because I've seen many cfop videos with really nice solutions for solving one or two pairs at once.
Damn, I knew EOLR/EOFB was good but yeah, 8 moves means 2 edges + OLLCP almost definitely isn't worth it

I'd be down to help flesh out certain SB scenarios with DR solved, since this seems like that's where the most improvement can be made (though I think a lot of these could be figured out if Rouxers just studied good CFOP solvers, haha)
 

GuRoux

Member
Joined
May 6, 2013
Messages
1,714
Location
San Diego, California
WCA
2014TANG03
YouTube
franktangtartharakul
I’m not super experienced with Roux so I want to ask your opinion on a variation of Pinkie Pie. Do you think there’s any potential to solving the front pair of the SB last, doing a WV alg to place the pair in and just orient corners, and then do an OLLCP alg for one of i think four different OLL shapes. That way the horrible issue of learning ~300 algs for marginal benefit is negated? I’m just a cuber super interested in methodology
so instead of [last pair + ollcp] we do [WV+ollcp(less cases)]. It doesn't seem good unless, for some reason, corners oriented ollcp just happens to be way better than the rest of the set. Doing WV is just adding extra recognition and extra moves compared to the original pinkie pie.
 

OreKehStrah

Member
Joined
May 24, 2019
Messages
125
so instead of [last pair + ollcp] we do [WV+ollcp(less cases)]. It doesn't seem good unless, for some reason, corners oriented ollcp just happens to be way better than the rest of the set. Doing WV is just adding extra recognition and extra moves compared to the original pinkie pie.
hmm okay. I didn’t think it would very good either
 

GuRoux

Member
Joined
May 6, 2013
Messages
1,714
Location
San Diego, California
WCA
2014TANG03
YouTube
franktangtartharakul
Damn, I knew EOLR/EOFB was good but yeah, 8 moves means 2 edges + OLLCP almost definitely isn't worth it

I'd be down to help flesh out certain SB scenarios with DR solved, since this seems like that's where the most improvement can be made (though I think a lot of these could be figured out if Rouxers just studied good CFOP solvers, haha)
that would be great since rouxers usually solve DR first for second block. Is there a place with a whole bunch of multi slot tricks compiled? or is it really just looking at lots of cfop solves?
 

KubkaNoobka

Member
Joined
Aug 30, 2019
Messages
2
I think this thread could be also used to discuss and potentially improve the advanced Roux tricks from this Kian's video:
Me personally I use the misoriented centres quite alot and I was experimenting with doing UF and UB edges first instead of UL and UR.
 

ottozing

Platinum Member
Joined
Aug 13, 2011
Messages
3,206
Location
Canberra, Australia
WCA
2012MCNE01
YouTube
ottozing
that would be great since rouxers usually solve DR first for second block. Is there a place with a whole bunch of multi slot tricks compiled? or is it really just looking at lots of cfop solves?
tbh I don't really think in terms of "multislotting" but rather in terms of knowing multiple ways to solve individual cases, and then choosing ones that force better cases for the next pair

could probably throw PSF2L on top of that if you want but i kinda doubt it's good for roux because it removes M slice possibilities

I will say that the "worse" an individual pair case is, the more viable multislotting options you're gonna find (so focus on finding hella ways to solve bad pairs first and work from there)

also I'm starting to think inspecting FB+DR+orient and place SB edges in U layer might be the best way to inspect going forward based on Sean's inefficient but fast solutions :p
 

Spencer131

Member
Joined
Feb 16, 2016
Messages
59
For the specific pair jay mentioned: R' F R F' R' U' R
I actually do R U' R' U S R' S' but I agree with the principle. We should be more willing to use solutions that "break" RrUM. Similarly, we could probably use some edge comms in lse, although it will only be beneficial for a couple eolr and eolrb cases.

According to this link, eolr with misoriented centers is 8.28 moves on average, so pinkie pie is only an improvement if you can get ULUR on bottom in a couple moves. It's difficult to compare because there are a lot of factors: a couple extra moves with MU moveset to set up pinkie pie, slight regrip into OLLCP, slightly slower alg because of OLLCP, and another slight regrip back into MU for lse.
Btw, you don't even have to have them joined with a U/D center. It can be any center, and if it happens to be F/B, you just do misoriented 4c.

As for the FB+DR+orient SB edges, I have thought about it and it seems dope, especially for oh and feet.
 

ottozing

Platinum Member
Joined
Aug 13, 2011
Messages
3,206
Location
Canberra, Australia
WCA
2012MCNE01
YouTube
ottozing
Another idea that I think could be worth exploring for improving Roux's inspection would be FB+DR+misorient SB corners & place in U layer, since having the D sticker of the corners facing up seems to lead to most of the worst cases. This would lead to a more "efficiency focused" SB finish instead of a TPS focused finish that would come with FB+DR+orient SB edges

Also, does any resource exist for solving all of the last pair cases once you've already solved FB+1x2x2 on the right? If not, it definitely should (I'm genuinely surprised it's not on algdb)
 

GuRoux

Member
Joined
May 6, 2013
Messages
1,714
Location
San Diego, California
WCA
2014TANG03
YouTube
franktangtartharakul
Another idea that I think could be worth exploring for improving Roux's inspection would be FB+DR+misorient SB corners & place in U layer, since having the D sticker of the corners facing up seems to lead to most of the worst cases. This would lead to a more "efficiency focused" SB finish instead of a TPS focused finish that would come with FB+DR+orient SB edges

Also, does any resource exist for solving all of the last pair cases once you've already solved FB+1x2x2 on the right? If not, it definitely should (I'm genuinely surprised it's not on algdb)
There is a resource for that somewhere in the roux method speed solvers facebook group.

I think solving DR+corner might be good as well. Since solving that pair (R U' M' U R' -like solutions), cmll does not change.
 

Sue Doenim

Member
Joined
Nov 9, 2016
Messages
446
Another idea that I think could be worth exploring for improving Roux's inspection would be FB+DR+misorient SB corners & place in U layer, since having the D sticker of the corners facing up seems to lead to most of the worst cases. This would lead to a more "efficiency focused" SB finish instead of a TPS focused finish that would come with FB+DR+orient SB edges

Also, does any resource exist for solving all of the last pair cases once you've already solved FB+1x2x2 on the right? If not, it definitely should (I'm genuinely surprised it's not on algdb)
Kian has his here.
 

efattah

Member
Joined
Feb 14, 2016
Messages
519
In the new method thread we discussed my favorite Roux variant, which is to leave the FR slot unsolved, then pre-locate the FR edge such that it will be auto-solved by the L5C algorithm. In this fashion you basically skip the entire FR slot. The drawback is that you can't use multiple CMLL's to prevent 6-flip scenarios.
 
Top