Pinkie pie is slightly better than EOLR

[Spencer131]

This one is for the Roux nerds. I was curious as to whether Pinkie Pie is a worthwhile Roux variant, so I did a Pros & Cons analysis.

By the way, here's an explanation of pinkie pie for those who are unfamiliar:

• Solve first two blocks as normal
• Solve ULUR to the bottom face (centers do not matter here)
• OLLCP (this leaves us at step 4b of the traditional Roux method with ULUR already on bottom)
• 4b+4c

Comparison table

In this comparison, I compare Pinkie Pie to EOLR, since EOLR is currently the best and most popular Roux variant.

Criteria	EOLR	Pinkie Pie
Move count (# of moves needed after blocks)	11 for CMLL + 13.5 for LSE = 24.5	3.7 for setup + 12 for OLLCP + 7 for finish = 22.7
Recognition	CMLL recognition is not great	Can predict corner case during setup, because the setup only uses MU moves
Ergonomics	1 grip switch (from CMLL to LSE)	2 grip switches (one from setup to OLLCP and another from OLLCP to the finish)

Overall

Pinkie pie saves ~1.8 moves and has better recognition. However, it has an extra grip switch as compared to EOLR. Overall this leads me to believe that Pinkie pie is slightly better than EOLR if you are willing to learn 300+ algs for OLLCP

Notes

The move count for pinkie pie setup was found by listing all cases exhaustively. There are 120 cases because the first edge can be in 6 positions, the second edge can be in 5 positions, and each edge has 2 orientations. 6*5*2*2 = 120

The move count for other steps was found by doing a bunch of LSE scrambles. There may be a more objective method, so feel free to prove me wrong on those if you want to compute more thoroughly.

What do you think? Did I miss any criteria?

 

[OreKehStrah]

“CMLL recog is not great” 1, skill issue, 2, OLLCP is literally CMLL recog and U layer recog on top of it lol.

Another thing to note is that the CMLL algs are a lot nicer than OLLCP algs.

People are also learning and using EOLRb so if you want to compare more advanced options that would be the better thing to compare against.


Regardless, I do think it’s underexplored and situationally good.

 

[Spencer131]

“CMLL recog is not great” 1, skill issue, 2, OLLCP is literally CMLL recog and U layer recog on top of it lol.

1. I average ~10 with Roux
2. With pinkie pie, you can predict corners during the setup because MU doesn't affect corners

 

[Swagrid]

. I average ~10 with Roux

I average ~9 with zz but still have rough eocross execution and bad F2L habits, as well as subpar zbll recog. Averaging a lowish number doesn't mean you don't have any skill issues.

 

[artless1der]

haha I've been using "Pinkie Pie" for about 2yrs now and didn't even know it lol
I actually thought I was just being smart making small changes from usu EOLR... can I get co-credit? pffft! jk (somebody oughta ditch that PP for A Good Name lol)

 

[Deleted member 36445]

Intuitive?
Nah
Learning 500+ cases?
ofc!

 

[quanloh]

Intuitive?
Nah
Learning 500+ cases?
ofc!

Nah, 1LLL CFOP is more fun

 

[BradleyGallella]

What do you think? Did I miss any criteria?

When it comes to the average movecount for the setup, I’m going to assume that number is with misoriented centers, but I’m guessing you didn’t account for influencing or even inserting whichever edges you’re placing in DFDB during SB? Also is it taking into account the fact that you can use UF UB edges and not just UL UR?

pinkie pie is almost definitely not all that worth it, unless you just use it situationally, but when I was using it every solve to get practice with it it definitely did not take me an average of 3.7 extra moves for the setup. It was usually like 3 moves max.

 

[Thom S.]

You trade a 1.8 Move advantage against 2 guaranteed Regrips.
Back in 2018(I think, may have been March) I was doing some research on Pinkie Pie and for what I want, the only Setup moves that I think are worth it are the M' U M(and variations) 3-1 Arrow case and doing a D2 so that you don't have to learn all the cases with one Edge at DB.


But in all seriousness, the Majority of solves(for me) end SB with an Insertion, in other Words, R, R', r, r', depending on Center Orientation. So with a little Lookahead, you can already make out 2 Corners, CMLL recognition needs 3. I fail to see where recognition is not that good like you say.

 

[GodCubing]

1. is OLLCP really 12?
2. is finish really 7?
3. I have heard OLLCP algs are bad ergonomically
4. 3.7 for setup? (idk where these movecounts are coming from. i would believe them i just need proof)

 

[BradleyGallella]

1. is OLLCP really 12?
2. is finish really 7?
3. I have heard OLLCP algs are bad ergonomically
4. 3.7 for setup? (idk where these movecounts are coming from. i would believe them i just need proof)

I just ran it, there is a guaranteed move to insert ULUR edges since you set them up on the bottom and add 0.75 for the AUF beforehand, 4c is 5.02 if you account for the guarunteed AUF after inserting ULUR. The move count for the end totals 6.77 however this isn’t counting UFUB cases so it’s likely slightly lower. Unless I’m stupidly missing something.

I also left another comment talking about the setup number already. I also thought OLLCPs averaged about 11 moves, but I haven’t actually looked at mine lmao.

 

[BradleyGallella]

You trade a 1.8 Move advantage against 2 guaranteed Regrips.
Back in 2018(I think, may have been March) I was doing some research on Pinkie Pie and for what I want, the only Setup moves that I think are worth it are the M' U M(and variations) 3-1 Arrow case and doing a D2 so that you don't have to learn all the cases with one Edge at DB.


But in all seriousness, the Majority of solves(for me) end SB with an Insertion, in other Words, R, R', r, r', depending on Center Orientation. So with a little Lookahead, you can already make out 2 Corners, CMLL recognition needs 3. I fail to see where recognition is not that good like you say.

I’m not arguing for Pinkiepie, it’s definitely only worth it situationally cause OLLCPs suck so bad, but you can influence or even in some cases control the setup case during SB, it’s relatively easy actually.

 

[efattah]

I think the ZBRoux finish is better than Pinkie Pie. It is also the only version of the three variants that can be completed in 2 looks.

 

[Celeritardum]

It is also the only version of the three variants that can be completed in 2 looks.

You could learn how CMLL affects permutation and then predict EOLR or EOLRb to get it in 2 looks

 

[efattah]

Really the only criticism or drawback of ZBRoux is (1) some ZBLL algorithms are not that nice or ergonomic, (2) very hard to maintain good speed & memory fetch time for such a large algorithm set. On paper ZBRoux certainly beats regular Roux, its incredible feature is the capacity to recognize the corner case during the pure MU step of solving DF/DB & orienting the edges, greatly reducing the ZBLL recognition time; in fact I would say this makes it the only method that has truly amazing ZBLL recognition time. However the drawbacks of such large algorithm sets are real, and make it fairly impractical for any except pro cubers who have all day every day to train & drill algorithm sets.

I know about 300 algorithms and doing 100 (3x3) solves per day is not even close to enough to maintain my algorithm sets. I'd estimate you need to solve your cube at least 4 times the number of algorithms (i.e. 300 algs = 1200 solves per day). Or just drill each set individually.

 

[OreKehStrah]

Really the only criticism or drawback of ZBRoux is (1) some ZBLL algorithms are not that nice or ergonomic, (2) very hard to maintain good speed & memory fetch time for such a large algorithm set. On paper ZBRoux certainly beats regular Roux, its incredible feature is the capacity to recognize the corner case during the pure MU step of solving DF/DB & orienting the edges, greatly reducing the ZBLL recognition time; in fact I would say this makes it the only method that has truly amazing ZBLL recognition time. However the drawbacks of such large algorithm sets are real, and make it fairly impractical for any except pro cubers who have all day every day to train & drill algorithm sets.

I know about 300 algorithms and doing 100 (3x3) solves per day is not even close to enough to maintain my algorithm sets. I'd estimate you need to solve your cube at least 4 times the number of algorithms (i.e. 300 algs = 1200 solves per day). Or just drill each set individually.

It sounds like you greatly overestimate the difficulty of ZBLL lol.
ZBRoux is bad because you give up the low alg count of traditional Roux and the fact humans are often able to come within 1-2 moves of optimal LSE solutions, while ZBLL is far from that. So you loose out in terms of movecount as well.

 

[efattah]

I would say the movecount of ZBRoux vs. Roux is basically the same except ZBRoux removes the worst part of Roux which is CMLL recognition time. As I mentioned, in ZBRoux you are solving DF/DB & orienting edges, while you are recognizing the corner case, which eliminates the CMLL recognition time; that is a huge advantage.