Is Petrus a dead method? Does the community still think it can be "fast"?

[Escher]

Off-topic and tongue in cheek but here goes:
XFOP any takers?!?
Might please the purists who don't like "Fridrich"

Fridrich is actually more descriptive imo
...Although obviously some people love to make the methods origins political.

The only thing I can imagine being useful and non-restrictive would be to call it 'slots', but that's a dumb name.


On-topic again: I might practice Petrus for a while and see what I can get.

 

[JLarsen]

@ Ellwd when I do Fridrich solves I X cross the vast majority of the time =3 If I were to switch methods, I think that's what I would do. The only reason I don't use Fridrich besides having to learn full OLL is that the cross is SO BORING.

@ Escher what kind of times can you get? I'm going for the brute force turn fast approach and it seems to be working so far. I've been getting mostly high 14 averages which is pretty good for instant improvement imo.

 

[Escher]

@ Escher what kind of times can you get? I'm going for the brute force turn fast approach and it seems to be working so far. I've been getting mostly high 14 averages which is pretty good for instant improvement imo.

I'm terrible at EO - I am good at recognising it but rubbish at solving it efficiently (I should really spend some time reading your tutorial and learning stuff), so my times aren't really reflective of what I could get with practice - about 12-13 if I skip EO, 15-16 if I actually go and do it. My 2x2x2 isn't as efficient as it used to be somehow :/

 

[oll+phase+sync]

The Petrus method has a lot of unique steps, all of wich require experience not brute force ( like "learn 57 OLLs" ).

I also thought it might be possible to get better at inspection time usage (and plan ahead the full 3x3x3 ), but this seems to be to much over the top (at least for me - and the majority of cuber, too)
( I think Roux is currently the winner in the category: inspection time usage )

Using (full) MGLS to simplyfy Step2 (or Step 4) is the most promising way to enhance Petrus, i currenty can think of. And I really would like to see some Petrus records.

But I believe currenty there is no way to be "worldclass" with Petrus.

 

[Godmil]

The Petrus method has a lot of unique steps, all of wich require experience not brute force ( like "learn 57 OLLs" ).

Using (full) MGLS to simplyfy Step2 (or Step 4) is the most promising way to enhance Petrus, i currenty can think of.

Given MGLS has nearly twice as many algs as OLL, would you not call that a 'brute force' step?

 

[JLarsen]

I'd definitely call that a brute force step.

From what I've read in this thread the most likely reasons (to me) that Petrus is slow compared to other methods are;

1. Edges are solved in 2 separate steps
2. Step 4 move count is really high at times

So there's 2 things we can do to solve these problems. We can either eliminate the bad edge + step 4 approach, OR, we can make the EO step worth doing by combining some of the other steps together. (MGLS, WV, and ZBLL come to mind.)

Noahaha recently came up with a "SuneLS" which is a last slot system meant to force Sune ZBLL cases the majority of the time. If somebody can come up with or justify another approach with EO, LL or something like that I'll go ahead and learn it. The only reason I haven't started to learn Sune LS yet is because it didn't seem to receive all that great of feedback.

 

[DavidWoner]

I'll just throw everything I have into TPS. Assuming my move counts are fine. Is 60 alright for a speedsolve?

No. 55-60 is an alright average for CFOP, but a Petrus movecount should definitely be less. 50, maybe. Without a 5+ move advantage, you can't even begin to argue that Petrus is in the same league as CFOP. Working on TPS is definitely important, but you're adding moves somewhere and need to stop.

I'd definitely call that a brute force step.

From what I've read in this thread the most likely reasons (to me) that Petrus is slow compared to other methods are;

1. Edges are solved in 2 separate steps
2. Step 4 move count is really high at times

So there's 2 things we can do to solve these problems. We can either eliminate the bad edge + step 4 approach, OR, we can make the EO step worth doing by combining some of the other steps together. (MGLS, WV, and ZBLL come to mind.)

Noahaha recently came up with a "SuneLS" which is a last slot system meant to force Sune ZBLL cases the majority of the time. If somebody can come up with or justify another approach with EO, LL or something like that I'll go ahead and learn it. The only reason I haven't started to learn Sune LS yet is because it didn't seem to receive all that great of feedback.

A couple things here:
Why would you force Sune ZBLL? Sune/Antisune is the worst set. T/U are much easier to recognize, and H has a smaller number of cases. Since the T/U sets have a lot of inverse pairs, you could easily learn those two and then just force two adjacent twisted corners. That would be much easier than learning to force a single set.

As for your reasons for Petrus's relative slowness, you left out the biggest issue - ergonomics. Building a 2x2x2 can be roughly equated with building a cross or a 1x2x3, as far as movecount and speed are concerned. However, 2x2x3 will likely require some rotations or some nasty moves, whereas CFOP pairs are generally nice, and second roux block is <rRU>. Petrus EO with sometimes require a rotation and are 3+gen, Roux EO is MU with no rotations. COLL+EPLL is more moves than OLL+PLL and generally quite a few COLL cases suck, whereas there are good algs for pretty much every OLL/PLL by now. Also, I think OLLCP and KCLL are more viable and useful advanced methods as compared to ZBLL.

I'm just not feeling optimistic for Petrus as a speedsolving method. It is pretty good for starts in FMC though.

 

[irontwig]

imho the hardest step in Petrus is the 2x2x3 expansion, which is why I only block build after the 2x2x2 if I have a free pair, otherwise I just complete the X-cross. With the EO you don't really save any moves on the F2L, and on the LL really only if you know ZBLL. So basically it's a monster method if you can plan out 2x2x3s and know ZBLL, especially for slower turning events like OH and Feet (or linear FMC). And yeah, for FMC 2x2x3 kick ass, but most of the time it's just better to get on with solving pieces than orienting edges.

 

[Kirjava]

Why would you force Sune ZBLL?

Because it's different.

 

[oll+phase+sync]

I
1. Edges are solved in 2 separate steps

What do you mean by this ?

2. Step 4 move count is really high at times

I think the optimal average for this step is 11 moves, but I don't see how a human can aproche this. (even slanting opportunities are too rare )
on the other hand I don't think step 4 is worse than standard F2L.


Regarding MGLS - for sure it is a brute force step, but that's not necessarily bad (like OLL)

also:
- all the algs can be done 2-gen
- mirror ratio is very high (OLL mirror ratio is very small )
- recognition is easy too (like OLL)
- still less algs than learning Sommer + Winter for FR and BR slot.

What really holds me back from even starting MGLS is that there is no solid 2 step approach ( If you don't know an OLL just do EO and CO seperate ) (not even an unsolid approach I know of)

 

[Escher]

What really holds me back from even starting MGLS is that there is no solid 2 step approach ( If you don't know an OLL just do EO and CO seperate ) (not even an unsolid approach I know of)

Just begin by learning all the one/two corners oriented cases (not that many), and then using sunes to reduce the 3/4 mis-oriented corners cases.

I learned the I and Im cases in about 15 minutes - it's really not that hard. I should finish off O and +/- sometime :/

 

[JLarsen]

Because it's different.

Lol. Well, I didn't come up with the idea, and I've never been good at cube theory. I really don't know anything about ZBLL or alternative LL methods, so your guess is probably better than mine.

What do you mean by this ?

It was something QQ mentioned and it made a lot of sense to me. It's starting not to make as much sense now lol but I took it to mean that we have two separate orientation steps; one for edges, and one for corners. That's not as efficient as solving LL in 2 steps?


@ Woner You made some good points. Thanks for the advice. I'll have to work on my move counts as well then. And like I said before. I'm no pro with Petrus. 15 seconds is decent but there's going to be lots of flaws in my solves still. I'm just trying to work them out.

 

[Sillas]

This thread is serious?
Petrus' Method is 'the father' of all block building method. It's also very efficient for Fewest Moves.
It maybe not can be fast now, but still very helpful and and definitely isn't a dead method.

 

[JLarsen]

This thread is serious?
Petrus' Method is 'the father' of all block building method. It's also very efficient for Fewest Moves.
It maybe not can be fast now, but still very helpful and and definitely isn't a dead method.

I really meant dead as a speedsolving method. Hindsight is always 20/20 but I should've probably left out the first part of the thread title.

Update on move counts; I just did some slow solves that were slightly more efficient than usual, some even quite lucky, and the lowest movecount I got was 55. I'm thinking that 60 really isn't as unreasonable for a Petrus speedsolve as Woner said. It might not be the greatest, but I don't think 50 is really practical as a goal. At least not in a speedsolve.

 

[Rpotts]

55 is very high for your lowest move count in slow solves. I do freefop slow solves everyday at work, counting moves. Mid 40s are common, 50s is average, and this is for some pretty cfop-y freefop.

 

[jrb]

Petrus could definitely be as fast as CFOP or Roux(IMO). I think the reason no one has gotten as fast times as with CFOP or Roux with Petrus is that not many people use Petrus.

 

[Kyle™]

I think I average 50-60 move count for my solves. My lowest move count is with my last slot variation, where I average 49 moves per solve. ( Taken from ao12 ). There are so many difficult cases with that step so I don't use that variation too often in real solves.

 

[JLarsen]

After some discussion in my example solve thread I've decided to try X cross for a while. Page 3 is really all you need to read for an explanation. I realized that while my 2x2 solutions are good, it's too difficult to take advantage of the freedom of 2x2x3 in a speedsolve. The result is ridiculous movecounts.