# What's the fastest method for 3x3?

#### Gaétan Guimond

##### Member
Don't ignore a funny April Fools joke.

Also, hasn't it been established that methods don't have speeds and one shouldn't be limited to learning just the "good" methods?
Learn a lot of them, then pick your favorite, or the one you think suits you.
So you forgot ZZ, Petrus, Waterman, Heise, Triangular Francisco,...
The biggest no name too The cube was asleep none media Fridrich too when I arrived

Last edited by a moderator:

#### JKNK

##### Member
Wow. Nice video, Maybe elaborate on how it is learned and progressed on, and how CFOP has so many cases to learn etc.
thanks, im going to talk about difficulty to learn and who it fits (etc roux fits blockbuilders and zz fits creative and fast thinking people)

#### bpwhiteout

##### Member
CFOP isn't necessarily the "fastest", but I would recommend it purely because it is the most common method, and therefore is has the most easily accessible knowledge. It's also fairly well open to other techniques. Using CFOP wont make you faster though. Or any method for that matter. Practice is what makes you faster. 200 Solves a day for 2 or 3 months might land you at sub 30. Just practice.

#### TDM

##### Member
I think the perfect method has:
1. Great Recognition
2. Great Fingertricks
But what about movecount? And number of algs?
Also, that means that people have been cubing for 40 years and not seen the perfect method yet. I'id have thought the perfect method would be obvious. I don't think any more very good methods are going to be discovered, and if they are it's very unlikely we'll get better than Roux, CFOP or ZZ, all of which don't cover all 3 of your points (Roux's lookahead is harder with the DB edge never solved until the end, and the other two's fingertricks aren't great with ZZ having the transition between L and R, and CFOP having both that and rotations (although because you can often do both at the same time, people ignore that you have to transition between L and R because you can rotate at the same time. Imo it can't really be called a disadvantage of ZZ compared to CFOP as the problem happens in both methods)).

#### Stefan

##### Member
Not true bro. You can't time methods.

I left my cube like this for the past 14 hours, and it's still not freaking solved.
Just because you don't see a way and did something useless doesn't mean there is no way.

Some method speed definitions:
- Fastest official average-of-5.
- Fastest average-of-100.
- Have 100 kids learn and practice the method for 10000 hours, then have each kid do 100 solves and define the method speed as an average of those 10000 solves.
- Average solve move count (like I did in my diploma thesis, though ideally using all possible scrambles).
- Average solve time using more sophisticated models (including grips like Gripper and recognition).

#### mDiPalma

##### Member
Just because you don't see a way and did something useless doesn't mean there is no way.

Some method speed definitions:
- Fastest official average-of-5.
- Fastest average-of-100.
- Have 100 kids learn and practice the method for 10000 hours, then have each kid do 100 solves and define the method speed as an average of those 10000 solves.
- Average solve move count (like I did in my diploma thesis, though ideally using all possible scrambles).
- Average solve time using more sophisticated models (including grips like Gripper and recognition).
-Not timing the method
-Not timing the method
-Not timing the method
-Not even timing
-Not timing the method

soz bro, looks like you are still DUNKED.

#### Stefan

##### Member
-Not timing the method
-Not timing the method
-Not timing the method
-Not even timing
-Not timing the method
- I disagree. It is a way to attribute a time to the method.
- I disagree. It is a way to attribute a time to the method.
- I disagree. It is a way to attribute a time to the method.
- Speed doesn't necessarily involve time.
- I disagree. It is a way to attribute a time to the method.

#### mDiPalma

##### Member
- I disagree. It is a way to attribute a time to the method.
- I disagree. It is a way to attribute a time to the method.
- I disagree. It is a way to attribute a time to the method.
- Speed doesn't necessarily involve time.
- I disagree. It is a way to attribute a time to the method.
Attributing time to a method is like attributing taste to geometric shapes.

It's not possible.

Methods are concepts. You can only attribute time/speed to the application of a method on a particular scramble.

In that situation, it is the METHOD APPLICATION that has a time. Not the method. There are trillions of ongoing external factors at place in every timed solve that are not relevant to the method being applied. Any speed assignment to methods from experimental means is largely misdirected.

U can argue all u want pl0x. At the end of the day, ur still gonna get dunked.

EDIT:

oh and by the way:

"Speed doesn't necessarily involve time."

Last edited:

#### GuRoux

##### Member
Speed must be used to evaluate which method is better. I think what stefan proposed is in the right direction and will at least tell us something insightful. Of course there must be analysis on the qualities of each method and how those qualities effect speed, but in the end, the quantitative evidence will hold unless there is a real argument to why the results turned out as they did. If the experiment is controlled, there shouldn't be much external factors that would greatly alter the results.

#### Stefan

##### Member
Attributing time to a method is like attributing taste to geometric shapes.

It's not possible.
Interesting, given that I just did.

Congratulations, you found a definition of speed that does involve time.

Here's another: My speed of realizing your close-mindedness. Was rather quick, only took me one post of yours. Notice that "one post" is not a measurement of time.

#### mDiPalma

##### Member
Speed must be used to evaluate which method is better. I think what stefan proposed is in the right direction and will at least tell us something insightful. Of course there must be analysis on the qualities of each method and how those qualities effect speed, but in the end, the quantitative evidence will hold unless there is a real argument to why the results turned out as they did. If the experiment is controlled, there shouldn't be much external factors that would greatly alter the results.
Even a controlled experiment contains near-infinite quantities of microscopic, atomic, subatomic, and spiritual factors (among others) with nonzero effects on the "timed" outcome of a 3x3 solve.

dunked?

Congratulations, you found a definition of speed that does involve time.

Here's another: My speed of realizing your close-mindedness. Was rather quick, only took me one post of yours. Notice that "one post" is not a measurement of time.
That's not speed, and you know it. It's some quantity.

Speed has units of SOMETHING PER UNIT TIME. It's defined that way.

It's common for those expecting to lose debates to withdraw with some aggressive comment. If you want to continue this hardly-productive and potentially abusive conversation, PM me.

Last edited:

#### Stefan

##### Member
I think what stefan proposed is in the right direction and will at least tell us something insightful.
Yeah, that's the goal. I don't like dunking my head in the sand, I prefer trying to do something useful if I can.

That's not speed, and you know it.
I disagree, and nope I don't.

Speed has units of SOMETHING PER UNIT TIME. It's defined that way.
Nope. Even the wikipedia page you showed disagrees (check its "Units" section).

#### GuRoux

##### Member
Even a controlled experiment contains near-infinite quantities of microscopic, atomic, subatomic, and spiritual factors (among others) with nonzero effects on the "timed" outcome of a 3x3 solve.

dunked?

That's not speed, and you know it. It's some quantity.

Speed has units of SOMETHING PER UNIT TIME. It's defined that way.

It's common for those expecting to lose debates to withdraw with some aggressive comment. If you want to continue this hardly-productive and potentially abusive conversation, PM me.
Sure there are near infinite number of factors (depending on how you count them) but they are still not going to be too significant over the course of many solves and many days. There is a very noticeable difference between sub 17 and sub 15 solve. External factors that are unable to be avoided will not make a sub 15 solver perform worse than a sub 17 solver. Probably anything under sub 15 can safely differ skill 1 second away from each other, maybe .5. If there is anything within that range, then there would of course need much more trials to have accurate results.

And just out of curiosity, what do you mean speed without time?

#### mDiPalma

##### Member
Nope. Even the wikipedia page you showed disagrees (check its "Units" section).
Are you blind? Or just trolling?

All of those units are in distance/time (or some are nondimensionalized by dividing by some known speed at those conditions [sound, light]). Either way, "one post" is not a speed, and you are fully aware. If it takes you "one post" to learn something, then the quantity of ONE POST has been READ for you to attain some understanding. There is no distance. There is no speed. And there is no nondimensionalization. (my major is relevant to Newtonian Dynamics, and this is a concept that even my 3 yr old cousin can understand).

Stefan, please don't read this spoiler if you have self-esteem problems or depression:

I now recognize why a vast quantity of members of the speedsolving community profess negative opinions about you. You feign some coolness by writing succinct inflammatory posts, and when confronted, you hide behind empty claims and reclaims of the same points, without offering any new support to your argument. Sometimes you are right, but often you are just posting to be mean or to make someone feel bad for either no reason or some selfish reason. When proven factually incorrect, you simply disagree. Either you are trolling, or you are acting really immature. I don't intend to ever respond to you from now on. Mods, you can remove this and temp-ban me if you want. I know Stefan's already seen it. And that's all that matters to me. Peace.

Sure there are near infinite number of factors (depending on how you count them) but they are still not going to be too significant over the course of many solves and many days. There is a very noticeable difference between sub 17 and sub 15 solve. External factors that are unable to be avoided will not make a sub 15 solver perform worse than a sub 17 solver. Probably anything under sub 15 can safely differ skill 1 second away from each other, maybe .5. If there is anything within that range, then there would of course need much more trials to have accurate results.

And just out of curiosity, what do you mean speed without time?
Well, the best (and perhaps the only) way you can compare the optimal time intervals for different solving methods would be with some Normal distribution. As you move the humps about, assigning appropriate mean and deviation values, there will always be some infinitely small overlap of the Petrus hump with the Roux hump. It could be .1% or it could be 50%. Either way, even if you could perfectly assign times to methods (which I'm arguing is impossible), there would be no definitive statement that one method is better, because there would be some solve, the Normal model suggests, that would be faster for PCMS than for CFOP.

I don't know what you are saying "Speed without Time." I am just saying that timing 100 guys solving with Petrus does not give a SPEED to the METHOD. But rather, it assigns a TIME to the APPLICATION of the method. There are two inherent differences, and those are the flaws in Stefan's argument.

#### GuRoux

##### Member
Are you blind? Or just trolling?

All of those units are in distance/time (or some are nondimensionalized by dividing by some known speed at those conditions [sound, light]). Either way, "one post" is not a speed, and you are fully aware. If it takes you "one post" to learn something, then the quantity of ONE POST has been READ for you to attain some understanding. There is no distance. There is no speed. And there is no nondimensionalization. (my major is relevant to Newtonian Dynamics, and this is a concept that even my 3 yr old cousin can understand).

Stefan, please don't read this spoiler if you have self-esteem problems or depression:

I now recognize why a vast quantity of members of the speedsolving community profess negative opinions about you. You feign some coolness by writing succinct inflammatory posts, and when confronted, you hide behind empty claims and reclaims of the same points, without offering any new support to your argument. Sometimes you are right, but often you are just posting to be mean or to make someone feel bad for either no reason or some selfish reason. When proven factually incorrect, you simply disagree. Either you are trolling, or you are acting really immature. I don't intend to ever respond to you from now on. Mods, you can remove this and temp-ban me if you want. I know Stefan's already seen it. And that's all that matters to me. Peace.

Well, the best (and perhaps the only) way you can compare the optimal time intervals for different solving methods would be with some Normal distribution. As you move the humps about, assigning appropriate mean and deviation values, there will always be some infinitely small overlap of the Petrus hump with the Roux hump. It could be .1% or it could be 50%. Either way, even if you could perfectly assign times to methods (which I'm arguing is impossible), there would be no definitive statement that one method is better, because there would be some solve, the Normal model suggests, that would be faster for PCMS than for CFOP.

I don't know what you are saying "Speed without Time." I am just saying that timing 100 guys solving with Petrus does not give a SPEED to the METHOD. But rather, it assigns a TIME to the APPLICATION of the method. There are two inherent differences, and those are the flaws in Stefan's argument.
I think your normal distribution curve may give some insight on how "good" a method is but , like you said, not definite. But it might have some telling results of how close in "goodness" methods are to each other. In an idealized world, we could have a large number of people who don't know how to solve a rubiks cube and have them train for similar amounts of time; then we compare them. But I guess that can't happen in a large enough scale to create a "definite" result, but perhaps close.

#### Phillip1847

##### Member
Yeah, that's the goal. I don't like dunking my head in the sand, I prefer trying to do something useful if I can.
10/10 for pun

Some methods are faster than others, intrinsically. LBL will never be as easy to get fast with as Roux or CFOP is.
However, among the top methods, I think all have potential to be equally as fast as one another. I'm talking mostly about Roux, CFOP, and ZZ.
Some of these haven't been realized because a. faz doesn't use them b. a lot of people use CFOP.

#### mDiPalma

##### Member
I think your normal distribution curve may give some insight on how "good" a method is but , like you said, not definite. But it might have some telling results of how close in "goodness" methods are to each other. In an idealized world, we could have a large number of people who don't know how to solve a rubiks cube and have them train for similar amounts of time; then we compare them. But I guess that can't happen in a large enough scale to create a "definite" result, but perhaps close.
If I give you and your clone from an alternate (but, up to this point in time, identical) universe the same exact scramble and THEN tell both of you to solve it, you will not get exactly the same time, even though everything seems the same. This difference will be massively magnified when comparing different people with different backgrounds, in different environments.

#### GuRoux

##### Member
If I give you and your clone from an alternate (but, up to this point in time, identical) universe the same exact scramble and THEN tell both of you to solve it, you will not get exactly the same time, even though everything seems the same. This difference will be massively magnified when comparing different people with different backgrounds, in different environments.
You might actually get the exact same solve according to physical determinism, that a certain initial condition will always follow the laws of physics; thus, lead to the same ending condition. Basically, if you have the same causes, you have the same effects.

#### mDiPalma

##### Member
You might actually get the exact same solve according to physical determinism, that a certain initial condition will always follow the laws of physics; thus, lead to the same ending condition. Basically, if you have the same causes, you have the same effects.
But there is a certain level of subatomic randomness that occurs at every infinitely small time step. Simply assuming that you start at the same state doesn't mean that the states will be the same after even 1 time step into the future.

You are right though; the times that you and your clone get will be nearly identical. To the accuracy of a timer, they will be the same, but in reality they will be slightly different. And that slight difference is what magnifies as you change environments, solver, solver background, and the numerous other legitimate factors that affect speedcubing times.

#### uberCuber

##### Member
Some method speed definitions:
- Fastest official average-of-5.
- Fastest average-of-100.
- Have 100 kids learn and practice the method for 10000 hours, then have each kid do 100 solves and define the method speed as an average of those 10000 solves.
- Average solve move count (like I did in my diploma thesis, though ideally using all possible scrambles).
- Average solve time using more sophisticated models (including grips like Gripper and recognition).
1. Based on your first definition, the speed of the CFOP method is 6.54. That is the method's speed (that number is a time, by the way, not a speed). What if Feliks breaks this world record at a future comp? Then by your definition, CFOP's speed would then be lower than before. But the speed of CFOP is 6.54! Without changing the method at all, how could its speed possibly change? Answer: it wouldn't. The only thing changing is Feliks' proficiency with the method. If a method itself actually has a speed, then that speed is a property of the method. If you don't change anything about a method, one particular property of it can't spontaneously change.

2. Based on your second definition, the speed of the Roux method is 7.00. That is the method's speed (that number is a time, by the way, not a speed). What if Alex starts practicing hardcore again and breaks this average? Then by your definition, Roux's speed would then be lower than before. But the speed of Roux is 7.00! Without changing the method at all, how could its speed possibly change? Answer: it wouldn't. The only thing changing is Alex's proficiency with the method. If a method itself actually has a speed, then that speed is a property of the method. If you don't change anything about a method, one particular property of it can't spontaneously change.

3. Say this experiment was carried out for CFOP, and the average ended up being 7.00 flat, with the fastest person's avg100 being 6.66 (made-up numbers, no statistical calculations made here). By your third definition, the speed of CFOP is 7.00. That is the method's speed (that number is a time, by the way, not a speed). But the fastest person in that group of 100 actually averages faster than that. How is it possible to use a method and be faster than what the method's speed is? How can you say that "the speed of the CFOP method is 7.00" if there is actually a person averaging below 7.00 over a whole lot of solves?

4. Okay, hold up. Your first definition states that CFOP is faster than all other methods. Your second definition states that Roux is faster than all other methods. Your fourth definition says that, for example, Heise is faster than both CFOP and Roux. But these statements are all mutually exclusive. It's also pretty much guaranteed that the kind of model proposed in your fifth definition would declare a different "Fastest" method than your fourth definition does, since an extremely low movecount method will inherently have bad grips (as compared to CFOP/Roux) and recognition. And if this sophisticated model did in fact declare the same "Fastest" method as any of your other four definitions, I strongly doubt it would also come up with the same specific number as did the other definition. Therefore, all five of your definitions for the speed of a method are mutually exclusive. So which, if any, is actually correct? (The answer of course, is none, since they aren't even giving speeds, but rather other quantities such as time or number of moves, but moving on).

A method does not have a speed as an inherent property of it. Something like the kind of model mentioned in your fifth definition would indeed make it possible to compare two methods to each other and say definitively "the CFOP method is better for human speedcubing than the Dan Brown method," but it would not let you say "the speed of the CFOP method is x.yz." For that kind of model to even output a specific number at all, it would have to include such factors as turnspeed. Well, that's a factor that depends both on the cuber's experience/current ability, and the qualities of the cube they are using. And both of these can get better over time as a cuber gets better and as cubes get better. So if you want the model to say "the speed of CFOP is x.yz," where will you get specific numbers to represent these factors? How can you say unarbitrarily "the best turnspeed achievable with the best cube that could be produced in the future is xx.xx"? Furthermore, even if you can answer that question, that is still looking specifically at human capability, which is separate from the properties of a method.

Last edited: