Find your best cube

[SpeedCubeReview]

I went on a mission to find what really is my fastest cube. I have four main cubes I use (Aurora, Fangshi mini, Guanlong, and Aolong V2). They all have matching shades of stickers. What I did was a Average of 100 and switched up the cubes every 10 solves. My rule was that I had to do a full set on each cube at one time so the lighting of the room or alertness of myself is consistent. I also switched up the order for each set.

I was mostly surprised at how close everything was.

Cube........Ao100...........Best
Aolong.....26.150.........19.023
Aurora......26.824........18.855
Guanlong..27.085........20.055
Fanshi......27.318........18.690

I would like to try this again with the Aolong and the mini V2 when it ever comes out. I think the mini Fangshi cube worked better for me but since the Fangshi is white with black caps I feel that my recognition was off.

Anyone else up for trying this? Post your results if you do!

 

[SpeedCubeReview]

Or, a different question... Has anyone done something similar to this?

 

[Ollie]

The average times are kinda meaningless without the standard deviations, so please post those if you still can.

 

[StachuK1992]

Unless you're going to do an average of 100+ (and I think probably 1000+, honestly), all results you're coming up with are pretty much useless.

Some scrambles suck. Sometimes you sneeze in the middle of a solve. Sometimes you're in the zone.

 

[SpeedCubeReview]

Unless you're going to do an average of 100+ (and I think probably 1000+, honestly), all results you're coming up with are pretty much useless.

Some scrambles suck. Sometimes you sneeze in the middle of a solve. Sometimes you're in the zone.

I totally agree. I tried to fix the "in the zone" issue with doing a set on each cube during every group of solves I do in one sitting.
100 will give you a much better average than 5 or 12 though and for now it could be a pretty close indicator of how well a cube works for you.

 

[Antonie faz fan]

Unless you're going to do an average of 100+ (and I think probably 1000+, honestly), all results you're coming up with are pretty much useless.

Some scrambles suck. Sometimes you sneeze in the middle of a solve. Sometimes you're in the zone.

pretty sure you added an extra 0 to each of those numbers, srsly 100 solves is accurate enough. during summer holiday i did probably 5 seperate avg 100 and each of them was between 10.07 and 10.32.

 

[TDM]

pretty sure you added an extra 0 to each of those numbers, srsly 100 solves is accurate enough. during summer holiday i did probably 5 seperate avg 100 and each of them was between 10.07 and 10.32.

... yet my Ao100s can be anywhere between sub-13 and sup-14. Just because you might be that consistent, it doesn't mean everyone else is.

 

[PenguinsDontFly]

Unless you're going to do an average of 100+ (and I think probably 1000+, honestly), all results you're coming up with are pretty much useless.

Some scrambles suck. Sometimes you sneeze in the middle of a solve. Sometimes you're in the zone.

Lol I love it when I sneeze in a solve! I keep turning and then I recover. It feelw so cool to see that look ahead has applications!

 

[rowan]

Stachu's 1000+ is even a joke amount. 2^21 is more like significant portion of solved states. (Too lazy to find out exactly how few you could do and still be in the clear.) To get a statisitcally significant amount of solves done on each cube would take years.

For ViloaBouqet, for instance, assessing four cubes at about 26 seconds (she would likely improve but I can't account for that accurately), the math comes out to 6.91 years. Add in sleeping, work, eating and it becomes even more impractical.

I think your "best cube" should be measured by enjoyment since the data really won't mean anything if you actually want accuracy.

 

[PenguinsDontFly]

Stachu's 1000+ is even a joke amount. 2^21 is more like significant portion of solved states. (Too lazy to find out exactly how few you could do and still be in the clear.) To get a statisitcally significant amount of solves done on each cube would take years.

For ViloaBouqet, for instance, assessing four cubes at about 26 seconds (she would likely improve but I can't account for that accurately), the math comes out to 6.91 years. Add in sleeping, work, eating and it becomes even more impractical.

I think your "best cube" should be measured by enjoyment since the data really won't mean anything if you actually want accuracy.

Agreed. Cubes these days are all fast, pop resistent, and can corner cut. Its all personal preference from anything as old as a guhong v1 to an aolong v2.

 

[Ollie]

Stachu's 1000+ is even a joke amount. 2^21 is more like significant portion of solved states. (Too lazy to find out exactly how few you could do and still be in the clear.) To get a statisitcally significant amount of solves done on each cube would take years.

For ViloaBouqet, for instance, assessing four cubes at about 26 seconds (she would likely improve but I can't account for that accurately), the math comes out to 6.91 years. Add in sleeping, work, eating and it becomes even more impractical.

I think your "best cube" should be measured by enjoyment since the data really won't mean anything if you actually want accuracy.

Nope, 100 solves per cube is sufficient. Statistical tests take variance into account. Independent groups t-tests will tell you if there is a significant difference between the cubes (or a one-way ANOVA, just need the individual times for each session)

 

[hawkmp4]

Nope, 100 solves per cube is sufficient. Statistical tests take variance into account. Independent groups t-tests will tell you if there is a significant difference between the cubes (or a one-way ANOVA, just need the individual times for each session)

That assumes that solving times are iid draws from a normal rv, and that variance is equal across both cubes.

This is incorrect. Solving times are highly skewed (long tailed to the right). Solving times also will improve over the course of a session, we all know this from experience. We also have no idea if variance is equal across both cubes. It probably isn't. A bad cube is going to have lockups and such, which will increase variance as well as the mean.

Let's pretend that a t test is remotely appropriate, just for ***** and giggles. "100 is enough" is a statement that needs a lot of support. Enough for what? How do you know?

 

[jokeeffe]

Or, a different question... Has anyone done something similar to this?

I'm in the middle of this now, with a group of five cubes -- Zhanchi, Aolong V2, Panshi stickerless, Aurora, and a Rubik's brand thrown in for good measure. Following a very similar methodology, targeting 100 solves on each cube, but rotating on every solve (not groups of 10). I've been trying to do 15-20 solves per day for this project, should be done in a couple of weeks.

 

[Ollie]

That assumes that solving times are iid draws from a normal rv, and that variance is equal across both cubes.

This is incorrect. Solving times are highly skewed (long tailed to the right). Solving times also will improve over the course of a session, we all know this from experience. We also have no idea if variance is equal across both cubes. It probably isn't. A bad cube is going to have lockups and such, which will increase variance as well as the mean.

OP wants to know what is the fastest cube out of a possible four. The main point of my reply to rowan is that you don't need to do tens of thousands of solves to work this out, given the cubes he is testing. OP has only given the Ao100 times without standard deviation or any other data, so for all we know he could be fairly consistent and the solving times might distribute evenly.

Let's pretend that a t test is remotely appropriate, just for ***** and giggles. "100 is enough" is a statement that needs a lot of support. Enough for what? How do you know?

And fine, I concede the independent groups error - the best thing in an ideal world is equally matched cubers doing n number of solves with the same scrambles each on a different cube which isn't practical for the OP's project. But my point was that doing 100 solves on each cube will provide some evidence for which cube is probably faster out of the four.. I'm not asserting that 100 solves per cube is enough to definitively prove one way or the other.

 

[IRNjuggle28]

Hmm, interesting. I doubt I'll try this since I'm very, very slow at getting used to new/different cubes and I hardly ever change my mains or practice with something that isn't my main, but still interesting to see your results. I'm not at all surprised that your times are this close with all of your mains. It really is the cuber, not the cube. For the most part, at least.

 

[Myachii]

I doubt that OP wanted a perfect statistically correct answer, rather just wanted to see how fast they were on each cube doing a fairly large number of solves on each puzzle. Bringing more Maths into this is pointless.

 

[SpeedCubeReview]

The average times are kinda meaningless without the standard deviations, so please post those if you still can.

I'm using ChaoTimer on my iPad. DI do not think it has that option on there.

 

[SpeedCubeReview]

I doubt that OP wanted a perfect statistically correct answer, rather just wanted to see how fast they were on each cube doing a fairly large number of solves on each puzzle. Bringing more Maths into this is pointless.

:tu

 

[SpeedCubeReview]

So I received my Aolong mini a couple weeks ago. I did 500 solves on each switching up which one I did first switching cubes every 10-20 solves and doing an equal amount on each every time I sat down.

Aolong 57mm average = 24.7 seconds
Aolong mini 54.6mm average = 24.3 seconds.

I felt like the 57 was clunky compared to the mini and was pretty surprised at how close the numbers were. I will be sticking to the mini. I have large hands, but I think playing the violin made me used to small finger movements that work better on the mini.