On January 1, 2017, I asked myself the question: Within one month of practice, can I solve a Rubik’s Cube in under 20 seconds?
On January 24, 2017, after 25 hours of practice, I found out that the answer was yes.
During the month of January, I documented my entire learning process in a series of 31 daily blog posts, which are compiled here into a single narrative. In this article, you can relive my month of insights, frustrations, learning hacks, and triumphs, as I strive towards monthly mastery.
It’s January 1st today, which means it’s time to start a new challenge. This month, my goal is to solve a Rubik’s Cube in under 20 seconds.
Why 20 seconds?
Two months ago, at San Francisco’s Ferry Building, I went to a book event and signing by Ian Scheffler, the author of Cracking the Cube, the predominant book on speed cubing (the competitive sport of solving the Rubik’s Cube).
At the event, I spoke with the author and the dozens of Bay Area speed cubers that attended. I wanted to learn about the major benchmarks in the sport.
Repeatedly, what I heard was “The most significant benchmark in speed cubing is the 20-second mark. Breaking 20 seconds is like breaking the four minute mile”.
At first, I just thought that the equivalence was symbolic, but, that day, I also heard that “If you look at the proportion of cubers who can solve “sub-20” to all cubers (i.e. people who own a Rubik’s Cube), it’s about equivalent to the proportion of runners who can run sub-4 miles out of all competitive runners”.
I’m not exactly sure how this was computed, or exactly how accurate it is. But, nevertheless, it’s clear that 20 seconds is the most notable benchmark in speed cubing.
Thus, I’ve set my sights on a sub-20 time.
My starting point
I actually have a long history with the Rubik’s Cube: I originally learned how to solve it when I was in Middle School, and, ever since, I’ve been casually solving it for fun.
However, I’ve never really tried to go fast. Instead, I prefer the more meditative, tactile approach of smoothly solving the cube.
In fact, I haven’t actually timed myself in eight or nine years. So, to properly kick off this month, I set up a camera and a computer-based timer, and completed five solves at top speed.
The average of the five solves was 48.29. (44.46, 51.83,46.15, 53.62, 45.41).
This isn’t horrible nor is it great. However, if you have no Rubik’s Cube experience, it’s possible to think that I’m not too far away from a sub-20 time.
But, to continue the running analogy, solving a Rubik’s Cube in 48.29 is symbolically equivalent to running a 7-minute mile. In other words, it’s a reasonably respectable time for an amateur, but it’s far from a competitive time for a serious athlete.
A month of deliberate practice
Over the past nine years, I’ve solved the Rubik’s Cube hundreds of times, but never with the goal of improving. As a result, my time has only improved by around 10 seconds from 60-ish seconds to 50-ish seconds.
This month, I’m taking a more deliberate approach with my practice, explicitly looking to shave another 30 seconds off my time.
My hope is that my effort this month reveals the overwhelming difference between nine years of mindless practice and 30 days of deliberate practice.
Time to formulate my plan of attack…
Over the next few days, I’ll be experimenting with a number of training techniques and exercises, to determine the best 30-day Rubik’s Cube training program. In particular, the program needs to get my Rubik’s Cube solving speed consistently under 20 seconds.
To understand my training program and the rest of this month’s posts, it’s important that you first understand the method I use to solve a Rubik’s Cube, which is called CFOP.
Intro to CFOP
CFOP is the most common speed cubing method and is the one that I originally learned.
CFOP stands for 1. Cross, 2. F2L, 3. OLL, 4. PLL, which are the four steps used, in this method, to solve the cube. These acronyms probably don’t mean much, so let me go through each one.
1. Cross
With a fully scrambled cube, I start by finding the white center, which identifies the white side of a solved cube.
I then find the white edge pieces, and move them into place, so that a white cross is formed on the white side.
It’s important that the edge pieces not only match up with the white center, but that they also match up with the correct color of the adjacent centers.
For example, the white and green edge piece should be placed between the white center and the green center. The white and red edge piece should be placed between the white center and the red center. And so on.
The cross is complete.
2. F2L (First 2 Layers)
Once the cross is formed, I put the white side of the cube on the bottom, which looks like this.
In this next step, F2L, or ‘First 2 Layers’, I’m focused on solving the first two layers of the cube, starting from the bottom.
Since the center pieces and ‘white + color’ edge pieces are already solved (from the last step), I only need to focus on solving the highlighted areas below.
After placing the correct pieces in these slots, the cube looks like this.
Here, in the solved state, it’s easier to visualize which pieces I moved during F2L: I positioned the four edge pieces (like the blue and red edge piece) between the corresponding centers (like the blue center and the red center). I also positioned the four corner pieces that have a white side and two other colored sides (like the white, blue, and red corner piece) in the correct slot on the bottom layer.
As a result, when I finish F2L, the white side is completely solved.
F2L is complete.
3. OLL (Orient the Last Layer)
At this point, the white side is completely solved and the first two layers of the cube (from the bottom) are solved. Thus, I only need to solve the last layer of the cube.
To do this, I break down the last layer into two parts. First, in this step, OLL, I solve the yellow side, which is also known as orienting the last layer.
At this point, on the yellow face, it’s possible for the yellow pieces to be configured in 57 different ways.
The fastest speed cubers have memorized the 57 different algorithms necessary to most efficiently solve the 57 different configurations (one algorithm for each configuration).
However, if you’re willing to give up efficiency, it’s possible to solve the yellow side with far fewer algorithms. I currently only know 9 of the algorithms, which means I’m highly inefficient at this step.
Upon completing the necessary algorithm, the yellow side is solved.
OLL is complete.
4. PLL (Permute the Last Layer)
After finishing OLL, the cube looks like this.
The first two layers and the yellow side are solved. The only thing left to solve, during this step, PLL, is the rim of the last layer, which is also know as permuting the last layer.
There are 21 different configurations for PLL, which can be solved most efficiently with the corresponding 21 algorithms.
I currently know 7 of the PLL algorithms, which is enough to complete PLL in twice as many moves as a competitive speed cuber.
After executing the appropriate PLL algorithm, the rim of the last layer is solved.
And as a result, the entire cube is solved.
Going into each month, I have a reasonably good idea how I plan to approach that month’s challenge.
This month, my plan was simple: Learn the 62 Rubik’s Cube algorithms I don’t know. Obtain a sub-20 time.
I figured that, since I know only 21% of the algorithms, learning the remaining 79% will drop my time low enough to meet my goal.
It turns out that it’s not this simple.
Right now, algorithms don’t matter
As a reminder, there are four steps to solving the Rubik’s Cube (The cross, F2L, OLL, and PLL), which I explain in yesterday’s post. As a simplification, there are really two parts to the solve: The intuition-based part, which includes the cross and F2L, and the algorithm-based part, which includes OLL and PLL.
My original plan was just to focus on the algorithm-based part, since this approach seemed contained, very measurable, and well-defined. I would learn three algorithms per day, leaving me one week at the end to practice and perfect all the algorithms together.
But there’s a problem: it takes me, on average, 30–35 seconds just to reach the algorithm-based part of the solve. In other words, even if I learned and could perform all 78 algorithms perfectly, at full speed, I wouldn’t even reach that part of the solve until much after 20 seconds elapsed.
Basically, the intuition-based part (the cross and F2L) currently takes way too long, making it unignorable. In fact, since it makes up the majority of my solve time, I will likely need to put much of my focus on these steps.
The 20-second breakdown
According to Rubik’s Cube forums, if I want to solve the cube in 20 seconds, my time should be allocated, more or less, in the following way.
- Cross: 2 seconds
- F2L: 10 seconds
- OLL: 2 seconds
- PLL: 6 seconds
Thus, I need to throw out my algorithm-based plan, and instead, create unique plans for each stage to achieve the necessary times.
Of course, I can give and take seconds from step to step, depending on where my strengths lie, but this is a good framework to start with.
Tomorrow, I need to start experimenting with F2L training techniques (since these past three days I’ve spent just learning new algorithms).
I have a hunch that I will be reprising November’s use of a metronome…
As mentioned yesterday, I need to decrease my speed on the intuition-based parts of my solve (the cross and F2L) in order to achieve a sub-20 time.
After analyzing a few of my solves, there’s clearly one major thing currently slowing me down: Cube Rotations.
What is a Cube Rotation?
If I turn the entire cube clockwise or counterclockwise around its vertical axis, I’ve executed a cube rotation.
Executing a cube rotation doesn’t get the cube any closer to the solved state, but still costs valuable time. In fact, not only am I losing time executing the rotation, but I’m also losing time getting reoriented to the rotated cube. (For those who read about productivity, this is a bit like context switching, which has time implications beyond just the act of switching tasks).
As a result, when solving, speed cubers attempt to minimize their number of executed cube rotations, aiming for one or two at the most.
Today, before analyzing some video, I imagined that my cube rotating was minimal (perhaps 4 or 5 rotations per normal solve), but after filming myself, I realize I’m an aggressive cube rotator.
Watch as I execute 13 cube rotations just while solving the cross and F2L…
I execute Cube Rotations for two reasons
The first is because I’m trying to find the pieces I need to complete the cross and F2L. I’ll address this problem, which I call Inspection Pauses, tomorrow.
The other reason I execute Cube Rotations is to gain positional leverage. In other words, it’s easier for me to execute certain Rubik’s Cube maneuvers when holding the cube in certain, better-practiced positions. Simply, I’m rotating the cube into more preferable orientations.
While I’m nicely compensating for my weaknesses, this isn’t a healthy practice (if I want to get universally faster). It’s as if I learned how to play a guitar solo only in the key of A major, and every time the band plays anything different, I need to stop the band and ask them to transpose the song into the key I know.
Like a good guitarist, if I want to solve the Rubik’s Cube fast, I need to be comfortable with the full range of possibilities and all the ways they can be presented to me — whether that’s all the possible musical keys of a song or all the possible ways a certain maneuver is oriented on the cube.
How I plan to get better (introducing FRS)
In order to practice cube maneuvers from all possible orientations, and also in order to reduce Cube Rotations, I’ve created a simple exercise called Forced Rotationless Solving (FRS).
During FRS, Cube Rotations aren’t allowed. Instead, I must determine new ways to solve the cross and F2L, from new vantage points, without ever rotating the cube.
I’ve completed about 25 solves today using FRS and am already developing a knack for these new maneuvers. In particular, I’ve gotten good at inserting pairs in the back. I’ll explain specifically what this means in a future post, but generally it means that I can now (more comfortably) solve pieces on the back face of the cube, rather than having to rotate the cube 180 degrees to solve them in the front.
During the rest of the month, I’ll continue using FRS as a warmup exercise before I practice my normal speed solves.
Hopefully, I can drop the number of Cube Rotations during my solves into the low single digits.
Yesterday, I realized that Cube Rotations were majorly slowing me down during the intuition-based parts of my Rubik’s Cube solve. In an ideal world, I’d have only one or two rotations per solve, but instead, I have about 10–15.
As a reminder, these are Cube Rotations…
Yesterday, I discussed the positional reasons for Cube Rotations (and introduced Forced Rotationless Solving, FRS, as a potential solution). Today, I’ll address the other reason for rotations: Inspection Pauses.
Inspection Pauses
During F2L (solving the first 2 layers), I must correctly place eight pieces: four pairs of corresponding corners and edges.
In other words, during F2L, I must create four corner-edge pairs, and then place each pair in the correct position (called a “slot”) on the cube.
If you watch the video above, you’ll notice that I execute a fast series of moves. Then, I pause, rotate/inspect the cube until I find the next pair of pieces, execute another series of fast moves, and so on.
Basically, I’m not able to flow fluidly from pair to pair during F2L. Instead, my F2L is broken up by these inspection pauses.
Looking into the future
If I want to eliminate Inspection Pauses, I need to search for the next pair while I simultaneously execute the necessary moves for the current pair. In other words, I need to look into the future.
This is tough.
So, to practice, I’ve returned to using the Pulse metronome app, which I used in November to become a pseudo grand master of memory.
Here’s the idea: I set the metronome to something modest, like 90 BPM. Then, I attempt to solve the cube by making a turn every time the metronome clicks. As a result, by the nature of the exercise, all Inspection Pauses are (theoretically) eliminated, and the solve is consistent and fluid.
Even solving at 90 BPM, in this way, is surprisingly challenging for me. My brain really needs to work hard to both execute the moves and search for the next pieces simultaneously. But, as a result of this brain strain, I can feel my cube vision improving.
Better cube vision = faster solves, so nightmare-inducing metronome clicks are back in my life for at least the next week.
When I first explained how I solve a Rubik’s Cube, I noted that I always start by creating a cross on the white side. This consistency helps with speedy pattern recognition, but may not be the most efficient method.
So, to address this potential inefficiency, I experimented today with Color Neutrality.
What is Color Neutrality?
Under color neutrality, rather than always starting a solve with the white side of the cube, I must assess each unique scramble and identify which color seems easiest to start with (where easiest is some combination of fewest number of moves and ease of execution). The easiest color could be white, but it could also be yellow, blue, green, red, or orange.
Thus, to be a color neutral cuber, I must be able to 1) Quickly identify the easiest starting color for a particular scramble, and 2) Execute my solve with equal competency and pattern recognition from any starting color.
To determine if Color Neutrality is something worth pursuing, I needed to answer a couple of questions…
How much time would Color Neutrality save me?
To answer this question, I timed myself solving 20 crosses on the white side of the cube. I wanted to see how much time I was saving between an easy cross and a hard cross.
I found that my cross time was between 2 and 4 seconds, meaning perfect Color Neutrality, at best, will save me only two seconds per solve. Of course, two seconds is significant if I want to beat the world record (or if I’m consistently stuck at 22-second solves), but it’s only a tiny boost in absolute efficiency.
How much effort would it take?
Since I’d only be gaining two-ish seconds from Color Neutral solving, I needed to assess the investment level.
To do so, I completed ten solves starting on a color other than white.
Surprisingly, I’ve never actually tried this, so I didn’t know what to expect. The result was worse than I expected: On average, my Color Neutral solves were about 30% slower than my solves starting with white.
This is a big hill to climb.
Of course, with practice, I can overcome this difference, but I’m just not sure it’s worth my attention during this month’s challenge. Especially since I’ll only be shaving 1–2 seconds off my time for doing so.
I’m glad I went through the exploration, but I feel my efforts are better spent elsewhere.
What should YOU do?
If you are just learning to solve the Rubik’s Cube now, or if you ever plan to learn in the future, I would encourage you to start practicing Color Neutral solves as early as possible (assuming you care about speed).
It seems that the better and better you get with one color, the harder it is to justify going backwards to gain Color Neutrality.
In fact, if I was just learning how to solve the cube now, I would probably force myself to start each new solve with a different color than the previous solve. This is an easy mechanism to understand and follow, and will reap nice Color Neutral rewards.
Today, I solved the Rubik’s Cube 100 times, which took about 2 hours (with a few breaks).
I started with 20 Forced Rotationless Solves, which helped improve my muscle memory for maneuvers in less common orientations.
Then, I executed 30 solves to a metronome at 100 BPM, which helped improve my look-ahead.
Finally, I completed 50 regular speed solves to put the pieces together (literally, I guess).
Adding algorithms to my training
Over the past few days, I’ve only been training the first part of my Rubik’s Cube solve, ignoring the second, algorithms-based part.
Now that the exercises to train the “first part” are well-defined, it’s time to start addressing new algorithms as part of my training regime.
I’m not sure if I can learn all the remaining algorithms in the next three weeks, so tomorrow, I’ll figure out which ones to focus on.
When I first described how I solve a Rubik’s Cube, I alluded to the fact that some parts of the cube can be solved using a set of 78 algorithms, but I didn’t say much further.
Today, I’ll explain the fundamentals.
What is a Rubik’s Cube algorithm?
Simply, an algorithm is a set of pre-determined moves that, when properly executed, accomplish a specific task (i.e. “move these particular pieces on the cube into this particular configuration without moving/messing up these other pieces”).
The most common way to express a Rubik’s Cube algorithm is using Basic Notation, which is depicted below.
Under the Basic Notation scheme, R means “turn the right face of the cube clockwise”. R’ (R prime), means “turn the right face of the cube counterclockwise”. And so on, with F = front; B = back; L = Left; R = Right; U = Up; D = Down.
A full algorithm may look like this: F R U R’ U’ F’
When can you use Rubik’s Cube algorithms?
Once you solve the first two layers of the cube, you enter the algorithm-based portion of the solve, which is focused on solving the last layer.
Here’s what the cube looks like at this point.
To solve the remaining part of the cube, you must execute two classes of algorithms.
The first class of algorithms is called Orient Last Layer (OLL), which solves the top face of the last layer.
There are 57 different algorithms to solve each of the 57 possible patterns.
The second class of algorithms is called Permute Last Layer (PLL), which solves the rim of the last layer (and, as a result, the entire cube).
There are 21 PLL algorithms.
Which algorithms should I learn?
To solve the cube, you only need to know four OLL algorithms and four PLL algorithms. But, to solve the cube fast, you need to know as many as possible.
Since I don’t think I can learn all 78 this month, I need to determine which subset to study. Tomorrow, I will finish figuring this out and will write about it.
How should I optimally practice algorithms?
This another question I haven’t yet answered.
Once I determine my target algorithms, I’ll experiment with some training ideas, and share out what I find.
Today, I was trying to select which set of last layer algorithms to learn, when I stumbled on something amazing…
Two-Look OLL algorithms
As explained yesterday, 57 of the 78 Rubik’s Cube algorithms are used for the Orient Last Layer (OLL) step of the solve. Thus, I suspected that OLL would make up most of my algorithm learning effort.
However, surprisingly, after today, that’s no longer true:
This morning, I learned about something called Two-Look OLL, which basically means “If I’m willing to string together two algorithms back to back (instead of using just one algorithm), I only need to learn 10 OLL algorithms to cover all 57 OLL patterns”.
Here are those algorithms:
At the beginning of this month’s challenge (and as of this morning), I only knew a random assortment of 9 OLL algorithms that I picked up over the past many years of casual cubing.
I have no idea why I know the particular set of algorithms I do. I just do.
When I learned about “Two-Look OLL” this morning, the first thought I had was “I wonder how many of the 9 algs I know match the 10 Two-Look algs”. If I just linearly extrapolate the probability (which is perhaps combinatoriclynaive), I would expect something like a 15% match.
However, unexpectedly, I already knew 70% of the two-look algs. Somehow, 7 out of the 9 algs I already knew perfectly matched 7 out of the 10 two-look algs.
And, on top of that, the three algs I didn’t know were only minor variations on others that I did. In fact, because these three algs were so similar to others (i.e. mirror images, one move apart…), I was able to learn all three new algs in ten minutes.
Normally, it would probably take a day to fully incorporate a new alg.
Thus, excitingly, as of today, I know all the OII algorithms I need for this month’s challenge (I’ll prove why two-look is sufficient below).
It feels oddly lucky how much overlap there was between the set of algorithms I knew and the set of algorithms I wanted to know, but I suspect there is some underlying reason that propelled me to learn these particular algorithms in the first place. I’m just not sure what that underlying reason is.
If you’re a speed cuber and have a reasonable guess, leave a comment. I’d love to know.
PLL algorithms
The remaining 21 algorithms (out of the 78) are used for the Permute Last Layer (PLL) step.
Unlike OLL, I’ve decided to learn all 21 this month.
I already know 7, so, over the next two weeks, I will learn one of the remaining 14 algs per day. This will afford me one full week, at the end of the month, where I can execute solves with full knowledge of all the relevant algorithms.
Will this work?
If I just learn Two-Look OLL and complete PLL, will I be able to execute sub-20-second solves? It looks like the answer is yes:
I reviewed the video of solves I made on Day 1 of this challenge, and extracted a few relevant statistics.
On average, I was using 3.2 OLL algorithms and 2 PLL algorithms to complete the five solves.
With two-look OLL and complete PLL, I’ll reduce the number of algs to 2 and 1 respectively. In other words, assuming time maps linearly to ‘number of algorithms’, I’ll be cutting my OLL time by 1/3 and my PLL time by 1/2.
In Solve 1 from the video, which is the most average solve (since I used 3 OLL and 2 PLL algs), it took me 6 seconds to solve OLL and 8 seconds to solve PLL.
Therefore, if I can reduce my times by the amounts above, I will theoretically be able to execute a 4-second OLL and a 4-second PLL, or, in total, an 8-second last layer.
In my post “I already need a new plan”, 8 seconds is exactly how much time I allocated myself for my 20-second solve:
- Cross: 2 seconds
- F2L: 10 seconds
- OLL: 2 seconds
- PLL: 6 seconds
I’m shifting some of my PLL time to OLL to compensate for two-look, but it seems like it will work out.
Tomorrow, I’ll start practicing the new algs and determine which training methods are optimal.
Yesterday, I determined that I need to learn 14 more Rubik’s Cube algorithms this month. Today, I learned one of the 14.
Here was my approach :
- On the train ride to work, I practiced the movements of the algorithm over and over again. While doing so, I wasn’t trying to use the algorithm in context or solve the cube. Instead, I simply wanted to learn the feeling of executing the algorithm. I repeated the algorithm about 200 times until I felt I could execute it without consciously considering each of the individual moves.
- On the train ride home, I executed the algorithm again, but this time, I watched what happened to the patterns on the cube. In particular, I was trying to learn how I could quickly identify the pattern and associated algorithm (based usually on only seeing two or three sides of the cube), so that I could execute the new alg in flow during a solve.
- Once I got home, I solved the cube completely 40 times. During these solves, the new algorithm came up three times, giving me the chance to assess how smoothly (and correctly) I could execute it. The first time didn’t go so well, but the other two times flowed perfectly.
While this approach isn’t too profound, it got the job done.
Over the past 11 days, I’ve solved the Rubik’s Cube probably 400–500 times, which is a lot. In addition, I’ve also focused on learning new algorithms, and training my pattern recognition abilities with a few different exercises.
Ever after all of this practice so far, I still don’t seem to be getting much faster. In fact, some of the speed gains I saw in the first few days seem to be reverting (i.e. I’m going slower than I did ~4 days ago).
I have four theories why this is happening:
1. My cube is getting worse at the same rate I’m getting better
As I practice more and more with my Rubik’s Cube, it’s out-of-the-box buttery smoothness is breaking down. I think this is from a combination of dust and general wear and tear. In other words, the cube is getting harder to turn and sometimes locks up when I’m going really fast.
It’s possible that, as my cube gets slower, I’m getting generally faster, so the effects cancel out and I’m left with an unchanging time.
The potential solution: Lube.
Yes, Rubik’s Cube lube is a real thing and very popular amongst speed cubers. I’ve ordered some on Amazon and will put it to use once it shows up.
2. My brain is getting overloaded
Two months ago, when I was trying to learn how to memorize a deck of cards in less than two minutes, between days 7–10, I dipped to my worst level of performance. My brain had been subjected to so much new information over the previous week, and hadn’t yet caught up. Once it did, after a few days, I made a big leap forward in my progress.
It’s possible I’m experiencing a similar phenomenon now.
The potential solution: Don’t get discouraged and push through. If I don’t see positive changes in a few days, I should reassess this theory.
3. I’m practicing poorly
I’ve noticed that, during real solves, I ignore a lot of the techniques I’ve been practicing. For example, I’ve been working hard to reduce my cube rotationsand improve my look-ahead via two main directed exercises. But, when it’s time to apply these principles to an actual solve, I freak out, forget everything, and spin the cube frantically. This is clearly not good.
The potential solution: Combine my training exercises with actual solves, forcing myself to actively look-ahead and keep the cube positionally controlled.
4. Poor pre-solve planning
Before every solve, speed cubers are allowed 15 seconds to inspect the scrambled cube. During this time, it’s advised to make some sort of plan.
Typically, most speed cubers can plan out the entire cross and one F2L pair during inspection. On the other hand, I seem to only plan, on average, two of the four edge pieces necessary to form the cross.
As a result, my cross is still too slow, and my transition to F2L is usually terrible. After I finish the cross, I pause for maybe 1–2 seconds just to assess the situation.
This is not good.
The potential solution: Practice planning out the full cross during inspection. Also, practice tracking an F2L pair while executing the cross, so I can smoothly transition from the cross to F2L.
(Not sure what “the cross” or “F2L” are? Read my post about how I solve the Rubik’s Cube)
Validating these theories
When learning new skills, especially for speed, it’s always a good practice to form hypotheses about where the major inefficiencies lie, and then attempt to validate/invalidate these theories.
That’s what I plan to do over the next few days, which will educate my training moving forward.
Yesterday, I took some time to reflect on the Rubik’s Cube progress I’ve made so far this month. I mentioned that I wasn’t progressing at the pace I was hoping, and I presented four theories as to why I thought this might be.
Interestingly, when I started writing the article, I only had three theories in mind. The fourth came to me while I was proofreading.
I tested this fourth theory today, and shaved two seconds off my average solve time. It also reminded me something important about life.
The Fourth Theory: Poor pre-solve planning
Yesterday, here’s the theory I put forward:
Before every solve, speed cubers are allowed 15 seconds to inspect the scrambled cube. During this time, it’s advised to make some sort of plan.
Typically, most speed cubers can plan out the entire cross and one F2L pair during inspection. On the other hand, I seem to only plan, on average, two of the four edge pieces necessary to form the cross.
As a result, my cross is still too slow, and my transition to F2L is usually terrible. After I finish the cross, I pause for maybe 1–2 seconds just to assess the situation.
Today, as a follow-up, I attempted to better use my inspection time, which I expected I could improve over the next week with deliberate practice.
However, as soon as I started actively using my prep time, I was able to plan for the entire cross during inspection without a problem. I guess this is something I was already able to do (and was previously just being mentally lazy about).
The result: My solves were two seconds faster, on average, which is a nice bonus and a great help in my quest towards a 20-second solve.
It’s interesting that I didn’t think about this inefficiency earlier in the month.
A hiccup in my approach
Typically, the way I approach learning a new skills is as follow:
- Breakdown the entire process end-to-end
- Identify which parts of the process I’m worse at / are causing the most inefficiencies
- Identify which parts of the process offer the most upside growth potential
- Overlay these two lists to determine how to prioritize my practice
- Practice
- Repeat, as appropriate
However, when I was originally working through this approach, I was completely blind to “the inspection phase” as part of the end-to-end process. In fact, in the post where I deconstruct the Rubik’s Cube process, I never even mention inspection.
This was a big miss.
Why did I originally miss this?
I’m not actually surprised I missed inspection as part of my deconstruction.
I fell into a very common learning trap. Or maybe it’s just a life trap in general:
In life, we are often so driven to succeed at our goals, that we jump right to execution without taking a moment to plan for and setup our success. As a result, even though it feels like we’re moving faster, we’re actually going slower.
Usually, this is the result of mental laziness. It’s easier to distract our brains with “progress”, than it is to thoughtfully plan and reflect on our goals.
My eagerness to solve the cube forced me into this not-usually-so-optimal behavior.
Anyway, I plan to use all 15 seconds of my inspection time moving forward.
On Day 1 of this month’s challenge, I filmed myself solving the cube. My average solve time was 48.29 seconds.
Today, I filmed myself again (solving the cube 20 times). During today’s session, my average speed was 32.88 seconds, which represents solid progress.
Here are three of my best solves from today:
Two days ago, I mentioned that my cube is getting a little stiff. So, I ordered Rubik’s Cube lube on Amazon.
Today, the lube showed up and I made an overly-cinematic video explaining my less-than-precise lubing technique.
It turns out, I used way too much lube.
The cube spins nicely, but I got the lube all over the outside, which is making it nearly impossible to hold on to the cube.
Luckily, I also ordered a few new cubes on Amazon, and they also showed up today, so I’ll use those until I figure out how to effectively remove the excess lube.
When learning something new, I typically pass through four general stages:
- The Basics: Learning is easy and plentiful
- The Slump: Progress seems to standstill or regress
- The Leap: The pieces come together and progress leaps forward
- The Final Stretch: Progress accelerates until the goal is surpassed
Somehow, while I slept last night, my training hardened in my brain, and I transitioned from The Slump to The Leap. In fact, this morning, I could consistently solve the Rubik’s Cube 5 seconds faster than I could when I went to bed last night.
Not only were my times faster, but I could consciously feel the difference in the way I see and track patterns on the cube. I’m not sure what the difference is, but I certainly feel it.
For anyone who’s built software before, the feeling is similar to the following: Your code isn’t working. You’ve been debugging for a few hours, and nothing seems to fix the problem. All of sudden, while you’re taking a shower, or eating dinner, etc., something clicks and you realize what the problem is and how to fix the bug. You haven’t yet made the fix or tested the updated code, but you are certain that it will work. In the meantime, you stand in the shower or sit at the dinner table, feeling this weird mix of at peace and tremendously optimistic.
This is more or less how I feel. Something has clicked, and I’m now feeling really good.
Of course, this feeling comes at the cost of The Slump, which, for this month’s challenge, lasted from around Day 3 until Day 14, significantly longer than the previous two months.
Anyway, I still have a lot of work to do in order to reach my sub-20 goal, but I now feel momentum is on my side.
Today, there’s a lot weighing on my mind. And, while I appreciate the fact that some days are just like this, it definitely makes it challenging to solve the Rubik’s Cube as fast as I know I can.
In other words, in order to solve the cube quickly, it’s important that I get fully in the zone, and today, my brain doesn’t want to do that. Instead, my mind is wandering in the middle of solves.
How to fix this?
Of course, the best way to fix this is to address the underlying problem, which I’m working on.
In the meantime, I came up with a clever little hack that’s been helping, which requires a small explanation:
In November, my goal was to learn how to memorize a deck of cards in under two minutes, which also required complete concentration. Thus, while I trained, to help eliminate external distractions, I would listen to a combination of white noise and rain sounds in my headphones. After a while, the white noise became a trigger for me to get in the zone.
Basically, every time I hear the white noise, my brain automatically gets into the associated fully focused state, ready to memorize cards (or in this case, solve a Rubik’s Cube).
So today, even though the noise is coming from inside my own head, the white noise actually helped. The trigger is still strong.
I can’t continue relying on this hack for too long. Otherwise, I’ll probably override the trigger or at least lessen its effect.
But, for today, it’s working for me.
When I first put together the list for my 12 Month to Master challenges, I picked things that were ambitiously out-of-reach. This was on purpose. I wanted goals that seriously challenged my interests, and would push my skills towards genuine mastery.
Over the past 78 days, as I’ve explained the project to many and written my daily blog posts, I’ve slowly become numb to my challenges. My passion is still there (mostly), but my gauge of the challenges’ difficulty has certainly dropped a lot.
This is probably the result of recognition bias combined with prematurely converting my goals to social reality.
If I don’t recalibrate my perspective soon, though, I fear that my passion will also become numb.
Luckily, today helped put things back in perspective: As I continue making progress towards this month’s goal, I realize “Solving a Rubik’s Cube in under 20 seconds is crazy hard. Especially in this timeframe”.
And while I’ve already articulated the difficulty of this challenge on January 1st, today was the first day in many days where I emotionally feel connected to that challenge.
I have a renewed hunger, and 14 more days. Time to make it happen.
With renewed passion for this month’s challenge, I filmed five new solves today and performed surprisingly well. The average of the five was 22.414 seconds, with a single time of 21.587, only a couple seconds off my target time.
I somehow messed up filming the first solve, but I was able to capture the last four on camera.
I’m getting excitingly close.
Yesterday, I figured out how to consistently solve the cube at my top speed. I just can’t do it consistently.
Let me explain by analogy…
The “golfing” analogy
I don’t golf very often. Perhaps, once per year, at the most. But, when I do go, I continue to relearn the same lesson over and over: If I want to hit the ball hard, I shouldn’t try to hit the ball hard.
In other words, a slow, smooth swing almost always performs better than a fully-muscled shot. This is especially true when you don’t play golf, which I don’t.
Thus, when I go golfing, I make sure to remind myself to swing slowly before each shot. Yet, for the most part, I can’t seem to listen to my own advice.
A typical progression of golf shots
For the first shot, I swing smoothly, and the ball flies off my club. Awesome.
For the next shot, I swing smoothly, and the ball again travels far and straight.
For the third shot, I now feel pressure to replicate the success of my first two shots, and, as a result, swing harder (to “ensure” the ball goes just as far). Of course, the opposite happens, I mishit the ball, and it falls short.
Thus, to compensate for the short shot, I swing even harder on the next one, once again mishitting the ball. And so on.
For the rest of my time golfing, I try to escape this downward psychological spiral.
This is the odd thing about golf: Even though I know exactly what I need to do in order to be successful, I can’t seem to convince myself to consistently do it.
Speed cubing is no different
Speed cubing is very similar to golf in this way: If I spin the cube calmly and slowly, my solve-times are noticeably faster than when I spin the cube as fast as I can. This is something that I fully acknowledge and believe.
Nevertheless, like golf, it’s not so easy to spin the cube calmly and slowly — especially when I’m trying to break my own personal speed record on every attempt.
Here’s what usually happens: I spin the cube too quickly. My solve-time suffers. I spin the cube even faster the next time to compensate. I enter the golf-like downward spiral.
This problem becomes especially pronounced when I take out my camera to film a few solves. In these cases, I really want to perform well, which usually means I perform worse. (Read about how I use lube to curb performance anxiety).
Yesterday, then, was a bit of a miracle. I started filming myself, remained totally calm, and nearly surpassed my speed goal.
If I can maintain this same mindset, I should be able to execute a sub-20-second solve any day now.
For most of my posts this month, I’ve described how speed cubing challenges my spatial awareness, pattern recognition, discipline, etc. Basically, I’ve categorized speed cubing as a cerebral challenge more than anything else.
However, today, I couldn’t help but notice the physical / dexterous component to cubing: Unexpectedly, my cubing muscles were sore.
This past week, I started working out again at full intensity in preparation for my upcoming physical M2M challenges. In particular, among other things, I did a lot of pull-ups, which challenged my grip, forearm, and finger strength.
As a result, my fingers (by way of my forearms) were sore today. I didn’t think this would be a big deal, given that speed cubing doesn’t seem to require much physical strength, but I was wrong.
As I battled the soreness in my hands and arms, my solve-times clocked in about 3–5 seconds slower than yesterday. Not good.
Hopefully, the soreness goes away tomorrow, and I can have a few unencumbered attempts at completing this month’s challenge.
I’ll hold off on more pull-ups until I’m officially sub-20.
Today, since it wasn’t raining for the first time in a while, I spent most of the morning and afternoon walking laps around San Francisco listening to an audiobook. Currently, I’m listening to Spaceman by astronaut Mike Massimino (if you’re interested, here’s my book list from 2016).
I got home around 5:30pm, and decided to film a few Rubik’s Cube solves before dinner.
On my second attempt, I solved the cube in 20.720 seconds, only 0.73 seconds away from this month’s goal and my fastest solve on video so far.
The solve before this one, I completed in 21.787. The solve after, in 21.146.
I’m so consistently executing solves around the 20-second mark that I’m surprised I haven’t filmed a sub-20 attempt yet. It seems like there should be way more variance and luck involved here. Especially since my knowledge and execution abilities of the Rubik’s Cube algorithms aren’t uniformly distributed.
I’m not complaining about my consistency. I think it’s a good thing. It just doesn’t seem mathematically possible.
Anyway, I’ll continue focusing on systemically lowering my average solve time. That’s probably better than relying on variance anyway.
Today, I solved the Rubik’s Cube in 19.632 seconds, officially completing this month’s challenge.
This was especially exciting because I caught it on camera — or so I thought. It turns out my cinematography today was less than ideal: The cube is only framed halfway in the shot.
In other words, I turned on my camera, executed a sub-20-second solve, and still failed to fully capture the moment. Bit of a bummer.
So, while I’ve completed this month’s challenge (nice!), I’m not going to celebrate until I have a better video to prove it.
Until then, I’m going to zoom all the way out, and keep trying.
Yesterday, I completed this month’s challenge (I solved the Rubik’s Cube in 19.632 seconds), but I did it in the most unexciting way (the camera was poorly aimed).
Today, I decided to take a break from speed cubing and just relax. I’ll film some more solves tomorrow (with the hope that I capture a better-framed sub-20 solve).
In the meantime, today, I rewatched my recent videos, looking for ways in which I can improve, and there are many. In particular, I’m still being very sloppy: Although my lookahead is much better, I’m still relying too heavily on cube rotations, just because it feels more comfortable.
Tomorrow, I’ll make sure my first 20 solves are Forced Rotationless Solves(i.e. no cube rotations allowed) at full speed. Hopefully this will inspire some discipline…
Today, I solved the Rubik’s Cube in 17.704 seconds, 2.3 seconds faster than my sub-20 goal. This month’s challenge is officially complete!
Technically, I already completed the challenge two days ago, when I solved the cube in 19.632 second, but the cube was mostly out-of-frame during the video. So, today, I made sure to have the entire solve fully in frame.
While this month’s challenge is complete, I still have a number of Rubik’s Cube goals, like sub-20 average of five, sub-10, sub-60 blindfolded, and fewest moves. I’ll discuss these tomorrow…
Yesterday, I officially completed this month’s challenge, solving the Rubik’s Cube in 17.704 seconds.
And while this month’s challenge is over, I still have further Rubik’s Cube aspirations…
Sub-20 average of five
In competition, there are two main speed cubing events: single solve and average of five. Yesterday, I completed a sub-20 single solve (i.e. I solved the cube one time in under 20 seconds). However, I have not yet completed a sub-20 average of five (where the average of five solves, excluding the top and bottom times, is less than 20 seconds).
This is the next goal I’m striving towards. It’s even perhaps possible for me to achieve this before the end of January.
Sub-10 single
Solving the Rubik’s Cube in single-digit seconds is a feat only reserved for a minority of top speed cubers. I have a lot of work to do to cut my current time in half, but, perhaps with a little luck, I could do it within another month or two of practice.
Sub-60 Blindfolded
I’ve previously experimented with blindfolded solving, but I’m horribly slow (it takes me about 10 minutes to memorize the cube and 2 minutes to solve). I would love to be able to both memorize and solve in under 60 seconds.
Fewest Moves Challenge (FMC)
Outside the realm of speed cubing is a Rubik’s Cube event called Fewest Moves Challenge or FMC. With FMC, you are given a scrambled cube, paper and pencil, and one hours; and your objective is to find the most efficient solve for that particular scramble. In other words, the goal is to find the solve that uses the fewest moves (that aren’t just a reversal of the scramble itself).
To me, the part of FMC that is particularly interesting is called God’s Number. Basically, some researchers at Google determined that every possible Rubik’s Cube scramble can be completed in 20 moves or less. Thus, the number 20 is considered God’s Number (since god should be able to solve any Rubik’s Cube in no more than 20 moves).
The implications for FMC are clear: Can I beat God’s Number and find an optimal solve that is less than 20 moves?
The crazy thing is… in competition, three cubers (Marcel Peters, Tim Wong, and Vladislav Ushakov) have outperformed God’s Number and solved the cube in only 19 moves.
I would love to be the fourth.
Anyway, I share these challenges because I think they’re interesting, I’m inspired to pursue them, and I wanted to demonstrate the diversity of pursuits available even in such a narrow discipline like cubing.
Over the past 3.5 weeks, I’ve significantly improved my speed cubing abilities (to the point where I can consistently solve the cube in around 20 seconds).
However, it was hard for me to fully appreciate my progress, until I watched side-by-side videos of solves from Day 1 and Day 24 of this challenge (which are effectively the before and after videos).
Although, out of the gate, my speed seems reasonably similar, my efficiency on Day 24 dominates my Day 1 solve. In fact, the Day 24 solve is nearly three times faster than the solve from Day 1.
It’s pretty striking… (keep most of your attention on Day 24)
In the video where I solve the Rubik’s Cube in 17 seconds, before I start solving, I scramble the cube according to a series of 20 moves.
After sharing the video, many people asked where this scramble came from, if it’s random, if I’ve practice it before, etc. So, I thought I’d address that today…
I use an app called ChaoTimer to automatically generate new, random scrambles on every solve.
Each scramble is expressed using Basic Notation, which is depicted below.
Under the Basic Notation scheme, R means “turn the right face of the cube clockwise”. R’ (R prime), means “turn the right face of the cube counterclockwise”. R2 means “turn the right face of the cube 180 degrees”. And so on, with F = front; B = back; L = Left; R = Right; U = Up; D = Down.
You can see in the screen shot above that I’ve only solved the Rubik’s Cube 853 times this month while using the app. There are 43,252,003,274,489,856,000 possible Rubik’s Cube scrambles, all of which can be generated in 20 moves.
Thus, I’m nearly certain that I’ve never had the same scramble twice.
Anyway, if you’re interested in speed cubing, I’d highly recommend downloading the ChaoTimer app. Not only does it generate scrambles, but it also keeps time and crunches a bunch of statistics about your solves.
Even though I surpassed my speed cubing goal four days ago, I continue to spend the same amount of time every day solving the Rubik’s Cube.
However, since completing this month’s challenge, I haven’t made any noticeable improvements to my skills. In other words, I’m spending the same amount of time speed cubing, but enjoying none of the gains I saw previously.
Sure, it feels like I’m “practicing”, but I’m not really practicing. Instead, I’m just going through the motions that are now comfortable and easy. My relationship with the Rubik’s Cube has reverted back to tactile distraction, and nothing more.
I’m not saying this is a bad thing: I’m happy to enjoy the cube in this more relaxed way. But, it does help me appreciate the value of painfully deliberate practice.
The truth is, to improve at anything, especially in a short period of time, it’s necessary to endure a certain kind of pain while practicing. I don’t think it’s a bad kind of pain (it feels good, like going to the gym), but it takes a lot of willpower to enter into and sustain this state of pain for an entire practice session.
Since I no longer have a measurable goal to pursue or a desired timeframe to contain this non-existent pursuit, it’s much harder to motivate the kind of practice that yields noticeable changes from day to day.
Luckily, I have the Month to Master framework in place to help enforce, structure, and inspire deliberate practice most of the time. In a few days, a new month will start, and I will regain my hunger to endure the happy kind of pain.
Until then, I’ll continue to enjoy my visit to the land of casual cubing…
Today, I wanted to do something creative. So, in the spirit of cubing, I made a giant floating cube out of blue tape. It took me about 40 minutes.
I honestly don’t have any further explanation beyond “tape art is apparently a good creative outlet”…
Here’s the cube from some illusion-breaking angles:
And here’s a timelapse of the creation process:
That’s it for today.
Since this month is almost over, I decided to go through my practice log today and tally up the time I spent on the Rubik’s Cube challenge. Here’s what I found…
In January, I solved the Rubik’s Cube 907 times. For each recorded solve, I averaged 25.4 seconds solving, 15 seconds scrambling, and 15 seconds inspecting. This equates to about 14 hours.
I also spent another 5 hours learning new Rubik’s Cube algorithms, and another 3 hours completing other training exercises.
So, in total, I spent 22 hours on this month’s challenge. In other words, every hour I practiced, I decreased my solve time by about 4%, which compounded quickly, until I executed a 17 second solve.
Today is the last day of January, which means it’s also the last day of “Rubik’s Cube Month”.
And while this project shifts its focus elsewhere, I certainly will continue cubing. There’s just something about it that’s so meditative. I get completely engrossed in it.
Sometimes, cubing enables the quiet the mind kind of meditation, but, more usually, it gets my mind into that actively creative state typically reserved for hot showers.
Especially now that I’ve softened up on my training, and instead, am relying on mostly automatically reflexes to solve the cube, I’m able to enter into that shower-like trance often. It’s a great feeling to control.
Of course, everyone’s experience with the Rubik’s Cube is different, and if you’re interested in learning more about the perspectives and stories of other speed cubers, I’d recommend checking out the book Cracking the Cube.
In the book, author Ian Scheffler integrates himself into the competitive world of speed cubing, sharing amusing stories about cubing’s main characters and weaving in his own pursuit of trying to break the 20-second speed threshold (It took Ian about 1.5 years to get his time from around 50 seconds to 20 seconds).
Ian also had the chance to interview Erno Rubik for his book, the inventor of the cube, which is also an interesting story.
Anyway, if you are still itching to learn more about the cube, this book is probably the best place to continue. If not, tomorrow I start a new M2M challenge: Landing a standing backflip.
