Home
Methods
Opinions
Videos
Shop
Sponsor Invitation
Snyder Cube Challenge
Copycats

Tony Snyder's Cube Solving Hobby


Methods Page

How I Solved It In 16 Turns
snydermind
speedsolving wiki

Introduction

This page introduces my cube solving methods, plus a bit of history on myself and how these methods came about. You can click on the individual links to get more technical information.

Snyder Method 1 (1981)

aka Tony Snyder's Simplest Solution to the Rubik's Cube

Back in June of 1981 I came up with a basic set of stages and algorithms within 2 hours of first picking up the cube, and I did this without trying to solve it first. I simply got out a few pieces of paper and figured it out with a pencil while studying how it was designed. I took it apart to see how the pieces worked, put it back together, but I did not mix and try to solve it until after I first designed a method for this on paper. I wrote this all down long hand, using crazy looking pictures instead of standard notation (I didn't even know about notation at this point). Then using this method my first solve took about 2 hours, as I had to adjust my thinking here and there. The original algorithms for this method were inefficient. So over the following few days I worked and worked until I figured out better algorithms to fit those same stages. The result was a solution that was very easy to learn and yet relatively turn efficient, making possible 40 second solves. This paper got me an A in physics, an A in English, and an A in computer science (yes, one paper for three class projects, lol). And I won the First Annual Semiahmoo International Cube-A-Thon using this method.

Snyder Method 2 (1982)

In latter 1982 I added some new innovations to my method. At first these changes were quite tricky, but as I got the hang of it I discovered a whole new world in turn savings and tricks, allowing my times to improve to sub 20 seconds within a few short weeks. I used to call this my advanced method, but now I call it Snyder Method 2.

My philosophy in cube solving from this point forward was to always perform a simultaneous permute and orient with all pieces at each stage. I'm convinced that mathematically this reduces the total turns to solve the cube (as compared to techniques that permute and orient in separate steps), and I will attempt to prove this over the coming months.

However, another advantage is that it trains my mind to better see a handy direct solve if I'm already close. This is because I am using algorithms that directly place and orient pieces in one step, so as I watch this happen I get used to how that looks, and eventually start to understand how to see a direct solve. You can get the same benefit from mixing a cube 3 turns, then solving, over and over, then increasing to 4 and doing that over and over, then to 5, etc. This type of practice helps you to understand the cube, which I think is far more valuable than pure memorization of algorithms (however, if speed is your only objective then it may be counter productive, depending on whether or not you can think fast enough for a creative speed solve). As a result of this type of practice, I have in recent years achieved two 19 turn solves, an 18 turn solve, a 17 turn solve, and one 16 turn solve (this is unofficially the best in the world). Each of these took me only a few seconds of study, and a few seconds to solve. However, though these solves roughly resembled Snyder Method 2, they were mostly the result of creative on-the-fly thinking, which is something that I usually do a little, and on occasion take to the extreme. Most of the time I don't have enough focus to pull that off. This to me is quite interesting, because I know that when I do have that kind of focus that something happens with my mind, its like I'm half asleep, or where the conscious mind somehow shifts half way to the non-linear areas used in sleep. I also use this non-linear mode when I'm half asleep to conquer the more difficult challenges in software engineering.

Here's the trick: first list all the variables and objectives, using as clear and concise language as possible, while NOT attempting to solve the problem. Then go to sleep, then wake up half way and visualize the answer. This technique has worked many times for me, and I have used it to come up with new ideas and to think through the details to those ideas. However, for me to trigger the ability spontaneously during a solve is rare.

I used Snyder Method 2 at the 1982 2nd Annual Ideal Toy Corp Cube-a-Thon in Chicago to solve the cube in 11-12 seconds during an elimination round. Unfortunately, my final place time was not that impressive due to a number of issues: their timers didn't work, the guy stopped his stopwatch about 3 seconds after I was done, I had to use their very loose cube that was nearly impossible to keep from falling apart, I only had a couple weeks practice on my new method, and I have a thing with crowds as my fingers tend to lock up and I have to focus more on palming than fingering when the camera is on me. Just the same my bad final time of 31 seconds took first place anyway. In contrast, the elimination round was easy for me because I was allowed to use my own cube, and as others were solving their cubes at the same time I had no problem with the camera. This is when I solved it in roughly 11-12 seconds, causing the announcer to go stone quiet for nearly a minute in front of a thousand people. Nevertheless, my final place time missed That's Incredible by 1.02 seconds. This was frustrating because removing any one of those issues would have easily made up that difference, not to mention the fact that the guy who beat me did so weeks later (at a different location), long after my average had improved considerably. He would have had no chance against me if we competed face-to-face.

I continued to improve steadily over the following months, while practicing my new method roughly 5 hours per day. By the time the 23 second world record was set I was consistently averaging between 15 and 17 seconds, and had a best time of about 9 seconds. Then about a year later I got my first 8 second time.

I didn't beat that until 2004, when I solved it in what appeared to be 6.5 seconds using someone's analog watch, and I then added 0.5 to comply with mainstream opinion regarding times taken with an analog timer (I agree with this one), arriving at a time of 7.0. The cube was fully mixed according to current rules, and later that same week I got a 7.5 in the same way.

Using Snyder Method 2 I generally hit the 10 second mark once or twice per day, and speeds faster than that would only occur once in a rare while.

It may be difficult for modern cube solvers to make sense of these claims. This is because mathematically balanced solving methods combined with modern fingering result in incredibly consistent solve times. In contrast, Snyder Method 2 is something I thought up from my mind, it is not at all balanced mathematically, and I use mostly wristing. Combine these factors with the intermittent issues I have with my fingers, plus switching at times to a special mind state, and my times are all over the board.

For several years I could barely feel the cube with my right hand (due to nerve damage), now nerve reception is intermittently better, and my times are improving yet still inconsistent. Where fast fingering used to come natural, I now have to really work at it, oftentimes with 2 hours warm-up plus ideal conditions to get the cube to turn fast again. Right now on a good day I'll average around 16-18 seconds despite all these issues.

Here are all the handicaps: nerve damage and misoriented bones in my right hand and wrist, I'm 48 years old, I use old school wristing rather than modern fingering, my algorithms are nearly all human originated, I have Asperger Syndrome, as the inventor of this method I continually rethink what I'm doing rather than using pure memorization, and I get nervous easily when people are watching (this is a weird one, because though I'm consciously not afraid of crowds, all I have to do is think about people watching me and I screw up).

However, I think probably the biggest handicap is mental distraction. On average I make all kinds of fingering mistakes and turn mistakes, and in retracing my thoughts I find that it was because I started to think about something else in the middle of the solve, and other times this happened because I needed time to decide which optional algorithm was better. I tend to decide these things as I go rather than planning it all out in advance within a documented technique. And when I time myself I make even more mistakes - there's obviously some subconscious issue there. I may get to the point of going extremely fast repeatedly (estimating 9-12 seconds per solve), then I pull out the timer and immediately my times are bad again. I have to conquer this one before going back to the contests.

Regardless of the inconsistencies, I'm convinced that my solve speed was approximately the best in the world from the Spring of 1983 through the Fall of 2004. That's 21 years that the rest of the world took to catch up with me.

Advantages

One of the key advantages to my method is turn efficiency. Once fully mastered you should be able to average less than 40 turns in a speed solve. Then there's the long-term advantage. By always learning algs that directly orient and permute in one maneuver, you are conserving your learning curve, because all of these algs will still be useful to you once you do a consistent LL direct solve. In contrast, the 2nd to last stage of the Fridrich Method teaches you how to orient and permute separately. And with the orienting stage, these algs will never be needed again once you are able to do a consistent LL direct solve.

For those who don't already know, a last layer direct solve requires knowledge of over 1200 unique cases, and without any tricks that means memorization of over 1200 algorithms. Most people don't want to even imagine attempting this, however, I am convinced it can be done through a combination of tricks in grouping these algorithms for easier memorization. Some of these tricks result in a true direct solve, while others result in an approximate direct solve. Either way, in my opinion enough turns are saved that it is worth the effort.

My point in not switching to Fridrich is that I plan on evolving my method towards a full LL direct solve, and that means only learning algs that simultaneously permute and orient in one step. If I first switch to Fridrich, then I'd be learning algs that orient without permute, and those have no future purpose.

Disadvantages

The last layer is seriously imbalanced mathematically. The 2nd to last stage requires far more algs for memorization than the last stage. Basically, the more imbalanced the math, the more algs total that you have to learn to gain the same turn advantage. Balanced math = more efficient algorithm utilization, and more consistent turn counts and times.

However, the long term goal of compensating for these scenarios with more LL direct solve algs ultimately removes these disadvantages. The idea is to focus on memorization of the direct-solve algs that specifically deal with worst case scenarios, thereby balancing out the math.

There is also a disadvantage in rapidly recognizing scenarios. When you get to the last layer, you need to take in the positions and orientations of all LL pieces in the first look. In comparison, with the Fridrich system you only need to look at the yellow stickers in the first look - this is so much easier, as you can focus on just one color. The Fridrich LL method is far superior when comparing ease of pattern recognition. However, if your long term goal is a one-look LL, then taking in all the LL orientations and positions is something to start practicing anyway.

Conclusions

To minimize the learning curve in becoming a master cube solver, Fridrich LL is the way to go. My method does not compare when a short learning curve is one of the objectives. However, if you feel like a challenge, and think you can learn a lot of LL direct-solve algs, then Snyder Method 2 may be a more efficient way of achieving that goal. From the start you'll be looking for patterns and using algorithms that will always be useful, and it paves the way for incremental improvements from there. Granted there is a lot of work to do before you learn enough LL direct-solve algs to patch up the imbalanced math, but its worth it for those die-hards (like myself) who are looking for a turn advantage over Fridrich LL.

Ultimately I hope this approach will help to balance the need for fast fingers to the need for sophisticated technique, making them of equal value to a speed solve. In contrast, I'm convinced that block building + Fridrich LL + modern fingering (the current best speed solve approach) results in finger speed impacting the resulting time by around 80%, while method sophistication impacts the solve time by around 20%. In order to swing this back to 50/50 I see as necessary a consistent LL direct solve using a variety of tricks to aid in memorization and execution, plus a major improvement to the block building process.

The current version of Snyder Method 2 achieves an approximate direct solve on the last layer around 20% of the time, and without having to memorize 20% of the last layer direct solve algs. Instead I use a number of tricks to accomplish an approximate direct solve in many of these cases. As a result I average about 40 turns on a speed solve, or about 36 turns when I take extra time to think it through. And I do this using almost entirely my own human originated algorithms. Once I apply software analysis it will be easy to shave more turns from here. I will then have the tools I need to document all the tricks necessary to a consistent LL direct solve, and a major improvement to block building. At that point Snyder Method 2 should become a real contender on speed solves, and make possible world records without as much a need for fast fingers.

Snyder Method 3 (2011-2012)

I'm convinced that I have the right ideas that once perfected (through use of software modeling) will produce the very best human cube solving method in the world, both for speed solves and for FMC. This is because Snyder Method 3 will drop the average turn count for a speed solve to around 30 turns, and it will do this with fewer looks. Snyder Method 3 will be harder to learn than the other methods, but it will be the very best method, making possible faster speed solves than ever before, great CC solves, and near perfect FMC solves.

While my previous methods were thought up entirely from my mind, Snyder Method 3 will be produced entirely from decisions made by software. The software will decide the best stages, the best algorithms, and even the best finger tricks, to achieve the fastest possible human speed solves.

Questions? Send me an e-mail: tony@snydermind.com

home
snydermind
top