Speed Blindfolded Solving

Solving a Rubik's Cube blindfolded using a regular F2L (i.e. Fridrich) method.
Written March 12-25, 2006
My current records: 8:59 memorization / 11.88 execution

IntroductionThe Basic IdeaExamplesCrossF2L: TracingFinishing up F2LF2L: Tips and TricksOLLCompound OLLPLL

This page will render incorrectly and slowly in Internet Explorer; use Firefox to view it.
You will also need a good computer running Java to view the demonstrative applets.

I've considered writing this guide for a while, explaining what I do for a speed BLD solve (this is currently not an official event, but I currently hold the unofficial world records for memorization and execution), and Mátyás Kuti asked me in an email, so here it is:


Note: This is not related the regular cycle method for blindfolded solving, to which most of my BLD pages pertain (you can reach them from the index). If you're just learning blindfolded solving, I really recommend Macky's page for 3-cycles. However, this method is essentially performing your regular method in your head, so if you know your Fridrich well, give it a try:

This page is intended as a guide for those who want to solve a Rubik's cube blindfolded using their regular method, or just perform the solving portion quickly quickly: "speed blindfolded solving" (that is, going for fastest execution). Since most advanced cubing methods are designed for speed, this amounts to the same thing (until someone comes up with a new idea).
This guide does pertain directly to F2L, and the concept of "looks" (as in Fridrich being a 7-look method). If you use Petrus, or CF, or Roux (especially if you do the last steps as a combination of memorized algorithms), this can still apply, but you'll probably need to do more thinking to track each move. Also, if you use a method with many steps/looks, such as layer-by-layer or one with a very long last layer, you will have to remember a much longer solution -it will take longer to memorize, and thus you will be more prone to errors.

By the way, throughout this, I'm postulating that even though you may have the cube available to look at in front of you during the entire memorization time, you cannot take any physical notes (though that might help, for practice). In addition to not using notes, you should not allow yourself to look up an algorithm, practice it on another cube, or twist the cube at all. The idea is to do everything in your head, and you should strive for that. Anything else is like lifting the blindfold during a regular BLD to make sure you're correct.
However, don't be discouraged from using aides during the first few times (some people try that for regular BLD). It's not as pure, but it can be reassuring. If you think it helps a lot, go ahead and take notes the first time you try it. But if you want to go for the entire challenge the first time (which always feels much cooler), be bold.
If you can already solve the cube blindfolded, you can actually "compute" the solve without looking at the cube: Memorize the cube using your regular method. When you need to find a piece to trace, recreate its location and orientation from your memorization info (this is actually quite interesting, and a lot more fun than counting sheep). I think it's quite possible to get sub-1 memorization+execution this way, even as low as 40-45 seconds.

A word on solving style: You might try this for several reasons; to use exactly your regular method to solve the cube blindfolded, to just use a speed method to blindsolve, to mentally create a solution as quickly as possible (speedcubing.com category), to solve the cube as quickly as possible while blindfolded (not a category, though it deserves to be), or a combination of any thereof. Speed memo versus speed execution is an interesting issue. They are related; a good constructed solution will help both. But in one (speed memo), you want to do as little computing as possible; for the other, you want to explore as many opportunities as possible to choose the fastest one. Depending on your goal, you will approach the solve slightly differently, but the same concepts (especially tracing) apply.

The Basic Idea

A regular F2L solve has 7-8 looks, you recognize at each step and execute. For speed BLD, you do the same, but you do each look separately: each look now consists of tracing the necessary pieces through the current partial solution, figuring out what moves/which alg will solve (place, orient...) them, and then continuing for the next step, with the solution extended. You don't trace all the stickers or perform each move in your head and rememorize the entire cube; it's more a process of computing, hashing, and caching. You need to keep track of at most 8 pieces at a time (less will work, but will take more time).
It might seem daunting, but it will not be difficult to remember the solution you construct; you spend a lot of time on each step. During the first few tries, I would recommend recalling the current solution often, until you trust yourself to remember it.
The cross is straightforward; just do it like you normally would: inspect and memorize the moves. Then follow each F2L pair, and "recognize" it in your head (if an F2L is difficult, try another). For LL, trace the pieces individually through each step, and try to recognize PLL in terms of cycles.
If you think you understand it, I really encourage you to try a solve now. Maybe just cross and one F2L slot -if you really like it, you can just continue. As with a lot of cubing, it's much better if you figured out something out yourself and completely understand what you're doing.


To demonstrate the ideas of a speedblindfolded (nice adjective, isn't it?) solve, I will use two examples throughout this page, both of which I successfully speedblindsolved (I got this word from Kuti) myself. They are pretty typical and will convey all the necessary concepts. I will primarily use example 1, though. Their scrambles are below:

Examples: Scramble
#1: The 19.69#2: The 18.04
Scramble (white on top, red in front):
U' L2 D F2 U' B2 R U' B' D' B R' U2 R' U2 D2 R2 L'
Scramble (yellow on top, orange in front):
F2 U' R2 D' L2 U' F2 D2 U B R' U L' D' F D' F U' B L2
Clicking on a cubie in the example applets will move it (and its layer) in the direction of mouse movement. Right-clicking or clicking outside the cube will rotate it. The bottom buttons are for navigation (reset, go back one move, play backwards, stop/mirror, play, move once, skip to the end).