Cubing resources are easy to procure, but they often lack the proper annotations to explain the context -- When should I learn this? How much faster will I be? Is X the only way to Y? What sort of progression should I expect? Is this the best quality resource out there? As a result, cubers might get confused about what to learn next, or spent time learning stuff way too advanced for their level.

That's the motivation behind the Roux Reader. It hopes to fill this void for Roux method speedsolvers: by being the missing user manual that explains how to pick the top quality resources that are best suited to your skill level, and how to set expectations when learning them.

Content-wise, this reader is a tight coupling of two things: an improvement guide and a resource collection. The writers offer opinions to help you pick the right resources and answer all the common FAQs you might have, with the ultimate goal that you can walk away with a much better idea of where and how to improve your times next.


How to Read the Reader?

First off, this reader is written for Roux users of all levels, as long as you know beginner Roux. (If not, check out this awesome beginner tutorial guide here).

There isn't a single recommended way of reading it, but here are some useful scenarios:

  • If you already have a topic in mind (e.g. FB turning, EOLR progression), go find the related sections, and read the explanations to help you select the best resources.

  • If you are just randomly looking for things to learn, skim through the book for whatever interests you, and read up the section overview to discover exciting skills, techniques or concepts to work on.

  • If you have a particular question in mind, it might have been asked by many before. That's why we have created the FAQ chapter. Try your luck there!


At the top level, the reader is organized by the four steps in Roux -- FB, SB, CMLL & LSE, plus a few chapters for generic topics that don't belong to a step. In Blockbuilding, we go over the three fundamental aspects of speedcubing -- solutions, fingertricks, and lookahead. For each of these subsections, we first provide an overview of what a fast cuber should be ideally doing, and talk about the progression -- what do I learn at different skill levels. This overview is followed by a curated list of resources with commentaries. In CMLL and LSE, this structure is generally followed, with the additional emphasis on sub-step variants and their algorithms. Lastly, while OH and 2H techniques can differ sometimes, they are usually discussed within the same subsections to avoid fragmentation.

How to Contribute

If you wish to contribute to the writing, feel free to contact any one of the authors. Please refer to the organization section above when making edits. A basic knowledge of Github and running simple commands from the terminal would be helpful. The workflow is simple enough. Basically, download the binary for mdbook, clone this repo and make edits to the Markdown files locally, generate the book by running mdbook serve on the command line, and submit a pull request when you're done! The book will be automatically generated using a CI workflow once new changes are checked in.

If you have any questions / doubts / suggestions to any content, please also reach out so we can integrate your feedback in future edits. Feedback is super important to us!

Version Log / Roadmap

  • version 0.1
    • This project is still in its early phase as of Sept 22. (v0.1) We hope to get most of the content up there in a few weeks, so stay tuned for updates!


This chapter talks about how to improve blockbuilding. We'll go over orientation, first block, second block, and alternative blockbuilding stragies.

General blockbuilding strategies

Read about blockbuilding for examples and explanations of the basic blockbuilding strategies.

First block vs. Second block: A brief overview

Coming from the beginner's method, you might be employing similar techniques of loading the edge and pairing it with the corner for both FB and SB. However, going forward, you should realize FB and SB are fundamentally different steps that give rise to different sets of challenges:

  • FB planning is static

    • You use inspection to thoroughly examine the cube and carefully craft an efficient solution.
  • SB solving is dynamic

    • You track pieces in real time and make decisions on the fly, and you might sacrifice efficiency for better fluency and lookahead.

Choosing your Orientation

Orientation describes what FB locations you would consider in inspection, i.e. what are the possible bottom and side colors for your FB. Orientation is usually denoted by one or two rotations: once you fix a particular FB you solve, all the other FBs you use can be reached by a combination of these rotations. For instance, x2y means all FBs that are {y, y', y2, x2, x2y, x2y', x2y2} away from the reference point.

Among the commonly seen orientations, x2y offers 8 FB options, while both y and x2y2 give 4. x2y is highly recommended. It is the minimal orientation that enables you to use ANY premade pairs. If you have to adopt y or x2y2 instead, the probability is reduced by half, so you'll get more difficult cases on average, but it's still somewhat acceptable. Anything below y or x2y2 would heavily limit your choices, and is recommended against.

Regarding CN: it gives you 24 options for FB, which is definitely a plus over x2y, but the benefits have not been proven as there has yet to be a world class full CN Rouxer. Some speculate the benefits would be minimal because there is limited time during inspection to analyse all 24 FB options. However, a counterpoint would be that for any first square, you have two choices of last pair to extend to a FB, making the last pair case much better on average. Therefore, once you find a good first square, you can usually settle on it and make better use of your inspection time by looking further into the solve. TL;DR: CN Roux is worthy of exploring, but don't force yourself.

If you are interested in the average movecount comparison of FB or FB + DR under different orientations, use the following resource:

  • Movecount statistics - Notice how x2y boosts the chance of an easy first square compared to y from 60% to 80%.

First Block


Step I: Basic Blockbuilding Strategies

Transitioning from the beginner method of edge loading spots, you should be aware of different blockbuilding strategies. There are broadly speaking two families of FB building strategy: square + pair and line + line, with the first one being more frequently used.

It is recommended to watch videos on blockbuilding strategies and have a basic understanding before you proceed to example solves.

Overall strategy:

Beginner-friendly Example Solves:

Relationship to Planning

Before we go on, note that planning = solution + tracking, with solution playing perhaps a more important role. For now stay tuned on getting your solutions right, and we'll cover tracking in the last section.

Understanding relation of each FB pair's pieces and how they can be solved

This is crucial for understanding many of the influencing solutions and determining how to solve second stage after predicting first stage.

The minimum information needed to know how to solve a FB pair:

  • FB pair's edge position and orientation
  • FB pair's corner's position and location of its D-face sticker

For example, if the front pair's corner is on the U layer with its D-face sticker also on the U layer, then to have the corner be in the same alignment with its edge (such that when they're adjacent, they form a correctly connected pair):

  • the edge can be placed in FR in an orientation such that it is an F2 from being solved.
  • the edge can be placed misoriented in its solved position (FL).
  • the edge can be placed in BR such that the edge will be solved with R2 F2.

These relations can be understood through experience and observation. Using a trainer as well as always observing the aforementioned minimum information of an FB pair before solving it helps.

The above logic can be applied to line solutions as well, especially with understanding how to pair the DL edge with an FB corner to build the D-line. For example:

  • DL edge is in DF, and is a D' move away from being solved.
  • FB's front corner is in UFR, with its D-face sticker on the R layer.
  • Thus, the corner can be connected with the DL edge using R U R'.

Step II: Improve Your Solutions

Next, you want to improve your solutions. There are several aspects you can work on, in no particular order. Not only are easier solutions faster to execute, but the ability to come up with them will also help bring down your inspection time.

  • Influencing
    • square + pair and line + line are both two-stage solutions, and the point is we want to find alternative ways to solve the first stage i.e. square/E-line so we get a better second stage case i.e. pair/D-line, respectively. 1
    • Influencing involves altering your initial first stage solution so that the second stage case is better than without any influencing, and your overall FB solution is also better.
    • In general, influencing can be one of the following:
      1. Solving the same first stage differently for a better second stage, as seen in example 1 and example 2.
      2. Adding influencing moves to an already existing first stage solution.
    • Influencing moves can be used to move second stage pieces with 2 different intents:
      1. So that they'll be affected by first stage solution moves to achieve better orientation, as seen in the majority of examples.
      2. So that they won't be affected by first stage solution moves, such as when second stage pieces are in a good orientation or alignment (either before or after first stage), and first stage ruins that characteristic, as seen in example 3.
    • Typically in first stage solutions, one of the final pieces of the second stage (edge or corner) cannot be influenced, or is not worth influencing. Predict where the uninfluencable piece will end up after the first stage, and attempt to influence the influencable piece to improve the second stage case.
    • Your influencing moveset will be restricted in some way so it doesn't affect your first stage solution, so visualise the effect of moves within this moveset on influencable pieces to figure out your influencing moves (if any).
    • Note that influencing is not always applicable or necessary.

Influencing examples:

Example 1:

U r B' builds the back square.

However, U2 R B2 as a first stage solution leads to a much shorter second stage solution.

Example 2

D' M U R2 builds the D-line, although the final piece of E-line (BL) is in a bad position.

Using r' instead of M leverages the R2 to orient BL for an easy B2 insertion.

Example 3

F2 builds the E-line, with DL being misoriented in the D-line.

Starting with M' causes DL to not be affected by the F2, and become aligned with the D-line after the first stage.

Example 4

B builds the back square.

Starting with R' sets up a much easier last pair solution.

Additionally, R B M2 F is another FB solution with influencing.

Example 5

D builds the back square.

Starting with F leverages the D to connect the front pair's corner with its edge.

Example 6

F builds the front square.

Starting with U2 leverages the F to align the back pair's corner with its edge.

Example 7

L2 D' builds the front square.

Starting with U2 leverages the L2 to connect the back pair's corner with its edge.

Example 8

B' D builds the back square.

Starting with L leverages the D to align the front pair's corner with its edge.

Example 9

L' B2 builds the bottom line of FB and solves the blue-red edge.

Starting with B' leverages the B2 to also insert the back edge, for a line + line block.

Example 10

R' D builds the back square.

Starting with U leverages the R' to align the front pair's edge with its corner, building the front pair after the D.

Example 11

D B builds the back square.

Starting with R leverages the B to align the front pair's edge with its corner.

Example 12

F D' builds the back square.

Starting with U2 leverages the F to align the front pair's corner to its edge after the edge moves from the D'.

Example 13

F' D builds the front square. Back pair's corner is clearly uninfluencable, so predict the back pair's corner's position and orientation, and then aim to influence the back pair's edge.

We can predict the back pair's corner will be at DBR after first stage, with its D-face sticker on the bottom.

We should know from experience that the back pair's edge needs to be misoriented in BR to pair with the back pair's corner when the back pair's corner is in DBR with its D-face sticker on the bottom.

Thus, starting with R connects the back pair's edge and corner after first stage, for an easy 3 move second stage.

Example 14

D' B' D builds the back square.

Adding an L after the D' leverages the D to align the front pair's edge with its corner.

Example 15

F' U F' builds the front square.

Starting with B' leverages the U to align the back pair's corner to its edge, and pairs them after the final F'.

Example 16

L' U L D builds the front square.

During the L' move, the final pair's corner is U2 away from being connected with its edge.

Starting with U' makes us do that desired U2 after the L' to insert the front pair's edge whilst connecting the back pair's corner with its edge.

Example 17

r U' M2 D' builds the front square. Back pair's edge can be predicted to be only a B move away from being solved after first stage. Without influencing, the second stage solution is 3 moves.

In an attempt to influence, we can try seeing the effect of specific U moves before first stage on the back pair's corner (U chosen as it doesn't affect first stage, rather than some B move):

  • U leads to the back pair's corner being aligned with its edge after first stage, but this is still a 3 move second stage solution, so overall solution is longer.
  • Similarly with U', we have a 3 move second stage solution, so overall solution is longer.
  • But with U2, the back pair's corner will be in ULF with its D-face sticker on the L face; this means it's a B move away from being solved, like the back pair's edge, and thus results in a 1 move second stage solution, saving 1 move overall.

Thus, start with a U2.

Note that this solution may have been harder to figure out if we tried R U' M2 D' to build the front square.

  • Optimize your Last Pair solution

    • FB last pair has lots of cases and elegant solutions that can escape your notice. Drill on these in the following ways: use trainers to get random cases and reference the solutions; do untimed solves and experiment around with different ways to solve the same LP case.
  • Watch Example Solves

    • Trainers can only help you so far -- example FB solves remain the best way to learn new ideas. Either watch example solves or go over text reconstructions.
  • Develop a Taste for good FB / FS

    • Taste is about being able to tell an easy FB from a hard one quickly without fully planning it out. This would help you take advantage of all the FB options given by your orientation.


  • Kian's Example Solves

Reconstructions (text):

  • Search for Kian, Alex or Sean's solves on speedcubing databases:
    • SpeedCubeDB - a modern database, roux solves yet to be added
    • CubeSolves - plenty of solves, but no longer maintained
  • More reconstructions can be found on Anto's subreddit


  • You want to have calm and fluid turning for two reasons:

    • You need to give yourself time to track SB pieces, and turning too fast can result in losing track of pieces.
    • FB is a bad moveset group that generally involve both F and B moves. It is intrinsically more difficult to fingertrick. If you rush, you could overshoot or get your grips wrong, resulting in bad lockups.
  • Plan out your fingertricks in inspection too:

    • Unlike the rest of the solve, FB fingertricking requires your active attention. You should limit your regrips and eliminate pauses between moves. Try to figure all these out in inspection --- This can be just as important as planning out the actual solution! Over time, you'll grow familiar and be able to plan fingertricks automatically.
  • Also, develop a preference for what pairs / triggers are easy for you to execute over time, and try to prioritize them in your FB solution. Use this to guide your FB planning instead of move count as this is more precise and correlated to actual times.

FB Planning / Piece Tracking

  • If you cannot plan FB entirely, try to plan FS first.
  • Planning LP:
    • Only need to calculate:
      • The edge's position and orientation after FS.
      • The corner's position and location of its D-face sticker (white or yellow if x2y) after FS.
    • From the above information, one should be able to determine the LP case and solve accordingly.
    • For example, if the LP edge (where LP is the front pair) is an F move from being solved, the LP corner can be placed at DFR with its D-face sticker on the R face to solve LP with an F move.
    • If you calculate that the LP case is bad or mediocre, then try to influence.
  • Similar logic applies with planning FB line solutions or two-stage solutions in general:
    1. Determine solution for the first stage.
    2. Calculate required position/orientation information of second stage pieces.
    3. Attempt influencing for better overall FB, factoring in fingertrickiness.
    4. Execute!
  • Tracking Trainer by Zhouheng
  • Partial SpeedBLD technique by Kian - This is arguably underrated as a practice approach. You should try to do them regularly as warmup before solves.

The "line + line" strategy refers to two lines: E-line (2 edges and 1 center on the E slice) and D-line (1 edge and 2 corners on the D slice). E-lines are usually formed first.

Second Block

Blockbuild Strategy: DR First vs. Freestyle?

DR first is used by most of the world-level rouxers today including Kian Mansour and Sean Villanueva. Most of the existing SB resources are built on top of it. It is easy to make progress with, as it greatly simplifies tracking and pattern recognition for SB blockbuilding.

Freestyle, in contrast, allows you to be flexible with how you build the second square. It is a superset of DR first. Strategies include make a pair then insert the remaining edge, or insert any edge then insert the remaining pair.

The consensus is while freestyle potentially leads to better efficiency, it is more demanding of your recognition and tracking skills1. Empirically, it has yet to be proven that freestyle can serve as a standalone strategy to attain a speed level comparable to those using DR-first.

In general, DR first is recommended to beginners and advanced cubers alike as the dominant strategy, while freestyle can act as a supplement, offering shortcuts to building SS whenever you happen to recognize a few easy freestyle patterns as they show up. These SS techniques will be mentioned in the section below.


Pair Progression

Pair Influencing

SS Shortcuts and Freestyle Building

Gilles Roux's website [insert link] features some of these algorithms.



FB-SB Transition

NMCMLL and Variants

1: (Most SS strategies are such that once you solve the starting piece/pieces the SS becomes fixed meaning you're left with no options or freedom to select in what order to continue to block-build, whereas DR-first is the only exception among these strategies to NOT fix the SS but rather enables you to choose any of the two remaining pairs (SP). The very existence of two competing options alone would boost the chance of there being a good case, and additionally create room for influencing techniques. Therefore, in order to beat DR first on average, you would need more information about SS, e.g you saw DR is bad and another edge is good and the corresponding SS corner can be paired in such a way that gives an easy SS finish, etc. )

Advanced: FB+DR & more

( in progress)


This chapter is about solving the Last 10 Pieces (L10P), which is divided into CMLL and Last Six Edges (LSE). We'll cover the algorithms, fingertricks, and various efficiency techniques that will help you succeed in this second part of Roux.

General Information


  • There are only 42 cases in CMLL, which isn't bad to learn compared to OLL + PLL in CFOP. See it as a motivation to begin learning full CMLL now!

  • Since the M slice is free in Roux, Rw moves are incorporated in CMLL algorithms, sometimes producing more efficient solutions to the cases.


  • LSE is 2-gen, meaning that only two kinds of moves are needed to solve this step, namely M and U moves. This gives a great advantage to one-handed solving.

  • LSE is intuitive. Avoid learning it by brute-force memorizing to make your life easier.

Transitioning from CMLL to LSE

For high-level solvers, a fast transition from CMLL to LSE is a key to faster solves. We'll discuss ways to improve this transition.


Recognition Method: From beginner's 4-tile to all angles

CMLL recognition is a little more tricky than something like OLL. Not only does one have to pay attention to the orinetation of the corners, one also has to pay attention to their colors. Despite that, an advanced Roux solver is able to recognize CMLL cases within fractions of a second. Here is the progression towards that.

I. Beginner Recognition

Most beginners use 2-look CMLL, so the colors of the corners are mostly irrelevant. As a result, it is adequate to simply get used to recognizing the orientation of the corners.

Practice recognizing the corner orientations fast - this will definitely come in handy later on.

II. Partial/Full CMLL and Peeking

Now that you have learned specific cases of CMLL, you are forced to pay attention to the colors of the corners. Practice recognizing these cases too. In addition, make sure you know beforehand which angles you should rotate to in order to see the corner colors. Keep in mind, you are only rotating the with U moves, not full cube rotations.

Rotations in recognition can be easily reduced by knowing how to peek to the side. Peeking is when you are 90° off from an angle where you can see all corner colors, you slightly tilt your cube to the left/right to see them. This can save at least one move.

It is perfectly fine to stick with this recognition method. It will give you good company on your way to sub-12, sub-10, or even beyond.

III. All Angles Recognition

This stage of recognition eliminates the need to peek or rotate by recognizing cases from every angle.

For "H" and "Pi" CMLL cases, this should be effortless, since all the colors of the corners is already visible on the top face.

For the other cases, a combination of the following is used:

  • Inferring the color of the stickers you cannot see
  • Knowing the color patterns of cases very well

It is important to note that on the four corner pieces, there are only a total of two stickers for each color (red, orange, blue, and green, assuming white and yellow is top and bottom). There is also a limited number of combinations of colors on each corner; that is, only adjacent colors appear on one corner piece.

Based on these properties, we may deduce the color of stickers we cannot see and take advantage of the unique color patterns that are formed.

Here is an example trying to recognize a "U" case from the given angle:

From here we can only see blue and green stickers from the top and the red and orange stickers on the side, but that is enough to tell us exactly which CMLL case this is:

  1. We notice that the two colors on the top, blue and green, are opposite to each other. Using our knowledge of unique color patterns, this brings us down to two possibilities: The "X" and "Bottom Row" U cases.
  2. We see that one corner in the back has an orange color on it, so we can infer that the other non-yellow color on that corner has to be blue or green.
  3. Out of the two possibilities, only the "X" case has colors at the back that are adjacent/opposite to the two colors on the top. Here we have blue and green on the top, and we concluded that a sticker on the back must be blue or green, therefore we must have the "X" case here.

Rotate the visual cube to check if we're right.

With enough practice, this thought process will become automatic. But where do we find all the "patterns" that we should know in order to get a case down to two possibilities? Kian's comprehensive guide on this topic will be your best friend.

Algorithms Resources - 2H

Algorithms Resources - OH


Predicting CMLL from SB last slot

Predicting EO from CMLL

Predicting LR Edges from CMLL

CMLLEO / Pinkie Pie / L10P Variants


Regular EOLRb insertions

Normal swap to bring top LR to bottom

For swapping the upper layer LR edge with the edge below it (to get both LR edges on the bottom), do an M/M' move to keep that LR edge on top, then a U2, then the opposite of the first M/M' move

Top LR with matching corners

Do an M/M' move to keep the top LR edge on top, do a U2 then repeat the first M/M' move.

Top LR with opposite corners

Do an M/M' move to move the top LR edge to the bottom, do a U2 then do the opposite of the first M/M' move.

EOLRb 4C Non-Cycle Influencing

The following solutions are for specific cases where 4C will be a non-cycle case. Influencing begins when EO is solved, and involves inserting LR edges in a non-typical fashion.

For cases not shown, one should insert LR edges normally as shown above. The solutions below typically insert them the opposite way.

The solutions either:

  • skip the dots case, resulting in a 4C skip,
  • or end up with a 2 or 3 mover 4C (U2 M2 or M2 U2 M2) instead of some 5 mover (e.g. M' U2 M2 U2 M), after AUF'ing when EOLRb is solved.

Either way, they're more efficient.

The cases are quite easy to recognise by only looking at the U-face colors and pattern; form your own recognition methods.


2 D-stickers on M slice and U layer

Dots Skip 1

Dots Skip 2

Note: Less efficient than inserting normally and solving dots with E2, but you may prefer this.

Dots Skip 3

The DFDB edge on top (blue white edge in this case) is adjacent to the center of the opposite color (green).

Dots Skip 4

The DFDB edge on top (green white edge in this case) is adjacent to the center of the opposite color (blue).

Dots Skip 5

Dots Skip 6

3-look vs. EOLR: better switch late than early

3-look Progression

EOLR Progression

It is not recommended to memorize algsheets for EOLR. Learn from text descriptions below:

MC vs. Non-MC: better early than late

Should I add EOBF on top of EOLR?

4c Prediction: battle of recog methods

Turning and Fingertrick - 2H

Turning and Fingertrick - OH


Beginner Resources

Tips on Switching from CFOP

Frequently Asked Questions

When should I start learning full CMLL/EOLR?

It is recommended to learn full CMLL as soon as possible, as you are going to have to learn it eventually. However, as a general guideline, the latest you should begin to learn is when averaging around 20-30 seconds.

EOLR should only be learnt after learning full CMLL. Similarly to CMLL, learning it as soon as possible is recommended, but a recommended time to start learning is when averaging around 15 seconds.

How do I improve my blocks?

We recommend looking at solves from fast roux solvers (Kian, Sean, etc.), and understanding how they plan and optimise their blocks (movecounts, ergonomics). Example solves are preferred as the thought process is explained, but reconstructions are sufficient as well.

We also recommend using the following block trainers to practice identifying optimal solutions for blocks


Here is a list of the contributors who have helped write this reader.

If you wish to help, feel free to reach out to any one of us! Alternatively, visit our GitHub Page and open up a pull request there.