ACMLL

ACMLL is a collection of algorithms for solving CMLL when the first two blocks aren't perfectly formed. Typically in the Roux method the user builds a 1x2x3 on the left and right side with all of the pieces in their correct positions and orientations. However, the blocks don't have to be perfect. There can be misoriented or unpermuted pairs or pieces within the blocks. Then the CMLL algorithm will simultaneously solve the last four corners and correct the blocks.

The goal should be to learn the block types that have quality algorithms. Not to learn every possible block type and its algorithm set. It is more about inserting pairs in the best way for the situation. This page will highlight a few of the sets that have been determined to be beneficial. It is important to learn only the ones that provide an advantage in solves. Algorithm quality is the highest priority and the movecount savings and other benefits are secondary.

If you want to generate an ACMLL set, that would be very helpful. If possible, copy one of the template sheets from an already existing ACMLL document from this page. That will make it easy for me to add your sheet to the rest of the ACMLL sets.

• More efficient: 2-6 moves can be saved depending on the block type. That's for just the second block. If incorporated into first block as well, the number of moves drops even further.

• Deeper inspection: The use of the technique can allow for planning more during inspection. Whether that means seeing a second block square that can be easily built using the technique, or incorporating the technique into the first block. The versatility of incorporating various block types is a major step toward non-linear F2B.

• Better blockbuilding ergonomics: Pairs can be inserting using just U and R moves. This is versus normal Roux where pairs often have to be flipped into place using something like r U r'. Slice turns can also be eliminated by just using U and R turns to make the pair and insert using the technique.

At first it may seem that recognition will be difficult. However, that shouldn't be the case. When using Straughan recognition, all of the patterns are the same for CMLL and ACMLL. When using U sticker recognition, there are only three possible corner orientation set types. Those are CLL, TCLL+, and TCLL-. CMLL already contains the CLL style orientation set types. So it is just a matter of learning TCLL+ and TCLL- recognition. There are no other weird corner situations in ACMLL.

It may also seem that with the various ways of building the blocks, once you get to the ACMLL step, you will have to recognize both the block types you have and the ACMLL case. This is also not true. You deliberately placed a pair misoriented, a couple of pairs swapped, or another situation while blockbuilding. You know what you did before you get to ACMLL. So there is no need to check the blocks. You also will memorize a few or several sets of ACMLL, not every possibility. So you won't be having to pull from memory thousands of cases.

This is when a pair is misoriented and positioned in a different slot. These sets are so far the ones that contain the best algorithms. In a solve when you build a pair, that pair will be either oriented or misoriented. If you don't worry about the orientation of this pair, you can just use U and R moves to align it above the edge or 1x2x2 that you have already added to complete a 1x2x2 or 1x2x3. It saves two moves on average to insert the pair this way versus aligning it above its correct slot and using moves like r U r' to flip it into place.

Movecount and MCC statistics:

Flipped pair FR = 9.81 moves / 11.81 MCC

Flipped pair BR = 9.67 moves / 12.35 MCC

Normal Roux CMLL = 10.02 moves / 12.31 MCC (algs from the 2H sheet in the Roux Discord server)

View Algorithms (Straughan Recognition) 

View Algorithms (U Sticker Recognition)

A two look U sticker recognition document has been created by junnoske:

View algorithms

When a pair is placed on the D layer outside of the second block. The advantage here is that in some solves when a pair is built it will be above the already built 1x2x2 but needs to be inserted with a long sequence such as U R' U' r U r'.

Movecount and MCC statistics:

Outside DFr = 9.63 moves / 13.11 MCC

Outside DBr = 10.05 moves // 12.47 MCC

View Algorithms (Straughan Recognition)

View Algorithms (U Sticker Recognition)

Pairs can also be swapped within the same block or between the two blocks.

View Algorithms (Straughan Recognition)

View Algorithms (U Sticker Recognition)

A good option that pairs well with the dFL and dFR pair swap is the outside pair version.

View Algorithms (Straughan Recognition)

View Algorithms (U Sticker Recognition)

These sets are when two adjacent pairs within a single block are swapped.

View Algorithms (Straughan Recognition)

View Algorithms (U Sticker Recognition)

A two look document has been created by junnoske:

View algorithms

There are likely other sets which would be speedsolving worthy. Having a solveing program automatically run every possibility and generate the average movecount for each ACMLL set would be a good start to finding the sets that should be learned.

One benefit of ACMLL is that it shortens and improves the final second block pair insert. After the final pair is built, it is always advantageous to insert the pair keeping its current orientation and not moving the already built 1x2x2 until the final move of the pair insert. The image below shows how this works. Based on this property and the quality of algorithms of each ACMLL set, below is a recommended learning order for ACMLL sets. 

1. Flipped pair FR and BR

2. Outside

3. Swap FL/FR

4. Front Cycle

Numbers 1 and 2 can be seen as a single group that works with a final pair. Numbers 3 and 4 are about simple pair swaps on the front layer. This also isn't limited to the final pair. The 1x2x2 of the right side 1x2x3 can have a pair such as these built in or any other possibilities. However, applying this kind of blockbuilding to the last pair at first is an easy way to get started.