Are Computer Generated Roux First Block Solutions Learnable?

[aquila_zyy]

Hello. CFOP solver trying to get into APB here. In case you're not familiar with APB, the first step of it is exactly the same as in Roux where you build a 1x2x3 on Ld.

This is the first time I've tried a method based on block-building, and it felt very different from the cross in CFOP. I have learned the various ways of building the FB (2F2L, 1x3 lines, half lines+F2L, etc), and can spot some common patterns like solved edge-center or corner-edge pairs.

Trying to improve my FB efficiency, I tried out some tools that generates optimal FB solutions to help me spot patterns of corner-edge relations. However, unlike training with a cross solver in CFOP, where I can almost always find a solution with the fewest theoretical moves, the solutions computer generates for FB feel very unintuitive, and I don't think I will ever come up with those 6 or 7-move solutions on my own. So my question is, is it possible for human to come up with FB solutions as efficient as the computer generated ones consistently at a high level, and is that something I should aim for? Any tips or general suggestions for improving FB is also appreciated. Thanks!

 

[abunickabhi]

Yes it is learnable.

This reminds me of 2016 when I used the Roux FB tool hosted on the cubegrass website.

 

[AJT17]

Hello. CFOP solver trying to get into APB here. In case you're not familiar with APB, the first step of it is exactly the same as in Roux where you build a 1x2x3 on Ld.

This is the first time I've tried a method based on block-building, and it felt very different from the cross in CFOP. I have learned the various ways of building the FB (2F2L, 1x3 lines, half lines+F2L, etc), and can spot some common patterns like solved edge-center or corner-edge pairs.

Trying to improve my FB efficiency, I tried out some tools that generates optimal FB solutions to help me spot patterns of corner-edge relations. However, unlike training with a cross solver in CFOP, where I can almost always find a solution with the fewest theoretical moves, the solutions computer generates for FB feel very unintuitive, and I don't think I will ever come up with those 6 or 7-move solutions on my own. So my question is, is it possible for human to come up with FB solutions as efficient as the computer generated ones consistently at a high level, and is that something I should aim for? Any tips or general suggestions for improving FB is also appreciated. Thanks!

Yes, many top Roux solvers can find the best FB on any given scramble.

The best way to train optimal FB is by using this website. It gives you the optimal solution for any colour, depending on how colour-neutral you are, and it remains very helpful as you use it and become more experienced with solutions.

Onionhoney's Roux Trainers

Roux trainer collection for all your cubing needs

onionhoney.github.io

Another approach you could use is by starting with a solved FB, then seeing how many 1-move solutions there are, then once you're comfortable with those, see if you can see 2-move solutions. Then keep doing that until you're comfortable seeing into longer FBs, and make sure to target the more unintuitive moves you're uncomfortable with. If you get those FBs you're uncomfortable with, you'll struggle, but if you continually practice the FB solutions/movesets that are your worst, you'll slowly get better and better at the difficult ones until none of your FB solutions is bad.

Another thing that will be immensely helpful in improving your FB solutions is looking at example solves and reconstructions from top solvers. You can look on YouTube for any example solves, and you can use reco.nz to look at any Roux solutions that have been reconstructed, so you can get an idea of what the best Roux solvers are doing.

 

[Renslay]

Bear in mind that the shortest solution is not necessary the fastest or most fingertrick-friendly.