New approaches that might finally get the belt working

[Cassiopaia]

Hello everybody.

I have some new and interesting approaches on the old belt idea. Maybe we can get a discussion startet on how to finally make it a decent method together

Here's the method so far as i came with it:

1.: Solve the e-layer (about six moves or so on average). The belt is created. While doing this, keep track on where on the bottom side top layer corners are. That's not asking too much.

2.: Seperate the corners into the top and bottom layers by U D and R2 moves (about five moves on average). A very good cuber might be able to combine steps one and two.

3.: Do CLL flip two times. (flip algorythms, my idea, solve one side but onto the other side, effectively flipping sides. The second time flips them back. that way you can work on two sides without manually flipping the cube). About two times 10 moves or so (they're more efficient than you think). Due to parity, flip algorythms have three times the cases of normal algorythm sets.

4.: Sort the edges onto the top and bottom layers. Around five moves.

5.: Do ELL flip two times. Again about two times 10 moves or so.

That's in total, if you add some alignment moves, somewhere around 59 moves or so. and since it is extremely algorythmic, TPS are high. On the other hand, there are four recognitions and sorting the edges has poor lookahead. But this might be the first belt version that can at least be taken serious. What du you think? Any thoughts on this? Ideas to improve?

 

[OreKehStrah]

Hello everybody.

I have some new and interesting approaches on the old belt idea. Maybe we can get a discussion startet on how to finally make it a decent method together

Here's the method so far as i came with it:

1.: Solve the e-layer (about six moves or so on average). The belt is created. While doing this, keep track on where on the bottom side top layer corners are. That's not asking too much.

2.: Seperate the corners into the top and bottom layers by U D and R2 moves (about five moves on average). A very good cuber might be able to combine steps one and two.

3.: Do CLL flip two times. (flip algorythms, my idea, solve one side but onto the other side, effectively flipping sides. The second time flips them back. that way you can work on two sides without manually flipping the cube). About two times 10 moves or so (they're more efficient than you think). Due to parity, flip algorythms have three times the cases of normal algorythm sets.

4.: Sort the edges onto the top and bottom layers. Around five moves.

5.: Do ELL flip two times. Again about two times 10 moves or so.

That's in total, if you add some alignment moves, somewhere around 59 moves or so. and since it is extremely algorythmic, TPS are high. On the other hand, there are four recognitions and sorting the edges has poor lookahead. But this might be the first belt version that can at least be taken serious. What du you think? Any thoughts on this? Ideas to improve?

This is unfortunately not new. It is also not good.

It is also not going to be high TPS just because it's algorithmic. The algs needed are not very good.

 

[Cassiopaia]

If I may ask, where exactly have you heard of flip algorythms? To what I know (though i don't know everything of course) that idea hasn't been around.

 

[bcube]

solve one side but onto the other side, effectively flipping sides. The second time flips them back. that way you can work on two sides without manually flipping the cube).

As a big fan of the 3-color method, I like the core idea of your flips, but I don't like your solving order which I find too restrictive.

However, neither Feather's nor Christos' nor Heise's approach start by solving the E-layer, so that's probably not something you are interested in / looking for.