best algorithm generator?

[Burrito]

what is the best alg generator?

I use mac, linux, windows, and chrome so an online app would be good

 

[efattah]

Most people use Cube Explorer, but the biggest shortcoming in the entire cubing community is the lack of any algorithm generator that can evaluate algorithms based on ergonomics, speed, and also that can inject wide moves and rotations into algorithms. There is no such utility, making algorithm generation an extremely slow and inefficient process.

 

[OreKehStrah]

Most people use Cube Explorer, but the biggest shortcoming in the entire cubing community is the lack of any algorithm generator that can evaluate algorithms based on ergonomics, speed, and also that can inject wide moves and rotations into algorithms. There is no such utility, making algorithm generation an extremely slow and inefficient process.

This is incorrect lol. You can use Trangium's batch solver to use any moveset (including roations, slices, wide moves) and it can evaluate algs using adjustable parameters to predict which algs are going to be faster for humans to execute (though it is not perfect). It is also browser based.

Link:

Batch Solver [Beta β]

Here's a video tutorial I made on using it:

And written documentation from Trangium:

Batch Solver Documentation

Puzzle: This field determines which puzzle is to be solved, therefore determining what the move notation is. 3x3x3, 2x2x2, Pyraminx, Megaminx, and Skewb are natively supported puzzles: just selecting that puzzle on the dropdown is sufficient. The batch solver also supports custom puzzle definiti...

docs.google.com

 

[efattah]

I'm skeptical. I'd like to see you or anyone re-generate (by computer) all 57 OLL's and 21 PLL's and maybe 50 ZBLL algorithms and if your utility can pick the well known best of these, then I'll believe it. If it is capable of even finding the best solution for half of those cases, I will be extremely impressed and use the utility immediately. Of course one of the big problems with alg generation is the shortest algorithm is not the fastest, even if it appears to be quick & ergonomic.

 

[OreKehStrah]

I'm skeptical. I'd like to see you or anyone re-generate (by computer) all 57 OLL's and 21 PLL's and maybe 50 ZBLL algorithms and if your utility can pick the well known best of these, then I'll believe it. If it is capable of even finding the best solution for half of those cases, I will be extremely impressed and use the utility immediately. Of course one of the big problems with alg generation is the shortest algorithm is not the fastest, even if it appears to be quick & ergonomic.

How do you think the evaluator was trained? Iirc with the tuning done it puts out the top PLL algs within the top 10 rankings every time. The point isn’t that it will tell you what is the best alg. It just narrows down the results of what is and isn’t worth checking.

 

[Burrito]

Is there a short and easy tutorial for this Batch Solver? Or is ksolve+ a better alternative?

 

[Thom S.]

I'm skeptical. I'd like to see you or anyone re-generate (by computer) all 57 OLL's and 21 PLL's and maybe 50 ZBLL algorithms and if your utility can pick the well known best of these, then I'll believe it. If it is capable of even finding the best solution for half of those cases, I will be extremely impressed and use the utility immediately. Of course one of the big problems with alg generation is the shortest algorithm is not the fastest, even if it appears to be quick & ergonomic.

How to lose a lot of the respect you have from me.

This thread in a nutshell.

What is a good Cube solver?
There is no solver that does what I want it to do
There is. Here's how to use it.
I don't believe it works. I'm also not gonna try it out.

But, to answer the question asked, you'll likely won't find a web app since Html isn't meant for such powerful computation.

 

[Burrito]

How to lose a lot of the respect you have from me.

This thread in a nutshell.

What is a good Cube solver?
There is no solver that does what I want it to do
There is. Here's how to use it.
I don't believe it works. I'm also not gonna try it out.

But, to answer the question asked, you'll likely won't find a web app since Html isn't meant for such powerful computation.

Thanks for the info guys!

 

[OreKehStrah]

I'm skeptical. I'd like to see you or anyone re-generate (by computer) all 57 OLL's and 21 PLL's and maybe 50 ZBLL algorithms and if your utility can pick the well known best of these, then I'll believe it. If it is capable of even finding the best solution for half of those cases, I will be extremely impressed and use the utility immediately. Of course one of the big problems with alg generation is the shortest algorithm is not the fastest, even if it appears to be quick & ergonomic.

Here are a couple screenshots of running PLL checking RUF RUD RUL MU to length 16

As you can see, for these Adj PLLs the top algs are the standard algs (Barring F perm since it didn't get to gen the R' U' F' into T since it goes over the search length but the 14 mover at the top has seen use as people's standard alg or AUF trick alg).

Then in EPLL, the 2 algs are the commonly used MU algs.

If it didn't work, why would I suggest using it? Plus as I said, if you don't agree with how it's evaluating algs, you can adjust the parameters to suit your taste. Many people have been using the batch solver for quite a while now. It's very convenient being able to tell it to gen something like full RUF ZBLL for OH overnight and wake up to a set of nice algs.

 

[Silky]

How do you think the evaluator was trained? Iirc with the tuning done it puts out the top PLL algs within the top 10 rankings every time. The point isn’t that it will tell you what is the best alg. It just narrows down the results of what is and isn’t worth checking.

This is very well said. Being able to reduce the amount of algs you have to comb through is huge. I still cross check with Cube Explorer and test algs by hand but things like this and MCC cut out a HUGE amount of leg work. It's imperfect of course. One improvement I'd like to see for MCC is to be able to put weights on certain moves/triggers. In several of the algsets I've generated there has been an over-prioritization of S and D moves and a under-prioritization of M moves.

@efattah I'm not sure if this was make clear but Batch uses MCC to sort algs (made by the same person) so if that is something that you generally trust than batch should be too. Probably doesn't mean much coming from me but I can at least vouch for it. It a tool in the tool belt for method dev in the very least.