Struggling to improve columns first

[Sion]

Well, here we are.

I'm trying to pursue columns first as my main method. I just love how flexible it is, and it goes a bit well with my turning style.

That said, I'm really struggling to get better.

I did try switching to cfop or roux, cfop was too restrictive with a cross bandaging many move options and just didn't feel correct to me. Similar thing for roux, but it's the fact that I need to solve my pairs in specific orders and locations. I just feel I grew so accustomed to the freedom alotted to me in PCMS I just really can't switch.

I also tried working around the method to things I just Like and dislike, such as I absolutely hate F rotations. I also like pll much more than cmll, so I decided I'd lower the algs halfway and increase the algs with what I think is a more feasible and recognisable set.

My goal is to be sub 17 by next year, and I don't know what to do. As of now I'm solving all over the 20 second range, with a pb of 17.57. I need help soon.

 

[DGCubes]

Hmmm, I personally don't have any experience with columns first, so feel free to ignore this post...

But! Have you ever tried ZZ? Absolutely no F moves (except for 1 or 2 at the beginning), you can use PLL, and you're not forced to solve F2L in any particular order.

 

[Solvador Cubi]

You could also take a look at LMCF for a corners first approach.
https://www.speedsolving.com/forum/threads/full-lmcf-3x3-method-now-available.63998/

It might have some similarities to a Columns First method and has many options for finishing.

 

[Sion]

How about the MZRG PCMS Variant? I'm currently learning PLL, so why not?

 

[Hazel]

If you like columns first and PLL but if your PB is still 17 seconds then maybe try this method I just thought of (I'm getting sub-20 times with it:
1) Solve 4 F2L pairs (with the centers solved too)
2)orientt LL corners
3) insert 3 first-layer edges with M, U, and y moves
4) insert the final first-layer edge while orienting all the LL edges with EODF
5) PLL
EDIT: Hey, just got a 16.47 full-step single with it

Does anyone know if a method identical to this already has a name?