K4 ELL?

[fanwuq]

I'm trying to get decent at K4. I can do it in about 3min now compared to 1:30 with reduction.
I learned about the method from Fallofshadows and Pembo on Youtube.
I do Opposite centers, 3 edges, all centers, pair 4th edge, Solve 1st and 3rd layer using Fridrich, insert 2nd layer, CLL (2 look), ELL (many many looks).
Total move count is about 170.

The only thing I know is RUR'U'rURU'R'r' and its variations with l, mirrors, and inverses.
According to
http://rxdeath.com/k4/step7.html
I should try to make 2 adjacent edges than continue from there. The problem is that I can only recognize the pairing up of BUl and UFr by using the inverse of the alg above. I end up doing [r',FR'F'U'R2U] for pairing up the last 2 algs. Then I end up doing EOLL and EPLL and the regular parities.
Are there any easy ways to approach this so that I can solve the pieces rather that just pairing them up. I would like a way where you do not memerize every single alg and it requires no more than 4 looks for ELL.
1.Do you solve 1 piece at a time or do you have know exactly which stickers to cycle (like a free style 3 cycle method)?
2.Would it be easier to just “orient” the single edges and do single edge U perms to finish?
3.Maybe I should just finish like it is r2 for 4x4 BLD? (same idea, but use r rather than r2 to keep it in LL)

Thanks.

Edit:
Wow! I just used the r2 idea and it worked! I got 128 STM. I only had to do 2 3-cycle commutators and Z perm for ELL. CLL was 1 look COLL. I didn't time it, but I have a feeling that it was very fast.

Edit2:
This makes the idea of cycling edges alot easier, but there are still some special cases for finishing last 2 dedges. I guess I have to learn a few more parity algs to do "OLL parity" and swapping pieces between edges in 1 step?

Maybe these algs? http://bigcubes.com/5x5x5/lastedges.html
(assuming that they do not mess up the rest of the cube.)

 

[blade740]

Yeah, that's pretty much how I do it too. Using r2 type algorithms gives you a lot of possibilities, and every alg you learn from the page can be adapted and set up in multiple ways, so even learning a few will come in handy. Work on solving two edges with every 3cycle. Also, I'd suggest direct solving rather than pairing + 3x3 ELL. Either way, you're solving 2 pieces at a time, but one way will require you to do an extra alg at the end. I only do a 3x3 alg if it will solve at least 3 pieces, or set up for an easy commutator right afterward.

 

[rachmaninovian]

I can do k4 1:40ish..hehe..faster than my reduction times of 1:45ish to 1:50ish.

 

[Kirjava]

Word.

I posted this recently because I've developed the way I solve the LL into something more concrete

So once you know a bunch of algs from my page and that link, the ELL step should take you no more than 3. I'll have to write an update for the site sometime with my new findings. The non-4-cycles are easy enough to learn since they're all pretty much the same thing.

The hardest part is solving any two edges in two algorithms, but 90% of the time it's just a three cycle for each one anyway.

By the way, pairing is more more difficult to recognise as you have to see the 3x3x3ELL you need aswell. Solving them directly is a less abstract solution, and you want to be able to easily know what algorithm you're executing next.