If you like tricks then be sure to keep a record of them in one place.
Yeah, I had to get them down in "off-chip memory"
. I have Google sheet for F2L with a column set aside for special cases like different destination slots and open slot shortcuts.
I will share the fact that I actually dropped a heap of tricks whilst pushing for sub-20 as simplicity was the key to my look ahead. It's only now that I'm re-introducing tricks for cases which frankly seem to take half a second longer than necessary.
I initially made good progress by simplifying my F2L so much that I only used cases where the cross color is facing up on the corners in the U face. That works the same basic way regardless of the edge orientation and boiling it down to just that case meant less thinking to set it up, even if it wasn't always very efficient. That took me from floundering around 1:20 to flirting with sub 1 min and then plateaued. From there I dug into more details but still grouped/simplified a lot of cases. Because my recog and lookahead develop so slowly I decided to learn full PLL at that time as well, the two together knocked another 10-15 seconds off the averages once absorbed.
I think that's a trick a lot of people do when they start to use sledgehammer but you should learn to do R' F R F' if you're to gain the full benefit. There was a time when I used to use F l' U' l but I stopped doing it a long while ago. Watch any fast person solving and you'll see they don't use that method of execution since it involves re-gripping with your left hand and if one hand should be fixed it's the left one. The wide l' also affects lookahead as it hides a lot of LL pieces at the back.
I do use R' F R F' for sledgehammer; F l' U' l is the inverse (hedge) vs. F R' F' R. Very ergonomic for me with a right index finger over the top for the F and the left does the rest. Same left hand grip for me as sledge. The lookahead is a downside but the contexts where i use it on 4x4/5x5 usually don't matter so much. L2e flip, for example:
Uw' (R U R') (F R' F' R) Uw, that hedge execution is awkward for me but the left grip shift to do the wide version is in a convenient place after the R U R', IMO.
I like that you're keen on tricks. Just be careful you don't over complicate your solves as your main priority should be to know everything blindfolded right now. That will lead to your biggest speedups as your look ahead will develop more quickly.
I'm trying not to get too fancy and sticking mostly with efficient/ergonomic solutions for all cases but it's admittedly easy to get distracted by shiny tricks that might be more flash than useful in a solve. The main thing for me though: recog is my bane; if I'm going to make my pairing "automatic" I want to bite the bullet and find the best solutions for me to execute now rather than turn around and break habits yet a third time at the next F2L plateau. I don't use "algs" for F2L at all... I mean, I initially learn them as a sequence of moves from a table, so they're "algs", but I fully understand the mechanics of everything I do in F2L. I'd guess about 80% are easily doable with my eyes closed because I can visualize, at worst, or muscle memory for the solidified cases.
Besides lookahead and sometimes just
finding my target edges, an emerging bottle-neck for my F2L right now is recognizing the case. Blind pair practice is going to do wonders for the lookahead side, I think thousands more solves is my only recourse for speeding up the recog. Also need better cross and lookahead into my first pair now. A few timed runs last night after warm-up and I think I'm nearly back to sitting on mid 40s for averages, but when my cross is also decent I'm seeing more 30s. Also shaved another tenth off my PB single with a 33.5 with no skips.
Glad to have you guys keeping me in line! It's easy to veer off on the focus *coughcough 5x5 cough*.
. 
