Logiqx
Veteran
Ok. I don't think I have anything revolutionary to say about practice but I can tell you how my approach has changed and the effects I'm noticing. The past few days, I've been averaging low 17's which is 2-2.5 seconds faster than I was in January! My practice used to consist of an hour in my car during my lunchtime, 4-5 times a week. I used to do some causal solves at other times of day but I'd say 75-80% of my solves were timed and perhaps 20-25% we untimed.
Nowadays, I get to cube about an hour in the morning and an hour in the evening on the train but at least 75-80% of these solves are now untimed and at most 20-25% are timed. The untimed solves are all about "smooth turning" where my goal is to eliminate pauses and create the nicest possible flow and smoothest finger tricks for cross + F2L. I execute my LL algs at 60-70% normal speed to keep the noise to a minimum but I still aim to make them as smooth as possible and minimise pauses!
In a nutshell it's just what people call "slow turning" but I'm forcing myself to track the pieces moving around rather than relaxing and thinking "no need to concentrate at this speed". The net result is that my look ahead and tracking skills are rapidly improving and the speed that I can turn whilst still looking ahead and avoiding pauses is also increasing. I'm also developing some "know ahead" skills whereby I know where the pieces of the next F2L pair are going to land before I've finished the current pair. At the very least, I seem better at knowing where a corner will appear after solving the current F2L pair.
Another advantage of doing untimed solves is the ability to actually stop and consider alternative solutions. I'll sometimes unwind an F2L pair and try something different to see if it makes the next pair any better. If I see a missed opportunity for edge control during the final pair, I'll also go back and use a sledgehammer / whatever to get a more favourable OLL case. When I do something silly or miss opportunities, I make a mental note and try not to repeat it!
I'm putting more effort into my cross planning, eliminating rotations where possible and making them more finger-tricky. I used to incorporate u / u' into my cross a lot but I rarely use them now and use D / D' instead as they are more finger-tricky and I have a better idea about where everything is when I'm starting F2L. I still don't attempt to predict the first F2L pair during most solves but my improved cross planning / execution and general look-ahead skills mean that I often see the first pair during cross execution and transition to F2L without pausing.
I'm also changing a few of my OLLs and untimed solves are the ideal time to practice. There is no temptation to revert to my original alg when they crop up as the time does not matter!
TL-DR - Slow solving, deliberate practice.
I feel there is still a lot more improvement to be made using this approach and I also have a long list of other skills that I'm lacking; x-cross, advanced F2L, two-sided PLL recognition, AUF for all PLL cases, etc.
