Immortalchaos - you mention getting it right half the time at your peak. Remember that we are aiming to be able to look ahead with perfect accuracy in any situation. Does it not make more sense therefore to work on trying to achieve 100% accuracy rather than working at a 50% level? If you can solve f2l without pause at 160 BPM, surely trying to get 165 absolutely right, rather then 200 BPM half right would be a better use of your time? Once you have 165 sorted, aim for 170 and so on. Practising at a level that's way beyond your ability is just teaching your brain to turn beyond your ability, missing the point of slow solving entirely. The whole point of metronome solving is to force yourself to do slow solves, and the benefit of the BPM figure is to measure your at slow solving, that is solving at the rate at which you do not need to pause. Practising at the level where you need to pause half the time, that is every other f2l pair, is by definition not true slow solving, so circumvents the entire benefit that metronome solving offers.
I definitely recommend ramping up to a goal maximum bpm speed. Consider them warm up sets. Get your brain in the right mindset for metronome training, locking into that click click click, its definitely the way to go as it allows your brain to adapt to the different turn style that metronome training requires and the necessity of keeping that formal, regimented turning. Going straight in there at max speed will probably involve a bit of fudging, as your normal solves require a bit of lookahead fudging (else why do you need to work on it?) and you apply that way of working to the metronome, and you dont grow in ability.
If you are getting it right 50% of the time in your slow solves (which is the point of the metronome training, forcing the formally accurate flowing slow solve) then when you return to regular solving your lookahead AT ITS BEST will be 50% of perfect. Usually less than this. This undermines the point of the practice method utterly. Go for 100%. If you get 10 out of 12 solves, try and figure out why the other two didnt work, and then adjust in some way to fix them. It may be learning a couple of new f2l algs, it may be slowing down a bit more, it may be trying to raise your awareness of what is in the rear slots, whatever. Just use the practice technique to become mindful of whatever the deficiencies are, and then you can fix them. Thats the whole point. If you can do 100 bpm perfectly, but 160 with 50% efficiency (or 70% or 80%) then you have identified that you CANT do this for a bunch of cases. You need to fix those by training your brain to maintain lookahead even if these awkward cases come up, and that usually means going slower, or figuring out a way to avoid a rotation. Its always hands or eyes. Eyes means go slower, hands means learn a better way for that case (that 10 move case with a y then a y', that sucks. Identify it. Fix it.)
So you want a higher hit rate than that if you can. The problem you get is that in an average of (say) 12, if you only get 6 right, thats a 50% failure to lookahead. So half the time an f2l case is coming up that you cant look ahead to, and instead of learning to do so (which can only happen at a slower speed), you are training your brain to ignore it because its one of the 50% of cases that is too hard to worry about. It is better to go slower and aim for 100%. I do two types of training: cross + f2l (specifically trying to train that cross to first pair transition, which for me is a weakness) and f2l "flow" training, where I begin with a completed cross, simply because the rest of my f2l is so much faster than the first pair. That way I practice looking ahead from one f2l case to another, partially predicting where the next incomplete pair will end up etc. These are two different skills, and completing f2l at my cross + 1 speed is pointless (because the rest of my f2l lookahead is much faster so it doesnt help), but trying f2l + cross at the speed of the rest of my f2l is currently impossible. Better inspection/tracking during cross execution would deal with this, so that is what I practice. Once BLD cross is in place, theoretically at least, your first pair should be no problem because you can look ahead in the same way that you can between pair 2 and 3. Actually theoretically its easier, because the cross take more time to build than placing a pair, so your eyes have more time to find the first pair, IF and ONLY IF you dont need to look for the cross. Crosses that end in a D/D2/D2'/D' move are always better because you can anticipate the D move and concretely lookahead whilst you execute it, you absolutely dont need to look at the cross pieces (assuming you know your colour scheme). You should strive to have the whole of your cross to the same level, so all your eyes are doing is anticipating the position of one previously identified pair (I say previously because the best find it in inspection, striving for merely during the cross is slightly less superhuman). Drawing an analogy with the gym, if you are trying to bench press a weight 12 times, but can only manage 6, then you have to reduce the weight.
So, try aim for training at a speed with 100% consistency. The best place to work is where you are slightly off that, maybe 95%. Let me draw the gym analogy again. If you have decided to bench press a weight 6 times, but can only do it 3, its too much. Pick the weight where you can do it 5 times and the last one is wobbly. This is a good weight for your training, with repeated attempts over several sessions you will find soon enough that the last isnt wobbly, so you should add more weight. By analogy, work at your 95% correct metronome level, until its 100%, then raise it. Thats where your practice should take place.
In terms of using the metronome effectively to get that extra 5%, there are a couple of things you can do to squeeze out the improvement. Having used metronomes extensively for musical purposes, I have adapted some of the standard techniques. Find your almost there level, lets say its 160. Start your session at 100, then work up in regular increments to your chosen challenge level of 160. You should hit your wobble at this point.
1)The first idea is to just solve at this level until its no longer wobbly. Its dull, but its effective. Its dull because you improve when you hit the case you cant do, but 95% of the time its not a problem, so you feel like its dull. You are however consolidating your skill (which is necessary and useful), but should also be identifying specific deficiencies - tricky cases, excess rotations, poor move counts on certain pair configurations, ignorance of rear slots, whatever the problem may be. This allows you to isolate and fix an issue.
2) Try going up by 4 and then down by 2, working up to and hopefully through your barrier. Up a bit, down a bit helps because your brain adapts to the manageably more difficult pace, (manageably because it is only slightly quicker, like adding a teeny bit more weight to your lift)/ It wont be 100%, but if you can maintain 95% for a small increment, you will often find that when you go back down you can manage the previous speed at 100 or 98% efficiency. This is forcing your mind to adapt and push through the barrier by repetitively slightly overloading its current capacity, then trying again at a slightly lower level. By comparison your reduced level feels easier, despite being faster than what you started with. Its a good technique.
3) Spend some time at your 160 level, locking into the regimented turning, and getting your 95%. Then just go 20 bpm higher, say 180. You will find that your hit rate becomes lower (obviously) maybe 70-75%. Keep solving at this level for a bit. Your solving brain-eyes-fingers system becomes overloaded, and desperately tries to cope. Fudging occurs. Your system desperately tries to overcome the new stress placed upon it any way it can. Your 70-75% creeps up to 75-80% as your solving mind begins to cope. Then drop back to your 160. If you have spent a meaningful period of time at 180, 160 should now feel incredible easy. You have forced a skill increase through forced overloading of the solving system. This doesnt always work and can lead to "panic solving". If so, the jump has been too large. Hang out at 160 again, then try 175 or 170 for a it. 20 bom jumps work for me, maybe 10s or 30s work for you. Use this techniaue sparingly, you dont want to encourage fudging too much, you want to have the accuracy you can achieve at 100 bpm (which should be absolutely perfect if you are practicing at 160) at your target bpm.
Contrast the use of these techniques if your working level only has a 50% hit rate. Cranking up by 20 bpm will be a case of practicing at a 30-40% hit rate for a while, (which lets face it, is basically messing it up unless you get an easy case) , and best case scenario you reduce back to your original level and find yourslef hitting 55%. Pointless, you have too much weight on the bar. You need to find where you can do it 100% perfectly, and then practice at or slightly above that level, so your 100% becomes no just doable, but triviallly easy. This is achieved by intelligent incremental advances, where perfection is always very close.
In conclusion, spend most of your time at or just beyond the point where you can do it perfectly, with good form, and your eyes arent behind your fingers. This is the ideal, and so should be what you practice. Practicing at a point where you just cant do it (50% say) is just getting better at failing. Incremental advances are systematic and effective. The increment should be one level above what you can already do perfectly.
The inductive solving idea is useful, and tends to be a skill that develops through practice at the appropriate level. The mental calculations involved in deducing the colours and orientation of a piece you cannot see will be a contributing factor to your 5% failure rate (when working at 95%). If you identify this as a specific weakness, then the 3 sided logic based solving exercise is most useful. A similar exercise is to attempt to solve the cube with only one rotation. That is do the cross and whichever side is facing you, solve as many f2l pairs as possible using 2 gen moves (which you can always do for pairs where the edge is oriented after the cross is comlpeted. You will solve 1,2,3 or 4 pairs. Then rotate y/y' and solve the rest. If you find you have missed one and need to rotate back, then you have failed the exercise, if you solve the entire f2l with no more than 1 rotation, then you have developed your ability to identify and work with edge orientation. By extension, correctly selecting sledgehammer insertions over regular insertions can further reduce the need to rotate, another skill that when developed to the level of unconscious competence will improve your f2l massively. As an associated side benefit, reducing rotations will further develop your ability to deduce colour and orientation of pieces that are not directly visible, so the exercise assists and ties in nicely with the 3 face exercise you mention Hypocrism. All of these practice techniques should be applied as a result of concretely identifying specific weaknesses and then sorting them out by working in a focussed manner on the identified weaknesses until they are no longer a problem. Then identify the next issue and work on that. This is the very essence of systematic training, and is the logically best way to continually improve.
It can be a bit tedious at times, and I am as guilty as the next guy at sacking off proper specific practice in favour of trying break my avg 100 pb, but I can assure you that 1) identifying your most problematic issue as specifically as possible, 2) fix it so its no longer a problem, 3) repeat from step 1, is the most effective way to get better at any skill.
Hope this helps.
