Personally if I don't have save states available I would refuse practicing seriously for any long period of time due to the large amount of time wasted but I'm probably a bit ruthless :P
but occasionally at events, I would repeatedly play just one game on PCB to try and get better.
I approach it then in a simple manner. The most important thing is to prevent waste of time, so you will want to reach for consistency to avoid repetition of easier parts. [it's not just to waste less time / learn faster, but also to make it more fun, because losing repeatedly for putting yourself in impossible situations repeatedly is not all that fun I think. These games are designed to be hard when you know where to be and what to do, so if you don't they will often put you in impossible or very volatile situations which don't involve your skills quite as much!]. At first, I just fly around trying to survive and at the same time be somewhere that doesn't seem disadvantageous to the scoring system. Then, anywhere I die I try to remember and either approach it differently, be ready to bomb, and I do bomb a lot then though not pre-emptively, trying to find a better way to approach or dodge, and try not to die with bombs in stock. The difficult thing to do here is to actually remember where you died and what happened shortly before that death so that you don't repeat a faulty path. Watching other players helps better than actually playing because you can observe a lot better. I think, taking time to study outside of play is even more important when you don't have save states. Basically experimenting is cumbersome, so outside study gains in relative importance. I've heard that players in Japan, when a new game comes out, they watch each other play a lot and even make some simple drawings of levels sometimes to figure out how/what to do. Whereas with savestates + video, you don't have to spend a lot of time watching the video at all because it is so easy to repeat experimenting with the section in the video. So you watch it once and try it 20 times, maybe watch it 3 or 4 more times and then you are able to do it yourself through experience. Whereas without save states you may need to see it done more times and map it in your head more precisely before you get your chance at playing it.
That's what I would say on this topic. If you follow the game's proposed approach, to reach for score extends as first goal, then the bonus extend trick, then get to the end boss (no need to all), then get more points, all the while doing outside study focusing on mistakes you make often and always try to avoid memorizing routes that are merely pure survival and completely detrimental to scoring, then you are on the right path I guess! Avoiding pure survival routes also ensures you are not playing a too-easy game that also don't improve your skill! That is also why you should really avoid any form of pre-emptive bombing! (but instead definitely practice your bombing skills)
Tbh I think when you own the game, probably the best would be to first perfect stage 1 by just always restarting at the end of stage 1 and doing outside study until you get a really good score and no bomb/death regularly enough. Then go to learn stage 2 by playing stage 1 and 2 and restarting at the end of stage 2, etc until you know the whole game. Not wasting time trying to fly around in further stages when you can't yet reach them consistently in strong conditions. If there is something to focus your efforts on at first, it would be those stages that you simply must play when you start a run, so the first stage, then the second, then the third... That would be the most efficient I think if you want to get a solid score over the whole game. To push this logic further, you would start learning stage 2 once you are at like 50% consistency in a known "perfect" or "wanted" stage 1 scoring route, so that every time you play stage 1 before stage 2 you are practicing consistency on stage 1 as well... as long as failure in stage 1 doesn't impact being able to practice the scoring route of stage 2 which depends on the game. Maybe just knowing the scoring path and having no consistency is enough to start learning the next stage while just practicing consistency............. like, the first time you get a perfect stage 1 with the high score you aimed for, then you stop restarting @end of stage 1 and start learning stage 2 and @same time just prax stage 1 consistency, that would be most efficient and fun, probably :D
For me efficient and fun go hand in hand really, just need to set your goals right, not too high or you may be stuck in failure forever, not too low or you will learn nothing and just take small pleasure in easy wins.
If you want something more simple, not optimized scoring, then I guess it would be fastest to try and learn everything at the same time by completing your runs and just looking over a bit of everything like I described before. That is likely more fun too if you have no intention of breaking very high scores. That's how you would get a ALL with a decent score fastest, but not how you would get a very good score fastest. It is also probably better if you have no video or other player to watch, because then you will have to make it all up. So yeah in that case, credit feeding repeatedly, sure, just catch any chance you get to learn something about later stages, makes sense.
Maybe you'd want to do the credit feeding thing until you get ALL and general idea of the whole game, then switch to the suggested sequential practice method. Doing all this credit feeding play beforehand would then help a lot not setting your goal too high or too low, and it would ease the process of learning any sections for precise high scoring afterwards because you have a general idea of everything already to use as a knowledge base. In the case of having no save states or level select, yes I would imagine that's great.
Scores, replays, videos || Currently playing Nothing
I have written a guide about getting good at shmups. Check it out !