All of that last part sounds good Cantus; I must have misunderstood you and thought you said you didn't do anything like that.
It's a shame about netplay (I have the same problem); it's not really something that can be solved without spending large amounts of money or getting lucky (by getting given a friend's old computer or something).
Theorycrafting is things like thinking outside of playing the game about how a certain things would work (like what strategies should be applied to certain matchups). Some really bad/basic examples for Melee Luigi are: "maybe I could go for more cross-up jabs to grabs on shield against Marth since his out of shield options are slow", or "wavedash up smash might be a better option than down smash at low percentages against spacies since it knocks down/forces tumble earlier and is more resistant to crouch cancelling". Basically you could work these things out by playing the matches and observing the results, but this accomplishes the same thing (you test out the theories when you play, of course) without having to run into the relevant scenarios in a match first. There's more to it than that (and there's probably a thread on it somewhere), but that's my understanding and poor explanation of it.
Since your setup is deliberately in an awkward position, the obvious solution (and the only real one) is to move it. If you don't want to move it because you worry about it affecting your work, then you're going to have to overcome that or give up. No johns etc. etc. You only said that you can't practice on your own because of your setup, not that you wouldn't practice on your own; does that mean that if you made your setup usable then you'd do it? Because that's a perfectly valid way to develop a lot of areas of your game; it worked for Mew2King, right?
Before I was asking about the top-tier matchups that you've already learned; you didn't mention any trouble with those, so what's to stop you learning the PM matchups like you've already done for Melee's relevant characters? What about rare characters in Melee that you can't just out-tier (which you said is your solution to unfamiliar characters) such as Ice Climbers; you must have a solution to those, right?
About knowing that your opponents' characters aren't as good as yours (this ties into theorycrafting): you must have some idea of why that's so besides them being lower on the tier list, right? It should be obvious quite quickly that Game & Watch has an awful shield, that Zelda has poor mobility, and so on. Therefore the reason you can beat those characters is because you're exploiting those weaknesses (by pressuring Game & Watch, or by camping Zelda...), whether or not you're aware of it. You should be able to apply this to any character, since all characters have exploitable weaknesses (though admittedly not always as glaring as those of a Melee low-tier) even if they're not bad overall.
It seems that you know what to do to improve and are able to do those things, but you aren't doing them because you're prioritising other things instead; that's fine, but in that case you'll have to decide between devoting time and attention to them and practicing Smash. How good do you want to get? If you want to get better then you'll need to put in more work. It doesn't even have to be that much work if you do it efficiently; see this:
http://www.reddit.com/r/smashbros/comments/2ebkwv/getting_better_at_smash/
I really would recommend playing each member of the cast to familiarise yourself with them; literally play one game with them so you at least have a vague idea of what their moveset and physics are like. That way instead of having "absolutely no idea" of what to do going into a match, you'll have almost no idea instead; that's worth the 4 minutes that it took to play that character once.
