If you need to ask this question then you should probably scale down everything until you know what parts you will need to make "what you want".

I know this sounds very confusing but I would recommend that you start with something like pacman, and when its a playable game try to see what parts the project can be divided into. Then see what different games could possibly be made with those parts.

Now you could try to make a new game (a bit more advanced maybe) and reuse the old structures but also make improvements and adjustments where you earlier code was performing badly.

Now you should be starting to see more and more common elements and can be creating structures that will work good for both kind of games, well now you are a good way to making your own library/engine/framework/collection of structures but most importantly you can vision broadly what elements a certain game could need,

To apply this to a fighter game make a really simple 2d platform game instead.

To apply this to a fighter game make a really simple 2d platform game instead.

I'm gonna agree somewhat with Kristoffer. Rather than making the actually game you should make a much simpler game. This is where I think our concepts diverge.

Say you want to write a game like the latest Mortal Kombat. Rather than creating the actual game now, maybe you should start with something more along the lines of the original Mortal Kombat and work your way up. You could start with something with only 4 player states (idle, punching, kicking, and dead). Then eventually you can add more moves, like some type of spinning kick and maybe a uppercut punch. Then maybe some sound effects, a main menu, a pause menu, different arenas/backgrounds, more characters, etc. Eventually you'll get to where your original concept was.

NOTE: I have never played any of the Mortal Kombat games but have seen videos and screenshots of it. This was just as an example.