Depends on the category.
Glitchless is by far the most popular category. Fundamentally it's a simple run, but being quick at it takes a LOT of honing. Getting used to movement and the rolling HP mechanic in battle is really important and takes a lot of practice as well. It's also long, so learning the whole route takes some serious dedication. If you are considering this category at all, you should check out the boogey% board and try learning that first. It's roughly 1/4 of the glitchless run and is a much easier run to digest.
Any% is not that popular, but that's mostly because it's pretty difficult. Most of the runs on the board use a fair deal of manipulation and require strong movement and input accuracy. Some of the glitches are pretty tough too. That said, there are times and routes on the board that are easier to do if you are not too worried about going as fast as humanly possible.
Those are the main categories. Any% Debug Menu would be a good one to look into if you want a shorter run.
I'll say forever that Glitchless is probably the way to go if you want to get comfortable with the way the game feels. Unfortunately it's a longer category but you get to experience the full game and learn all the movement tech, and there's no need to perfect any fancy-schmancy tricks, glitches, or anything like that. (Boogey% is just the first quarter of the full glitchless run, it covers the basics as well like despawning and stutter-stepping)
I've always compared EarthBounds barrier of entry to Link to the Pasts in that its a relatively easy game to pick up in terms of routing and tech but very difficult to master. EarthBound has been many peoples first speedrun, myself included, so I think its a good starting point.
I recently started speedrunning as well, I picked this game because it was one of my favorite as a kid. I have only run the boogey% as it only takes around 2 hours the first couple times you run it, I actually found that speed running this game can be easier then playing it normally once you learn the route.
