Well, to be fair, since the possibilities are so numerous in just the match settings alone that it would only be fair to consider it infinite in possibilities, standards have to be set for "accepted" gravity to create a baseline. It doesn't detract from what you've made if you make a sweet replay in 0 gravity, but for the sake of establishing a framework, it's pretty convenient to have two "accepted" gravity levels. Plus, most gravity changes have such small changes in most situations as to be near unnoticeable. In addition, standards allow us to create objective measurements on certain achievements. Surviving a fall from 2.0*10^999 height means nothing if the gravity was 1.0*10^-9999999.

And progress in the multiplayer scene is different from the singleplayer scene. MP is dictated by what's effective, so people constantly shun "style" in favor of efficacy. An extremely fancy move that wins 50% of the time is terrible in comparison to an extremely simple move than wins 70% of the time. Therefore, if you have a certain style of move that's winning a lot (queue jumpkicks in wushu being countered by sweep kicks at one point (which in turn got countered by helicopter kicks)), that move will inevitably see many different variations that try to milk every extra 0.1% winrate out of it. Meanwhile, singleplayer tends to shun efficacy unless efficacy is the focus of the replay. Singleplayer is more open to various different styles because the goals of singleplayer are not so one-dimensional as multiplayer's one goal of winning. You can only win in so many different ways (the 3 types of gamers argument inserted here (power gamers, combo gamers, and one other type that i can't remember)), but you can aim for a huge number of goals in single player. The 3 types of gamers thing still applies to single player, but it's not constrained by the goal of "winning is the desired end result".



To put it to an analogy, it's like the game is a test question. The question changes though based on what you're taking. Multiplayer is the abstract math question of the travelling salesman, except it's the pursuit of the perfect move instead. It's theoretically impossible to solve because of how many possible combinations there are, and it's impossible to prove because of the same reason. That being said, you can approach as close as you can to the perfect move within a certain degree of accuracy. And thus, the test question can be "solved" in a sense. Multiplayer is the constant pursuit of getting a higher and higher degree of accuracy.

Singleplayer is an essay question for an opinion paper. You can take any goddamn opinion you want, just so long as you can back it up with a solid foundation. Likewise, you can make any damn type of replay you want in single player, and nobody will really care, so long as you make the replay following a certain standard. Minimalist movement? Sure, acceptable hurdle. Most dismembers possible? Sure, acceptable hurdle. Spinning ridiculously large amounts of times? Sure, knock yourself out. Doing absolutely nothing except spinning fakey 460 noscope backflip tricking speed runs in parkourmod#90billion? Sure, there's a market for that.

The issue is just that, like with an opinion paper, if somebody else has an opinion and backs it up with a really good argument, you can be sure other people are going to adopt that opinion. So if somebody makes a really good replay, and backs it up with aesthetically pleasing movement, or extreme realism, or what have you, there will inevitably be imitators. And it makes sense, if it's good, you want to prove to yourself that you can be just as good, but in your own way. So you adopt the idea that was established as good, but then put your own take on it, through trial and error or flair or whatnot, and then show it off to the world. It's probably similar in terms of simplistic descriptions, but the little differences that you struggled through make it your own. And that feels good. So people do it, and they get rewarded for it because they see something that they admired become created by their own hands, and they think "damn, I'm pretty good at this", and they see other people who see it and think "damn, that's pretty good, I want to do it" and that further reinforces it, while continuing the cycle. Until somebody else comes up with something else that's impressive.

Historically, you can see that effect on a lot of styles. Realism had only a simple interest when the first "walking" replay was made. It was clunky, it was stiff, but it was a walk. People were interested, but nothing really came of it. It took somebody, in this example Nuthug, refining the style to the point that it didn't look pedestrian or mediocre and showing that, yes, it could indeed be done and it could look damn good too. Similarly, the use of smaller turnframes didn't exactly catch on until ChezDa, noweb, and some other oldies started showing the complexity that could be achieved from using turnframes <10. Now the only place you really find replays using only >= 10 turnframes are multiplayer replays.