against predicting speedrunners won't do things

instead, predict they might eventually do things

Sometimes, speedrun strategies are very hard¹. It is common to describe such strategies as “TAS only”, “not humanly viable”, or perhaps “not full-game viable” when they can realistically be done in Individual Level (IL) speedruns but not runs of the full game.

And this is perfectly reasonable as far as it goes! Sometimes the trick is hard enough that it isn’t worth it to try. That's a very sensible attitude to have, a lot of the time. The issue arises when people expect that state of affairs to never change, which is a sort of thing I’ve seen proclaimed from time to time.

Sorbetti Skip

In Super Mario Galaxy 2, there’s this really obnoxious midair rock jump thingy. For the longest time, no one really had any idea how to do it without TASing the game. One of the stars it saved time on was Freezy Flake 2, Sorbetti’s Chilly Reception. You could essentially skip from the first planet to the boss fight at the end, in theory. This is known as “Sorbetti Skip” even though it does not, in fact, skip Sorbetti. It saves enough time that the star can be routed into Any%.

Eventually, people figured out that Sorbetti skip can be done by a human, and got some IL times with it. As soon as this happened, it was very obvious to me that it would someday be implemented into runs. From what I remember, the Super Mario Galaxy speedrunning Discord server was very suspicious of my stance here. But it seemed to me that, even if it was very hard, and even if it didn’t actually save that much time, and even if it wasn’t close to worth it when doing record attempts given how many other time saves were available… eventually, the record would be more optimized, and maybe the trick would be relatively easier as people fiddled with it more.

And this did indeed happen—I think maybe in 242 Star before in Any%, because it saves more time when it’s straight up cutting time off the star instead of making Sorbetti’s Chilly Reception fast enough to replace some other star that was already somewhat faster than doing Sorbetti’s Chilly Reception without Sorbetti skip in the first place. On YouTube one can find a funny clip of legendary Super Mario Galaxy speedrunner Jhay missing it:

Reflection B-Side

The first player to seriously implement 6B in Celeste Any% was Revolucoin. As I remember it, everyone at the time thought Revolucoin was simply nuts. And they undoubtedly had a point—I don’t think this actually was helpful for Revolucoin’s performance at the game. He probably lost time from it, on most runs—it takes very consistent play on a level that’s rather difficult to get through without slowing down, if you want to make 6B faster than 6A. It’s a little easier in All Red Berries, where you need four Crystal Hearts, but Revolucoin was running Any%. Even 5B was pretty sketchy at the time, I think he started going for 5B before Chillino Drop was even found.

But looking back nowadays—sure, maybe Revo jumped the gun, but there certainly was a gun coming! Even players of my skill level have often found 5B more appealing than 5A, and that’s as of several years ago when I still speedran the game regularly. And 6B is still really hard, but it’s ubiquitous at the top level. The strats only attempted by the madmen have a way of eventually making it into the meta.

Of course, that 5B or 6B would end up implemented was less of a clear call at the time than something like Sorbetti skip. It was far more up in the air, at least for 6B, whether it was even optimal in theory, compared to Sorbetti skip’s status as something very hard but obviously theoretically time-saving.

The Future

I think one ought not make a post like this without being willing to stick their neck out with some bold predictions. Of course, most of my historically correct predictions have been more of the form “X thing will happen eventually” than of the form “X thing will happen within the next decade”—the former is much easier to call. And whether we get to observe “eventually” is a little up in the air, since I don’t actually know for certain how long we have until society is fundamentally transformed in some way, whether it’s for the better or for the worse.

But here are some things I’d expect to see by the end of my natural life, if the speedrunning community got to chug along the whole time in a state not too diminished from its current one:

Super Mario Bros. will go the way of Dragster

The game Dragster is famous within speedrunning because someone was proven to have cheated at it through the very hilarious method of showing their claimed time to be too fast to be possible. It’s only a few seconds long, so it’s not fundamentally all that difficult to simply perform a perfect Dragster speedrun. This results in a very funny leaderboard—usually when people say “perfect speedrun” they don’t actually mean it, but 31 people have literally performed a perfect Dragster speedrun.

The game Super Mario Bros. is famous within speedrunning simply because it is a very popular and very optimized game. Although it’s been the work of many years, No Minus World Ending² records in the current era reach the final level, 8-4, at the same time the TAS does³. The record is currently tied between Niftski and averge11, who both lose only 11 frames to the TAS throughout 8-4. Also, I’m pretty sure 8-4 has been played perfectly as an IL, though you should maybe double check that before quoting me.

I think someday someone will do it. It’s hard to say how long until then, but I think it’ll happen.

Super Mario 64 Any% will be done with only One Key

People used to think Carpetless, a significant time save in Rainbow Ride for 120 Star, was not realistically viable. It is now done in runs, they pulled it off in the last GamesDoneQuick marathon—someone found a setup.

The same person found a setup to get into the Vanish Cap course early, the first part of the Moat Door Skip route used in the Any% TAS.

And some parallel universe strategies have been performed in real time, so doing the rest of Moat Door Skip might not be totally out of the question. So, one starts to wonder…

My understanding is that the remaining barriers are still in fact very hard. Someone would have done it by now if they weren’t. But, though I haven’t looked into it spectacularly hard, I’ve never been convinced they’re insurmountable. It’s not like you’d even need to execute the run perfectly or anything, as long as your Moat Door Skip didn’t take longer than simply playing Dark World. I think maybe this can be done someday.

1. Citation needed

2. Frequently called “Any%”, despite the Minus World Ending being faster

3. I can’t say this is actually strictly required, because there are ways to lose less than an entire frame rule, I think. But time in SMB1 is—outside 8-4—usually lost in 21 frame chunks, and the record is within 21 frames of perfect.