Aaron, you are a worthy competitor! Please continue working on a <21min run! I'll be working to improve mine, but whatever happens SDA is eventually going to get a kickass run to post.
Thanks, and absolutely, competition just means an even better published run. Did you say you submitted another one after the 23:37?
I'm slowly getting back into running this game again, after a long break. I'm making a lot of those dumb nervous mistakes like exiting out of the wrong level or starting to play x-18. Gotta build back up.
I think I'm starting to see evidence of repeatable Larries behavior, though I've thought that before and it could always be psychological. When the first worm starts exploding, I start to jump up and down following the sixth explosion sound effect, before the seventh. The second worm follows immediately after about 75% of the time. It's better than my previous strategy, which was "jump a lot and pray the speedrunning gods are in a good mood".
I've been thinking that Larry has been repeatable too. The movements I use in my runs, when done correctly, seem to offer a higher chance of that perfect pattern than what other people seem to anecdotally see.
Random number generators aren't random yo. The seed is based on SOMETHING. Whether that be your character's position, the number of miliseconds elapsed, or what have you. Haven't you ever seen a TAS run, or manipulated a boss into doing what you wanted when he can pick from many moves?
Sure they are based on SOMETHING but the seed itself is random. The character position for example only takes a small part in the final number (if at all. in most cases it doesn't). If you dont trust me use a memory searcher and follow the stack operations back to the initial rng adress.
P.S.: Most rngs update every frame so what you just mentioned would mean you'd be using a frame-exact copy of your original input at exactly the same starting rng-value.
P.P.S.:
Quote from CMiller:
Haven't you ever seen a TAS run, or manipulated a boss into doing what you wanted when he can pick from many moves?
It works for TAS' because they are able to replicate frame-by-frame inputs every time. The example you mentioned would be an artificial intelligence thats entirely based on (for example) your own position compared to the enemies position without making use of the rng. So in essence they are set patterns which occur based on where you stand. That has nothing to do with random numbers whatsoever and thus can be manipulated in a real-time run.
That states they are random, but it's a wiki site so someone probably just entered that in because it looked random to them. I doubt they looked at the game code.
It's very obvious that the boss is based on something that is not repeatable by human input (most likely purely by a random integer value from 1-3 or 0-2).
All I know is I seem to get better results when I jump around than when I just stand there after many hundreds of attempts, so I'm gonna keep jumping around. Don't scoff, everyone who does long runs with luck-based sections has their rituals, it keeps you from getting discouraged if nothing else.
On another note, I've developed a habit of suddenly not being able to do certain levels I've done 1,000 times flawlessly. I had to change my strategy to 3-14 because for some reason, I started falling off the third moving platform more than half the time. A lot of my route is strongly tied to muscle memory, so I keep getting in these situations where I start to think about how to do something, and suddenly I can't do it anymore. Eventually I relearn it, and then some other level starts giving me trouble, like the speedrun equivalent of whack-a-mole. wheeeeeeee
I'm compiling many of the deaths and close calls from my recorded runs for an eventual bloopers video, I kind of hope I get a 2-minute one at some point so I can drop the entire thing in there and gradually speed up the video as it just keeps on going.
I've gotten close to that bad on Larry, but not quite that bad.
I just ran a real end run for the first time today, and got a 58'58" with close to 100 deaths. I think a real end run could easily get sub 50 with a good run. I'm gonna run some more this week to see if I can get a decent time.
I don't usually get that bad, but it's happened a few times lately and it's kind of discouraging. I tend to finish a run either way, but if Larries doesn't go down in five turns or less, I'm not gonna make my time.
A Real End run would be fun as hell, I've actually never tried it. What's your route (if you don't mind sharing)?
Right now, I'm doing two extra light world stages to save time in dark world (1-15 & 1-19). Once I get better at some rapture stages, I'll probably drop those.
Other than that, I do 2-19 instead of 2-14, because 2-14X takes forever, and I do 3-19 instead of 3-13 for the same reason.
You're only supposed to beat 85 dark world stages to unlock the fetus stage, instead of 90 light world stages, so you skip 5 stages for free. I'm skipping 6-2X - 6-5X, 5-3X, 5-12X, and 5-14X, then doing the extra world 1 stages. I think I'm just gonna do 5-12X and 5-14X once I get better at them instead of 1-15X and 1-19X.
I forgot how much harder the dark world stages are, I've been playing SS any% almost exclusively for months now. Did a dry Real End run and stopped after an hour with 227 deaths, 65 dark world stages beaten, and the fetus stage unlocked. I'm really rusty. Sub-45 might actually possible here, if not even lower, depending on how many 20s+ dark world stages we can throw out.
I just practiced those two stages a bit more, and they're pretty easy now. Next run I'm gonna drop the extra forest stages, leaving 6-2X - 6-5X and 5-3X the stages to skip.
I also just got my first sub 22 any% run today (final time 21:58). I had about 10 - 12 deaths (from what I remember), probably all of which were dumb, and good Larry luck (4 cycle).