I MESSED UP BIG TIME! *Fixed* Wall Jump Manip

[@alisson_302]

SMB3 is a insane speedgame I watched the 50:57 run by Kirua and 50:55 by Mitch and those runs are so different. Non Early Hammer vs EH, Fire strats vs Hammer strats, 7-9 clips vs Skipping 7-9, and now we have EH manips, a new route and clip/wall jump manips This is mindblowing (sorry for english mistakes)

[@OmnipotentEntity]

Do you think the reason why the acceleration is different is because the end of the fortress was ice instead of because is was a fortress? I'm thinking perhaps the properties of the previous ground was applied for a frame at the beginning of the level.

[@gailonebell2154]

It is completely awesome that you do all this work and then share it with the community rather than just hoarding it for yourself. Really an example to keep and make the speedrunning community great.

[@EmeraldEdge72]

You said that you played the level nearly over 400 hundred times repetitively to make sure you got it right. That is truly the mark of a dedicated man especially one who learns from their mistakes and improves on their mistakes!

[@JonathanHolt1988]

"But the good news is I just have to redo....everything." Umm, I guess that's good news? lol

[@mamonsin]

The good news is it can be worked out.

[@hurkyl1560]

Knowing what to do next is a big advantage over not knowing what to do next!

[@gigaherz_]

Bad news would have been: "the manipulation is impossible".

[@linkbombchu168]

1:33 you see that Hollywood? Such a natural acting talent. Hire this guy right now!

[@danielkirienko1701]

Mitch, upvotes and love for correcting it. That's science, man. People make mistakes. You owned it and I am so impressed. Great job!

[@Zalied]

is it all fortress that do that or just that one. Maybe it has to do with starting a level after the ice physics level

[@billbiff]

The mic sounds awesome

[@Gaant]

This is so nerdy. I love it. Keep up the great work.

[@tophatjones179]

Happens to the best of us. Thanks for the update. :)

[@Hankola]

Dr. Mitch Flowerpower with the seminar. SMB 500 level course for 5 units and a lab.

[@xzZZZ799]

even though i don't do speedruns these videos are so interesting man keep it up

[@apollolux]

Scientific method applied to the most noble of pursuits!

[@danneye-raye1995]

Keep up the good work, Mitch, we all make mistakes and I got mad respect for you and this video update 💯

[@WillowEpp]

I think you might be able to calculate your output subpixel based on the tile width of the hills you run down with a fair degree of accuracy because it's +3 spf for the width of each tile (256sp). From what I understand, that's +18sp/t for normal running (40spf) and +13sp/t for P-speed running. Count the hill tiles you run down, tally your gains with your starting value and modulo 16. Actually, I have eyes, here's a quick try on that run at 12:47 in detail: 1) Your starting value is 2 2) You're +5sp, which seems to be from acceleration, shortly after starting before the first jump 3) After the first hill, you go from 7 to 9 (18sp%16=+2sp) which is what we expect 4) It's not clear what _precisely_ happens when running uphill and you were also accelerating to P-speed at the same time, but the net result is you hit the top of the third hill with an sp of 3 5) Running down the hill you're at (3+13)%16 == 0... which is not _quite_ what we observe (15). Huh. I'm not sure if this is related to P-speed acceleration or a rounding error or if I'm making some other mistake, but we're only off by 1 from theoretical at this point, which is not too bad for a back-of-the-napkin calc like this. 6) The final hill is 4 tiles and you take it at full tilt, netting you another 52sp: (52+15)%16 == 3. ...Well, would you look at that!

[@EebstertheGreat]

Is it just because you are standing on ice when the level ends? Acceleration is different on ice of course, so maybe that affects the start of the next level somehow.

[@hysterheister0947]

Doubtful since x speed is 0

[@hurkyl1560]

@@hysterheister0947 ​ @Bradley Walker There could be other factors at play. I can imagine an implementation that sets a flag when you're on ice so that it knows to use ice physics. And then when you start the next level... it checks whether you're still on ice to clear the flag *after* the first frame of acceleration. Then you'd get one frame of ice physics before grass physics.

[@hysterheister0947]

@@hurkyl1560 plausible. Sure. I guess we need to know if it applies to all fortresses, or just this one.

[@AniVegMinMan]

this absolutely fucking ruled

[@KyrosQuickfist]

Yes, but how do you consistently set the subpixel value? 3:57 - 4:23 you just wouldn't know in an RTA run.

[@theonewhojustis6541]

damn, nice job with the spreadsheet, that's legit mitch

[@femkeyt2635]

I think you should give yourself a little bit of a break. These are uncharted waters you're exploring so mistakes might happen. It's just amazing how much work you are putting into these strats and freely releasing to the public, that's really cool of you imo. Your strats are so next level, you're a genius man!

[@MikeMorris0484]

This video is something you see out of a MLM campaign. I feel like Mitch is about to sell us something

[@fruityliciousk2704]

Love u 4 putting energy to that kinda stuff

[@thexbigxgreen]

You mean I've been operating all this time with the wrong information?! Good thing I don't run SMB. You're a boss dude, keep up the good work!

[@keirangtp]

Should be a data scientist

[@TrentKallust]

It's easy to follow that if you start at a certain value you can manipulate it from there, but when you're doing an actual run, how do you know/control the starting value? It seems hard enough to set it up when we can see it in the coin counter, but when you are doing a run do you still have access to a tool like that in-game or do you have some other superhuman trick up your sleeve? Asking for a non-speedrunner friend...

[@McCrippin]

Hey Man What is the name brand of your PC or did you build it yourself

[@elliottabaza6348]

Respect

[@JD-mz1rl]

How do you know if the final subpixel value received on the fortress also won't be influenced by what happened on the level preceding it? i.e. in your fortress testing, weren't you always reloading from the same save state, meaning that on a different playthrough with a different save state, the final subpixel values at the end of fortress might end up being something completely different than what you tested?

[@blanksymortimer4088]

there is no manipulation for the preceding level. you set up your pixel at the end of the fight and then do the manipulation. that's why he's testing every subpixel because you can't tell what pixel you'll get before the fight.

[@JD-mz1rl]

@@blanksymortimer4088 yes, but he thought he was doing that on the level after the fortress, but then realized the fortress was impacting that level too. What's there to say that the level *before* the fortress also doesn't influence the value received *at* the fortress? Wasn't all his testing at the fortress all done from the same save state? If yes, that would suggest different values will be received on the fortress depending on how the previous level ended.

[@blanksymortimer4088]

@@JD-mz1rl there is no subpixel needed before the fortress. The subpixel value is made in the fortress and is not affected by previous levels. The subpixel can be any random value up until the end of the fortress. Any level before the fortress is irrelevant to the fortress subpixel

[@JD-mz1rl]

@@blanksymortimer4088 ok, so if the level prior wasn't a fortress, then how is the subpixel manipulation determined?

[@blanksymortimer4088]

@@JD-mz1rl bruh I give up on you just do something else.

[@o0shivashakti0o]

Sounds good.

[@tctrainconstruct2592]

4:37 It's totally 15 and 3, not 13 and 5 :P

[@Jesusholmes64]

i didn't know you move faster in the fortress, doesn't the world 8 airship have different mario physics also?

[@jamesgagne7895]

give your damn spreadsheet a title!

[@meddles123]

Thanks

[@subliminalmemory]

why do i keep trying to scroll down your chart? 4 TIMES. not once, or twice, not even three times, but four times i tried to scroll

[@WiseSquash]

@4:46 wth? how did that spike die??

[@KnuxMaster368]

When you make the h jump H H H H H H H

[@chitownsuperfan]

I can't even fathom the amount of work this video had to take to make. Is smb3 even fun for you anymore?

[@mikemoore7232]

You're fucking nuts. I love you.

[@soythenesmaster]

Amazing

[@barsaf9989]

Hard to focus with the creepy world music in the background lol.

[@Themikemont8]

Love it

[@NoLifeDax]

11k view 85 comments wow

[@coltengoll9371]

Nice

[@SSgtBigSwole]

What?

[@StepBaum]

Great video and explanation. Good thing youre redoing your vids when you realize you fucked up. Glad to see whats gonna come of this! but pls lower the music next time, would also be good for your throat since you dont have to yell that much

[@hotdogwater7037]

how could you, man? i feel like everything I know is wrong....jk

[@mjsantia4]

Is this another April Fools joke Mitch? XD

[@Jiren420]

Omg you absolutely did not have to keep showing us 8 freaking times we got it the second time

[@gburlo]

did i do it. first

[@theSato]

Yes first

[@04groombra]

Still not entirely sure how this is helpful in a run. When you can't see your sub pixels in the normal game. The way the sub pixel thing works basically needs you to be a human tas! Playing every level perfectly. Or being able to correct your sub pixels on the fly.

[@featherpony]

Speed runners like to use the term "subpixel" to refer to the units of movement less than one pixel. But in reality, there is no such thing as a subpixel. I will explain. Graphically, the processor doesn't process less than 1 pixel per frame. Instead, what is happening is that the processor is running instructions in cycles per second (hertz), which is based on a unit of TIME. This becomes evident when too many enemies are on the screen at once and the game's clock slows down, but the CPU clock is still running at the normal frequency. This discrepancy has, somewhat inaccurately, been called a "lag frame". But the CPU doesn't care about frames or even pixels. From the CPUs viewpoint, it is just processing a consistent stream of instructions. The "sub pixels" you see are not graphical in nature. Those are remaining clock cycles left over after the screen has updated a pixel. Think of it as buying a package of 10 hotdogs and a bag of 8 buns. The processor, not caring about what you do (or don't) see on your screen is already processing more instructions ahead that have not resulted in a frame update, yet. This includes calculations based on the CPU clock (not the game clock). To humans, it seems easier to look at values as pixel positions and frames. But the real values have nothing to do with either. It is based on what cycle number the CPU is on. And this CPU clock would still be counting, even if those pixels and frames didn't exist.

[@WillowEpp]

Is this some speedrun copypasta I've never seen before? Needs more Bel Air.

[@featherpony]

@@WillowEpp No, I wrote it all myself, just now. I know, it seems impossible, right? All millennials know is how to write 280 characters or less. Can you imagine what news articles, magazines and books are going to look like when Millennials are the oldest living generation? I pity the future. Your comment was 75 characters, by the way.

[@hurkyl1560]

Mario's actual X-position can easily have a finer resolution than a pixel -- in fact, it is a relatively standard technique for the CPU to compute with more precision than can be displayed. Given the actual explicit display Mitch is using of the data, it seems fairly certain that the resolution the SMB3 engine uses is, in fact, 1/16 of a pixel.

[@WillowEpp]

@@featherpony I didn't think it was copypasta because it's long. I thought it was copypasta because it's a bizarre non-sequitur with just enough technical correctness laced in that I can't call it _entirely wrong_ without getting a bad taste in my mouth. It was an amusing feat that reeked of intention but lacked a punch line outside of the misplaced confidence of its delivery. As for the rest of your screed, I have no idea why you think this is suddenly about "Millennials" or whatever, but if you're going to be condescending, at least point your rancor in a productive direction. So yes, if you're truly not trolling, then it is with great, _gleeful_ Schadenfreude that I inform you: Mario's X subpixel position in SMB3 can be found at address 0x074D. This is not an accident. Representing game object position with higher precision than raster scanout can display is a very common technique and the jargon surrounding it is in no way new. (And if you want significantly more than 75 bytes where I reify this concept into a practical examination of the game as-played, look for my top-level comment on this video.)

[@featherpony]

@@WillowEpp First off, the term "schadenfreude" is thrown around so much only by people hopeful to impress others. But everyone knows this word as it was published and talked about by a very large KZbin channel. This same thing happened when "geniuses" started using the word "oxymoron" after learning it from Danny DeVito in a movie. Very unimpressive. Secondly, your "top rated" comment has no thumbs up and is buried several pages deep. Perhaps you are unaware that KZbin elevates your comments towards the top ONLY for you, because you wrote them. Not because they are top rated. All this time you thought your comments were the best on KZbin. Embarrassing. Lastly, underclock or overclock your NES's CPU and watch your whole subpixel theory explode and sink. That memory address is not being written by the in-game timing. Lag affects it.

[@travisflesher89]

Whatever you say goes Mitch. Your SMB3 critical analyses are always entertaining. But don't I distinctly recall your repeated affirmations and video proof of absolute wall skip mastery via subpixel manipulation in the previous vid? Anyhoo, keep up the great work bro!

[@Wheatsolo]

Do you think RTA runs are about to understand clips/wall jumps where adding more becomes a boon rather than a headache?