Wall Jump Manipulation Finally Explained EZ PZ *YE*

[@hitzcritz]

With this kind of quality, you'll hit 100k in no time. Phenomenal explanation, Mitch!

[@andrewut7ya511]

100m

[@Nick_Whiskey]

Mitch I want you to know that I have learned something today that may be actually useful at my QA job.

[@madladkobra9186]

Dude I love your energy, enthusiasm, and overall showmanship of this game SMB3 lately. The thing that got me hooked on your stuff was how genuinely HAPPY you are (it shows) to teach us these new things your brilliant mind comes up with. That's what I subbed to 2 years ago, that's why I support you now, supported you then, and will continue to support you. You fucking rock man.

[@Mitchflowerpower]

that means so much to me thank you

[@Marsuru1]

“Best Level ever, world 6-9” I never thought about that.

[@carrottopadc]

Mitch, I haven’t played smb3 in probably 11 years, and this information will never be useful to me, but your content is still enjoyable while being informative. Your content is great man, keep it up

[@hurkyl1560]

As for why this works -- speaking for general game mechanics (i.e. I don't know the specifics of SMB3), I imagine hills do different things when you reach them with different subpixel value. So, what you're seeing is *coalescence* -- two different choices of subpixel value on the input can lead to the same choice of subpixel value on the output. So, when you run up and down the hills, you have a chance of eliminating some of the 16 different possibilities for the output, and will reach some possibilities with different probability than others. Put differently, consider the first hill you use. You have 16 different ways you can enter the hill, and 16 different ways you can exit the hill. But if two of those entrances lead to the same exit, some of those 16 exits are unreachable. It might be possible to get better -- maybe even 100% perfect -- results if you try different combinations of which hills you run up and down to see if any other method performs better.

[@juggarnaut64]

Someone might have mentioned this or you might already know, but watching it looks like 2 of the hills are smaller than the others. Most of the hills minus your sub pixel by 5 each time. The sub pixels that are flipping back and forth are 8 sub pixels apart. That means on the first hill if you are at 6 and 14 which was the first value you showed after the hill you were at 1 and 9. Now since there are 15 sub pixels that you can hit if we have a number of hills * 5 that divides without a remainder by 15 then you will have the same sub pixel you started with at the end. To illustrate if there were 6 hills you would times 6 by 5 and then divide that by 15 which comes back to 2. This means that you are at the same sub pixel values you started with. If you think of it as something like a circle and splitting it into 360 degrees it might be easier to visualize. In this case though we are talking about a circle with only 15 degrees. Whenever you have a multiple of 15 you are back to the starting point on the circle. Now here is the harder part because there are 7 hills from my count. This would normally be a problem if all the hills were the same size, because 5 * 7 is 35 which divided by 15 is 2 with a remainder of 5. Again to our circle example you would be 5 pixels further in the circle from the starting point. (I believe it would be a positive direction so +5 pixels to your original one) But if you remember my first point there are 2 smaller hills. I believe at least. I didn't do an exact pixel count. So if those 2 hills equal the same amount of pixel change equal to one other hill then you would minus 2 hills and add 1 as their combined sub pixel count. If that's the case 7 - 2 is 5 + 1 is 6 which is the exact number of those hills that you need. The last slope is longer, but I don't know if the length of the hill matters, the height of the hill, the direction of the hill (up or down), or all 3 but that is a perfect sub pixel loss to get you into the right range. This was just me looking at your video in slow motion and pausing it, so I might be wrong and would have to test it in the game itself, but I think I'm right. That means that you could only run 3 of the hills instead of 6 and still get the same result. This would also have other implications as you could start mapping the other hills in levels to see how much it affects your sub pixel and you could be more consistent in levels where you don't have to turn back. Above sub pixel loss could be sub pixel gain of another value and up or down might change whether it's a loss or gain, but I don't know. Those hills are just perfect for this setup as you perfectly get back to your original sub pixel from the start of the level. Then the last slope perfectly removes or adds to the correct amount.

[@Drpepperer]

Love it when you get into the nitty gritty stuff. Keeps me coming back for more.

[@KaneYork]

That gentle slope at the end... it always adds 2 to your cycle! It must be the steep 1 slopes earlier that are causing the subpixel consistency & probability collapse, while the gentle slope adjusts the results into the desired range.

[@hurkyl1560]

Eyeballing it... isn't the main problem with that first big hill that you need to jump off of it, and can't control *exactly* when you jump off, and thus can't control your subpixel value? I would imagine that second hill would give predictable results, if you run all the way from top to bottom. Of course, I can't guarantee they would be results that work well with the coalescence you see from the final hill.

[@jayjasespud]

These videos are exactly what speedrunning should be: A collaborative effort to push a game far beyond what anyone thought was possible.

[@kenoz765]

That hat chroma really got me...

[@Mitchflowerpower]

looooooool the hat is sooo good i can't not wear it

[@hypnolegsub]

The hat is clipping into the video

[@AtheistJr]

I like it. I think it's a really cool looking effect, and it's kind of like his signature look now

[@tyo6896]

Almost the first, 2022, world 7-1 wall clipping.... tough... thx for all the tricks Mitch.

[@I_Love_Om_Nom34.]

0:01 The hammer bros really got me.

[@Darxide23]

Meetch developing those mad strats.

[@SSGSSVADER]

Ok so the way you explaned everything is so easy to follow great job mitch

[@ricklewis2068]

Without this manipulation, is the wall jump like the one in 7-1 where anybody that just throws themselves at it enough times will get it?

[@masterprawns]

Thanks Mitch I’m sure this video will help a lot of people get a more consistent 6 9

[@lokiwider]

isn't randomness based partially on subpixel value ? Like for early hammer or no hands ?

[@WillowEpp]

P-speed is 56 spf (3.5 pxf) and that hill looks like it has a width of 4 tiles; 64 px. But you gain 3 spx downhill, so it's 17.356 frames during which you gain almost exactly 52 spx (52.067). 52 spx is 3.25 px, which means running down that hill advances your subpixel value by 4 spx. I'm not quite sure why the subpixel _matters,_ but I _do_ have a guess. From what's documented on tasvideos.org, all you need to wall jump are horizontal speed >= 33 spf and to hit the wall on a tile boundary. This makes sense intuitively because wall ejecting is 16 spf (which nominally gives you a two frame window, and you probably lose one to the ordering of ejection vs. input processing), but your experience suggests otherwise. I don't have time to dig deep on this, but I suspect the more correct description is you need at least 33 spx of wall incursion, and your initial subpixel value helps shift that because it probably doesn't start wall ejection until you fully cross the pixel boundary.

[@DannyIO]

with this new information, is that fed into the tas to make the tas faster? or are these ways for humans to catch up? or can we even hope to catch up to it?

[@Kavukamari]

the ghost hat always entertains me

[@justinkendall8341]

Down to a science, it's amazing to see these games & routes broken down and planned out with such precision.

[@MattJohnsonPlus]

So, essentially jumping that last hill doesn't change the oscillating values from the rest of the run?

[@DouglasZwick]

Hey Mitch, you also do a small Mario run. Does this strategy help there too, or does the fact that you can't run 6-8 the same way ruin it? I guess you probably can't use a star in the small Mario category....

[@plutoburn]

Hey Mitch, enjoying these YT vids as well as your twitch livestream. This might be a lot of work but was wondering if you could make a video on the current warpless WR attempt route, esp. W7, what the total time save could be against the WR.

[@dreamlesss]

can u give us the link to the sub pixel practice rom?

[@reveillain9479]

Still don't get how i'm gonna know my value on the boss kills to make a decision of moving or not moving 1 pixel left

[@Mitchflowerpower]

Always move 1 pixel to the left. After you kill boom boom stand still and tap left until you move one pixel, doing this will roll you past 0 to a value depending on how hard you tapped, if lightly (which everyone would) you will land somewhere between 9-15 but 0-1-2-3-5-6 also work too if you pressed to hard

[@reveillain9479]

​@@Mitchflowerpower Thanks, i thought the number is changed linearly (e.g. you stand with 15 value and hard tapping gets you -16 and you get 15 again)

[@l1u1k1e2]

When you're playing these levels for real, and not with the subpixel counter, how do you know what subpixel you get after fighting the boom boom?

[@ozma6918]

Maybe I didn't get it right from your video, but how the heck can you keep track of your sub-pixel value all the time? I mean, you look like you modified the game so that it shows it to you, but would you do it if you weren't able to just have the value in your face the whole time? Also: what if you play a randomizer? You don't necessarily get 6F2 and 6-8 to set it up, yet you almost always seem to get it right regardless. I'm kinda confused on that and would like to know, maybe to pull it off afterwards :) In the meantime I'm trying it live, hope to get it x)

[@TheInfiniteMiseryJumper]

Mitch: the Paul Dirac of retro gaming.

[@Slormer]

Super Manip Bros 3! Keep killin' it bro

[@coyrex1480]

Awesome content Mitch!

[@danielkirienko1701]

That's super cool. How long till you're manipulating subpixels three worlds in advance?

[@cephelos1098]

You can manipulate subpixels as many levels ahead that you can complete with no turnbacks

[@Mitchflowerpower]

@@cephelos1098 or finish the level with a for sure method of controlling them

[@AlanNess]

Felt like I'm watching a science video Don't get me wrong, it's very well-explained and intriguing to know the results and the process of trying with every single subpixel Great video

[@cyberneticwarmachine]

Hey Mitch im new to this( only watched videos) how can a person practice a certain level? Emulator?

[@SpecificallyDanielSwan]

I have never completed Super MArio Bros 3. I don't speedrun. I understood some, but not all, of what you said. I've no idea why I find these explanation videos so interesting. But I do. So thanks!

[@_madism]

Great video as always. The audio was a little too low. Thought I'd let u know

[@AnimatingAnimator]

Another very well done explanation video!

[@Mullet345]

How soon till sub pixel manips are found for 5-Fortress #2 clip and 6-Fortress #3 floorclip?

[@Mitchflowerpower]

i looked into 6-ft3 but it seems like it might be a little to tricky

[@superben1755]

It's still not clear to me how you will manipulate the sub pixel in the boom boom fight during your run. When you don't have the coin counter telling you the sub pixel, how will you know what you're on?

[@DavidJao]

He explained this in a previous video. If you lightly change position by just barely enough to move one pixel then your subpixel is likely to be near the edge of the pixel that you changed position by. Boom boom is the only place during a speedrun where you have extra time to sit around tapping the d-pad until you just barely move a pixel.

[@Mitchflowerpower]

in this video i mentioned about facing left and moving one pixel, when you move left one pixel you have rolled over sub p value of 0 to 15 and then you start moving down from 15 same works for moving right. so if i wanna hit a value from 15-9 i'll move 1 pixel left and most always hit 15-9 and in this case if i miss i'll still most likly hit 0-1-2-3-5-6 its awesome

[@SPbakerbrange]

amazing! keep it up Mitch

[@jochenwuerfel]

Would be extremely interesting if a SMB3 code would analyze what you did.

[@jap274]

Mitch how do you know what your sub pixel value is during a run without using this hack?? So, when you kill boom boom, how do you determine how much to move?

[@Mitchflowerpower]

Always move 1 pixel to the left. After you kill boom boom stand still and tap left until you move one pixel, doing this will roll you past 0 to a value depending on how hard you tapped, if lightly (which everyone would) you will land somewhere between 9-15 but 0-1-2-3-5-6 also work too if you pressed to hard... sorry to copy pasta but its the same answer :)

[@jap274]

@@Mitchflowerpower ahhh so no matter where you are standing, it's always just 1 pixel left. Got it, thanks dude

[@Mitchflowerpower]

Drink every time I say sub pixel values

[@Mitchflowerpower]

already wasted

[@sophiegrey9576]

@@Mitchflowerpower mitch we're two minutes in and you're already wasted

[@youssefl.7574]

Please, we need a new world record

[@Sonicgott]

Nice chromakey ball cap. :)

[@jasonkoch3182]

Can't wait to see a warpless run with these new strats in play. If coronavirus starts forcing me to work from home, I'll definitely be checking out a Mitch stream, too.

[@bhavikabachwani5833]

He is literally trying to get a 48 instead of 49 in warpless

[@LightOfEon]

If the subpixel value ends up being something you don't want from the boom boom fight, could you opt out of running down the final hill in 6-8? Great video btw mitch, I've been watching you for many years now.

[@sophiegrey9576]

Well, you can't really see your subpixel value in the official game, so that'd be a dumb risk to take

[@LightOfEon]

Which brings up the question of how you know what the value would be when running on vanlla cart then? It's probable set up most the time is my guess

[@Showsni]

@@LightOfEon He explains this at 9:14 - basically, each pixel is divided into 16 subpixels from left to right. So if you move left enough that you just about move a whole pixel then stop, chances are you'll be somewhere towards the right of the pixel, in the subpixel range from 9 to 15, which all work.

[@Mitchflowerpower]

@@Showsni ^^^^^^^^ this guy knows

[@omfgbunder2008]

Volume really quiet

[@Mitchflowerpower]

so turn it up :)

[@SuperBazsle]

Just curious. Any reason you're changing the sound effects you use in your intro? Great video regardless!

[@Mitchflowerpower]

yeah! i have multiple copies with different sounds, i just like changing it up :) also some of them have a beard for when i have one and when i dont have one

[@ImVeryBrad]

GG Dude!

[@Glitchman24]

Look maw, no hands!

[@lunaraura2828]

damn thats a nice find!

[@thebiggestoofbob]

Wuts wrong with ur hat

[@theonewhojustis6541]

i was wondering how u was getting that wall jump so many times lol

[@perry4754]

very impressive

[@darkzer0670]

So obviously.....The underside of your hat is green.

[@SPECTER0459]

What? It's transparent, not green

[@ahumanperson3649]

**YE**

[@marixdai8780]

Raided from twitch?...

[@threehead99]

So when is the Jesus clip manipulation coming into play :p

[@KnuxMaster368]

threehead99 Kappa

[@KnuxMaster368]

*YEE*

[@johneygd]

Can’t wait till you do again speed runs untill you set up a new world record in smb3.

[@hdsunn4498]

Yay!

[@alisson_302]

mfpRAID

[@saynotop2w]

Yeet

[@JeffreyBuilds]

Avoid wearing green hats when using a green screen its super distracting.

[@Mitchflowerpower]

not possible, the hat is so good

[@JeffreyBuilds]

@@Mitchflowerpower All good still gonna watch keep up the great content.

[@robd8160]

6-9 mfpKEKW

23rd bitches!!!!

[@XiomaraGarcia-wz5lj]

You take so freaking long to explano!!!!!!!

[@alvaraitor1987]

RAIDDDDDDDDDDDDDDDDDDDDDDD