Quick history from a classic level
Read Full Transcript
This is stage 4-2 in the original Super Mario Bros. It doesn’t really look that menacing. An area at the start with no floor, sure, and an elevator gap just
after. And as soon as you enter 4-2, you’re able to exit it. If you hit these blocks to reveal this fine, you can climb up it and you’ll be above ground.
And there you can warp to either world 6, 7, or 8. This stage has seemed so pivotal to speedrunners who have been trying to beat Super Mario Bros. as fast
as possible for years. It’s a gateway to world 8, arguably the most dangerous part of the game. So, this is a level that’s going to make or break your
speedrun. But as short and simple as this level may seem, little do most people know that 4-2 has caused more pain and agony than just about any other
level in the game. [Music] Come on, [Music]
man. Oh, [Music]
[Music] welcome to the mid200s. This is when the first Super Mario Bros. speedruns were completed. Runners such as Scott Kesler
and Trevor Sean were the pioneers, and they knew that 4-2 is a simple level with a simple goal. run through it, climb up the vine, and warp to world
eight. Doing that quickly wasn’t too hard. Doing it as fast as humanly possible, on the other hand, definitely was.
It was obvious to most back then that the start of 4-2 is really straightforward. You jump across this gap, stomp on a gloomba, and then
continue jumping forward until you’re on the platform. Then you jump off the platform to hit the block that makes the warp zone vine appear. But this is where
it got really tough. You have to tap left to slow down to nearly a stop and then jump to reveal this invisible block. Then you tap left again just
enough to completely stop, landing just to the right of the invisible block you revealed. Finally, you jump on top of the block, onto the top platform, and
onto the vine. Doing this in real time at full speed requires a ton of precision, and it took runners quite a bit of time to get good at it. This is
Scott Kesler’s 510 from April 2004. As you can see, he hesitated a bit while climbing on the blocks and clipped through the vine by accident when he was
trying to climb up onto it. Yeah, the hit detection in Super Mario Bros. is a little bit weird at times, but runners eventually got better at this section.
In Trevor Sean’s 507 from a few months later, you can see that he was able to get a smoother climb to the vine, ultimately ending the section with a 376
on the timer, which counts down from 400 when you start the level. And when he improved his run to a 506 a few weeks later, he once again improved 4-2,
getting it smoother than ever before with a 377 on the timer. So, all right, speedrunners had established a way of getting through 4-2 very quickly. And
given how short the level was, you might wonder what else could possibly be found for speedrunners to go through the level quicker. After all, you already directly
to the warp zone as fast as you could. But if there’s one thing you should never do, it’s doubt the ability of speedrunners to figure out something
even faster. [Music] When you’re playing through Super Mario Bros., you’re occasionally able to leave
the screen you’re on to enter another screen. Sometimes this other screen is an underground coin cache, sometimes it’s a bonus section, and sometimes it’s
a warp zone. Every time you can exit the screen you’re on to access one of these zones, be it a pipe, a vine, or whatever, I’ll be referring to as an
exit zone. Now, 4-2 is a little bit unique. Almost immediately after beginning the level, it has two of these exit zones back to back. The first one
is the vine that you take to the warp zone, and the second one is this pipe a few blocks down that takes you to an underground room with coins. But there’s
a catch. The game can only have one of these zones loaded at the same time. It can’t say if you go up the vine you go here, but if you go down the pipe,
you’ll go here. So, the way the game works around that is by changing what the one destination is depending on how far the screen is scrolled. For example,
if you’re standing here and manage to activate an exit zone in some way, the game will send you to the above ground warp zone. Now, when we move forward
closer to the pipe, suddenly the next zone gets loaded. So, if you manage to activate an exit zone, then you’ll go to the underground section. What this means
is that if you’re somehow able to get into this pipe while the game still has the warp zone loaded as the exit zone destination, the game will send you to
the warp zone without playing the long climbing the vine cutscene that it normally does. The only known way to do this is by moving Mario closer to the
right side of the screen. The game thinks Mario’s farther back than he actually is, so it loads the exit zone destination from where the screen
position is. In this case, it’s the warp zone. So players had to figure out how to get Mario closer to the right side of the screen for this trick to work. You
couldn’t just run forward since whenever you pass 112 pixels from the left side of the screen, the screen automatically starts scrolling, locking Mario at that
112 pixel position. One of the earliest methods of doing this was worked out in the mid200s by tool assisted speedrunners. Toolass
assisted speedruns are a little different from regular speedruns. They’re not performed in real time by a human, but rather somebody goes inside
an emulator and manually tells the game which buttons will be pressed on each frame from start to finish. In 2004, Mana published a tool assisted speedrun
that showed Mario clipping into the wall at the start. This pushed Mario forward exactly 20 pixels. Those 20 pixels were just enough to stop on the first pixel
of the pipe, go down it, and be taken to the warp zone. any farther than the first pixel of the pipe and the screen would have scrolled too far, causing
Mario to go into the coin cache. This wall clip, however, was essentially impossible for a human to ever pull off, which is our main focus of interest
here. It required jumping in just the right way at just the right angle and then mashing the jump button in just the right way once you’re in the wall.
Nobody thought that these inputs could be done back then. Simply put, this method of moving to the right was completely out of the question for a
human. But another method was found also in 2004. Michael F posted a tool assisted speedrun that confused the game into wrapping Mario to the other side of
the screen using the vine. This method saved about 2 seconds over the standard method of climbing the vine, and it was a humanly possible trick. However, still
realtime speedrunners didn’t use it. You see, this method of vine warping was considered to be a glitch. And for many years, because of rules from websites
like Twin Galaxies, glitches were discouraged and usually outright banned from humanly played speedruns. So, the 2000s continued to pass with all
speedrunners playing through 4-2 by running through and climbing up the vine. But eventually, one guy had enough. That one guy was Andrew G. While
he had followed the glitchless rule set for many years, by 2007, Andrew decided to ignore it. So, he began to look at the options he had in 4-2. Method one,
the block clip method, wasn’t humanly possible, so he quickly ruled it out. Method two, the vine teleporting glitch, did save a couple of seconds, but Andrew
wanted something better. 4-2 is such a small stage, though. Was there really anything else he could squeeze out of it? But after a while of searching,
Andrew found something that would forever change the way speedrunners played 4-2. This was method three, the easiest and fastest humanly possible
strategy, the backwards bump method. All right, like I mentioned before, Mario cannot naturally move past the 112th pixel from the left side of the
screen. This is a ROM hack of Super Mario Bros. that displays how many pixels you are from the left side in the upper right corner where the time
usually is. See, it’s locked at 112 pixels. But check this out. If you run right in front of a wall, tap left to turn around, and then jump while
brushing up against the top of it, this number moves forward quite a bit, meaning that Mario has moved forward that many pixels. You usually get Mario
to move forward somewhere between 7 and 10 pixels from doing backwards bumps like this. And once again, you needed to be 20 pixels further to the right for
the wrong to work at all. And the more pixels past 20 you were, the more pixels you had to go down the pipe. If you managed to get Mario to exposition 132,
you could go down the first pixel of the pipe. If you manage to get him to exposition 133, you could go down the first or second pixel, and so on. Since
it was very difficult to immediately stop on the first pixel or the first few pixels, Andrew wanted to give himself some leeway with his pixel gain. So, he
elected to do three backwards jumps in order to guarantee somewhere between 21 and 30 pixels of gain, letting him stop basically anywhere on the pipe that he
wanted. The puzzle that Andrew G then had to figure out was where to do these three backwards jumps. Look at how long the stage itself is. You wouldn’t think
there’s too many different combinations he could try. But some jump locations gave smaller bumps than others. Some involved waiting for piranha plants to
move, and some were impossible to get if you did others first. This was the first three bump method used in April 2007. [Music]
Notice how he’s able to go down the middle of the pipe and the wrong warp still works. That’s because these three bumps were able to push him far past the
minimum 20 pixels needed. When he returned to the game a bit more than 3 years later, he had a new route. Brushing down on this section and then
the same two pipe jumps as before. But finally, a year later, in December 2011, Andrew was able to get some help from another guy, Weirwindle 111, to figure
out the ultimate threebump strategy. [Music]
[Music] It was able to get him into the pipe at a 375 instead of a 374 and ultimately ended up saving one frame rule over the
previous three bump methods. Oh yeah, you might not know what a frame rule is. [Music]
Because of what runners call the frame rule, you can only lose or gain time in increments of 21 frames or approximately.35
seconds in Super Mario Bros. in every level other than 8-4. This is because at the end of every level, the game only checks for completion every 21 frames
instead of every frame. To quote current world record holder Darbian, it’s as if at the end of every level, a bus arrives every 21 frames to bring you to the next
level. Even if you miss the bus by just one frame, you still have to wait 20 frames for the next one to arrive. Sometimes though, you’ll arrive just
before a bus comes, and you’ll hardly have to wait at all. Anyway, for years, this particular three bump method was the fastest method that any runner was
attempting to complete 4-2 in a speedrun of Super Mario Bros., but Andrew G did know of one potential faster method. If done correctly, you would go into the
pipe at 376 and save a frame rule over the fastest three bump method. However, at the time, this was by far the hardest and most random method any runner had
ever come up with to complete any stage in Super Mario Bros., let alone just 4-2. It would lead to years of resetting speedruns over and over again in 4-2
just to save those.35 seconds. Ladies and gentlemen, I present to you the two bump method, better known as fast 4-2. [Music]
If you recall from before, you need to get Mario to exposition 132 from exposition 112 for the wrong warp to work. Those 20 pixels were possible to
obtain from two bumps. Since you get between seven and 10 pixels from each backwards bump that you do. So if you got 10 pixels from each one, then you
could be exactly on exposition 132, stop on the first pixel of the pipe, go down it, and be taken to the warp zone. Doing only two bumps instead of three would
save another frame rule over the old three bump method. Andrew figured out that the two bumps that gave you the greatest chance of getting 10 pixels
each were right here. You could sort of control how many pixels you got from each bump, but not reliably. It was so precise that getting 10 pixels was
essentially just luck. On average, you’d get the 20 pixels you needed, maybe one in every four tries. But that was just the beginning. If you happen to make it
to exposition 132, you then had to stop on the first pixel of the pipe and go down it. If you stopped even a pixel too soon, you couldn’t go down the pipe. And
if you’re a pixel too late, you would go down to the coin cache instead of to the warp zone. Believe me when I say that doing this was ridiculously hard. But
doing this combined with needing 10 pixels from each of your two bumps nearly impossible. And to illustrate just how difficult I’m
talking about, I’m going to refer to a video from a guy named Patrick Lemieux. Lemieux took a stream of hundreds of world record attempts from Andrew G and
overlaid them on top of each other. Like technically could be possible. Like I’m not I’m not usually Whenever an attempt died, Lemieux simply
removed that run from the overlay. For the first three levels, hundreds of Andrew G’s were tied with each other to the frame. Those same Andrew G’s entered
4-2 to attempt fast 4-2. But out of all those hundreds, just one made it to the warp zone on fast 4-2 pace.
Whenever a run happened to make it past 4-2, Andrew fell silent. The voices of a 100 Andrews that were active just a minute before now didn’t say a word.
[Music] These fast 4-2 attempts continued for months and eventually years. Andrew G lost at least 95% of his speedruns that
made it to 4-2 because he needed to save a third of a second. Eventually, runners stopped trying for fast 4-2 because other easier time saves were found in
other levels. But by 2016, the world record got back to the point where the fast 4-2 frame rule needed to be saved for a good chance at the world record.
But runners hated doing 4-2. It was stupidly hard and largely out of your control. Derbian, Cosmic D12, Andrew G, none of these guys really wanted to do
Fast 4-2 anymore. Was there anything else they could squeeze out of such a small stage? Remember before how I mentioned that Weirwindle 111 helped
Andrew discover the best three bump method to fast forward S2? Well, it turns out back in 2011, Weirwindle also discovered something else.
[Music] Using a task, we window was able to clip partially inside of a block and get pushed forward several pixels. Runners
didn’t look into this trick anymore back in 2011, however, because it didn’t actually push you enough pixels ahead to get the 20 pixels needed, only about 17.
And the inputs that Weir was doing required pushing left and right on the controller at the same time, which wasn’t possible for a human. So, runners
forgot about it for a couple of years. But one day in 2016, Andrew G realized a couple of things. One, you didn’t actually need to press left and right at
the same time to get this to work. And two, you could just do a bump earlier to make up for the pixels you’re short of 20. If you did all this really quickly,
you would potentially have a few more pixels to go down the pipe and still be taken to the warp zone. Unfortunately, it was really difficult to get enough
pixels from this clip while still being fast enough to save the frame rule needed to match fast 4-2. So, Andrew G went back to the drawing board. He
remembered the old tool assisted method of clipping into the blocks at the start. His attempts at this trick from years in the past nearly universally
resulted in him not clipping in. And on the extremely rare occasions when he did clip in, it was far too slow to match the fast 4-2 frame rule. But that had
been a long time ago. He had gotten a lot better at the game since then. So he tried this method again one day in early 2017.
Oh my gosh. Wait. [Music]
Holy crap. I can’t believe this. About 100,000 more times. What in the world? [Music]
Holy crap. Holy crap. Unbelievably, the once considered impossible task method was now humanly possible. If done correctly,
it could just barely be fast enough to match the fast 4-2 frame rule. And while Andrew G did find some immediate success with this trick, in the long run, his
success rate with the trick started to go down. and no other top level runner, Cosmic T12, Derbian, or some West for instance, was having any notable success
with it. So 4-2 still needed some universally agreed upon fastest and best method. The block clip and wall clip methods were just a bit too
inconsistent, and standard fast 4-2 was just ridiculous as well. But then in 2017, the ultimate method for 4-2 was found, and it came from perhaps an
unlikely source. This guy’s name was XX420 Blazit XXX. He noticed another old tool assisted method of clipping into the wall a few blocks above the ground.
The reason why speedrunners never attempted this was because clipping into a wall in this way almost entirely depends on what Mario’s sub pixel is.
Sub pixels are values smaller than the pixel that tell your position on a very precise scale. You cannot normally control what your sub pixels are as they
rapidly change as you play a level. So most people simply ignored this wall clip method, but not 420 Blaze it. He realized that your sub pixel values are
always the same when you start a level. And this wall was right at the start of the level as well. And it turned out the sub pixels you needed to clip into the
wall just so happened to be the same that you started the level with. So 420 Blazet ran up to the wall without doing anything else first and jumped into it
while tapping left to clip in. A couple more quick jumps and boom, he was in and pushed forward 20 or more pixels to the right. He usually had a few pixels of
leeway to still go down the pipe. And yes, it was fast enough to match fast forward for D-2. Runners were a little skeptical of this method’s viability,
but soon runners like Mav 6771 were showing it off in practice with great success. He could get it one in every few attempts, and slowly other world
record contenders began to implement it into their runs. They realized that this was it. A decade and a half of playing 4-2 and this method was able to send
them to world 8 in the fastest time with the best consistency. And it was all thanks to a guy named 420 Blaze It over the course of 14 years. The amount
of work that speedrunners have spent on 4-2 is remarkable. Game developers designed this level over 30 years ago, and I doubt they could have imagined how
many people would tear it apart, to discover every little jump, every possible region that could be exploited to save time. Every few years, something
new has been discovered that makes the stage even faster and more consistent. And I doubt the journey is done quite yet. Thanks for watching.
[Music] Hey, it’s time to plug my Twitch again. You guys should come follow me at twitch.tv/summoningsalt.
I speedrun stuff there sometimes. Thanks.
Apple Podcasts
Spotify
Google Podcasts
iHeartRadio
RSS