Submission #7176: DyllonStej's Wii Super Smash Bros. Brawl "The Subspace Emissary, 1 Player" in 1:03:43.98

Wii
The Subspace Emissary, 1 player
Dolphin 5.0 Lua + TAStudio
916834
239.75891106210463
276103
Unknown
Super Smash Bros. Brawl (Europe) (En,Fr,De,Es,It) (v1.01).iso
Submitted by DyllonStej on 7/24/2021 8:05:04 PM
Submission Comments

Emulator/Sync Settings:

  • Emulator Used: Dolphin 5.0 Lua + TAStudio (Syncs on Dolphin 5.0)
  • Dual Core and Idle Skipping Off
  • DSP HLE Emulation
  • Use PAL60 Mode
  • Use 1 GameCube Controller, 0 Wii Remotes
  • ISO MD5 Checksum: c2255d53e08d6e1e6d25114be9f1bec1

Game Objectives:

  • Completes the main story as fast as possible
  • Uses hardest difficulty
  • Takes damage to save time
    • Takes damage to save time
      • ...Like, WOW, that’s a lot of damage
  • Uses death to save time
  • Heavy glitch abuse
  • Heavily manipulates luck
    • Heavily manipulates luck
      • Breaks my $12 scientific calculator when trying to calculate the full odds
  • One player in a multiplayer game
  • Genre: Fighting
  • Genre: Platform

Game Overview:

Super Smash Bros. Brawl is a fighting game released for the Nintendo Wii in 2008. While its core mechanics are the same as its predecessors (Where the genre’s traditional health bar is replaced with a percent counter, with the goal of knocking opponents off the stage instead of reducing their HP to zero), Brawl changed many aspects of the game from Super Smash Bros. Melee, being generally considered the “slowest” game in the series as a result of some of them. Brawl is also the first game in the series to have a dedicated story mode, called “The Subspace Emissary”, and is the main focus of this TAS.
This is an improvement to my previous submission, thanks to several things found in the process of making the old one alongside new things found in the process of this one.

Region Choice:

PAL was used over NTSC because of two version exclusive death warps; all other changes are purely aesthetic.

Difficulty/Category Choice:

Factors like enemy damage and autoscroller speed increase as difficulty increases, making Intense Difficulty actually faster than easier difficulties in most parts of a TAS. The only thing that slows Intense down in comparison to lower difficulties is the fact that bosses have more health and characters take slightly less knockback, but they go down so quickly that it hardly matters.
It is possible to skip the difficulty selection at the beginning and save about a second, but this brings the game into a glitched difficulty state. Notably, the speed at which particular block chains break (Seen in Room 5 of The Lake among other places) is so glitched that only a single block breaks, creating a softlock; this requires one to exit the game and re-enter to choose a proper difficulty to proceed.
This is a single player TAS of a mode that supports up to two players, but the addition of a second player reduces damage output from both characters by 0.6x (Which means all enemies effectively gain 67% more health).

Game Mechanics:

Basics

A Gamecube Controller was used. Primary movement is done by moving the control stick - this includes Running (Fully Left/Right), Walking (Half Left/Right), Jumping (Up), and Crouching (Down). Pressing the A button in combination with the control stick allows characters to attack with a variety of moves (Including Jabs, Tilt attacks, and Smash attacks on the ground, and aerial attacks in the air), and pressing the B button in the same manner allows unique Special attacks to be used. The Z button is used to grab opponents, and the L/R buttons are used to shield against or dodge enemy attacks. The C-Stick allows Smash Attacks to be used on the ground, and aerial attacks in the air. The D-Pad is used to perform character exclusive taunts, but they don’t do anything meaningful aside from looking flashy.

Other Techniques

  • Dash Dancing: By turning in the opposite direction in the first six frames of a character’s dash, the character dashes again. While its usage in combat is severely limited in comparison to Melee (Which has a larger time window), it can be quickly used to generate dust and advance the RNG (Explained below).
  • Fox Trotting: The act of repeatedly entering a character’s initial dash animation (Instead of transitioning into a run) by rhythmically tapping on the control stick. Several characters have Fox Trots that are faster than their run speeds.
  • Turn-around Jump: The act of jumping in the middle of a character’s turn-around animation; this is used to better position needed aerial attacks.
  • Shield Drop: By holding one’s shield and slightly lowering the control stick, one can drop through a platform. It’s one frame faster than the traditional method (Holding down for two frames while in an idle or turnaround position), and also retains momentum from a dash.
  • Directional Influence (DI): The ability to influence the trajectory one follows when launched by an attack. It is affected with precision down to individual control stick values (With the exception of the control stick’s deadzones) when bouncing off of walls or ceilings, and precise to cardinal and diagonal directions otherwise.
  • Smash DI (SDI): The ability to influence one’s position before being launched by an attack, possible during the attack’s “hitstun” period (When the attacker and attackee appear frozen). Unlike DI, SDI is only precise to the cardinal and diagonal directions. One can input a diagonal after a cardinal (I.E Left, then Left+Up), but not the reverse (I.E Left+Up, then Left).
    • Shield SDI: Similar to SDI, but when shielding an attack. Can only move left or right.
  • DACUS: Short for “Dash Attack Cancelled Up Smash”, performed when a character cancels a dash attack with an Up Smash. Can provide significant boosts in speed, depending on the character.

RNG:

RNG is a deceptively large factor in the process of TASing this game, ranging from individual characters’ randomness (Peach’s Turnips) to the entirety of the item drop system (Explained below). Enough about the RNG was known for me to make an RNG brute-forcing LUA Script - many of the spectacles in this TAS would have been considered impossible without it.
The RNG’s RAM address is 0x805A1E8C, with values taking the form of a 4-Byte Hexadecimal. New RNG values are almost always generated by taking the preexisting value, multiplying it by 0x41C64E6D, adding 0x3039, and using the last eight digits (The lone exception is during room transitions, where the RNG changes differently from this pattern and “jumps” from one RNG sequence to another). The RNG value changes many times in a single frame, with the exact amount depending on the amount of things calling it (Many enemies on the screen at once bring the call-per-frame number to over 40, for example). Numerous player actions also change the RNG (Generating dust from running, jumping, attacking, pausing and unpausing, etc.), which makes manipulating it a relatively quick process.
Thanks to RNG’s dominating presence in almost all rooms and its nearly nonexistent stability (Only remaining constant between level selection), it’s effectively impossible to hex inputs without RNG desyncing.

The Item Drop system:

On the frame an enemy is defeated, they drop a random item the character can use. The general item probabilities are dependant on the enemy itself (With specific enemies having their own individual item pools), the stage or room one is in (With some specific items not showing up at all, like Super Mushrooms in most of The Battlefield Fortress), and the game’s difficulty (With higher difficulties increasing the chances of Trophy Stands and Stickers dropping, for example). As RNG is called numerous times per frame, and that every value corresponds to its own item, many items are initially “trapped” in-between frames, requiring RNG manipulation to make them appear on a whole frame. Notably, pausing and unpausing the game up to three frames before an enemy dies prevents the RNG from moving up to six values (Two values per pause frame) in comparison to normal; this means a given frame of enemy death has four different item outcomes - one regular outcome, and three pause outcomes.
The brute-forcing LUA script I created takes a starting RNG value, generates a sequence of RNG values from it, and iterates through each value as an enemy dies - logging the items that appear from their respective RNG values. This simulates killing the enemy on different frames, and also highlights the items obtainable from surface-level or pause-level manipulation. Its only downside is that it has to go through every value one at a time (Since any other pattern would miss potentially useful items), making it inefficient when RNG is called a high enough amount every frame. The script also only properly predicts when items appear if the RNG changes the same amount in-between frames; deviations from this makes the script fail, and was a notable hassle to fix in several areas (Like near the end of the Great Maze, as the people who watched my progress on it can attest).

Item Odds

While the exact item odds still aren’t known, I was able to use my script to generate a list of 10,000 item drops across a variety of levels and fights (The RNG is slightly different between a horde fight or an ambush in comparison to regular non-fight side scrolling areas); this should give a good enough approximation for the actual item drop odds from a normal enemy. I also logged some data for other enemies and item drop situations (That don’t correspond with the regular item pools) that were used in the TAS, so I could get a complete set of data for the total item odds.
(POST-PUBLISH UPDATE: Exact item odds (Not including sticker subtypes) were eventually found for every individual room, but the tables remain unchanged due to the hassle of fixing them; they're still on the same order of magnitude in terms of the correct values)
ItemAverage Percentage
Sticker29.72%
Nothing21.93%
Trophy Stand19.98%
Food12.98%
Bob-Omb3.28%
Super Mushroom1.54%
Hammer1.31%
Poison Mushroom1.27%
Blast Box0.74%
Motion Sensor Bomb0.63%
Lipstick0.605%
Smoke Screen0.58%
Ray Gun0.59%
Super Scope0.55%
Unira0.55%
Star Rod0.58%
Soccer Ball0.55%
Screw Attack0.53%
Beam Sword0.61%
Smart Bomb0.52%
Mr. Saturn0.49%
Hothead0.49%
Fire Flower0.54%
Green Shell0.47%
Franklin Badge0.18%
Superspicy Curry0.15%
Fan0.14%
Cracker Launcher0.135%
Homerun Bat0.13%
Golden Hammer0.095%
ItemAverage Percentage
Nothing56.29%
Sticker7.33%
Trophy Stand5.60%
Bob-Omb4.47%
Blast Box3.67%
Smart Bomb3.37%
Food2.52%
Poison Mushroom1.69%
Super Mushroom1.57%
Green Shell1.02%
Hammer0.93%
Lipstick0.93%
Super Scope0.93%
Unira0.92%
Soccer Ball0.85%
Fire Flower0.83%
Mr. Saturn0.82%
Hothead0.78%
Ray Gun0.78%
Smoke Screen0.77%
Motion Sensor Bomb0.66%
Star Rod0.59%
Screw Attack0.59%
Beam Sword0.51%
Fan0.45%
Homerun Bat0.38%
Cracker Launcher0.36%
Curry0.29%
Golden Hammer0.10%
Franklin Badge0.00%
ItemAverage Percentage
Nothing77.23%
Sticker3.92%
Trophy Stand3.17%
Bob-Omb2.28%
Blast Box2.23%
Smart Bomb2.09%
Food1.30%
Hothead0.64%
Motion Sensor Bomb0.57%
Soccer Ball0.55%
Ray Gun0.53%
Green Shell0.49%
Smoke Screen0.48%
Screw Attack0.48%
Mr. Saturn0.46%
Fire Flower0.46%
Lipstick0.44%
Unira0.42%
Super Scope0.41%
Hammer0.37%
Star Rod0.30%
Curry0.30%
Cracker Launcher0.22%
Fan0.20%
Homerun Bat0.19%
Beam Sword0.18%
Golden Hammer0.09%
Super Mushroom0.00%
Poison Mushroom0.00%
Franklin Badge0.00%
ItemAverage Percentage
Bunny Hood8.88%
Super Mushroom7.93%
Poison Mushroom2.22%
Soccer Ball24.6%
Trophy9.64%
Mr. Saturn23.1%
Smoke Screen23.6%
ItemAverage Percentage
Sticker51.3%
Food25.4%
Trophy Stand20.9%
Super Mushroom1.3%
Franklin Badge0.91%
Metal Box0.09%
Star0.09%
ItemAverage Percentage
Stitch Face Turnip1.79% (1/56)
Mr. Saturn0.625% (1/160)
Bob-Omb0.417% (1/240)
Beam Sword0.208% (1/480)
The results I’ve found show that several items seen in the TAS (Green Shells, Hotheads, the Cracker Launcher) are actually quite rare items in comparison to all the other ones, with even the more common items (Bob-Ombs, Super Mushrooms) still only having a few percent chance at most in comparison to things like Trophy Stands, Stickers, Food, and nothing dropping at all.

Stickers

Alongside the series’ usual Trophy collectible items, Brawl also introduces the concept of Stickers. There are 700 stickers in the game, and while they look like nothing but more decorative collectibles, they’re actually another large factor in optimally TASing this game. Each sticker has a purpose in The Subspace Emissary, where applying them will give characters unique benefits. This includes increased attack power to a given type of move (Arm moves, Leg moves, etc.), increased defense, and so on. The holy grail of stickers for this TAS is the “Bullfrog” sticker, which has a character spawn with a Bob-Omb in every single room. It’s the only sticker with this effect, and allows for incredible feats of damage boosting when combined with a specific technique (Explained below).
Each sticker has its own unique sub-ID (The Bullfrog is 0x2BB, for example), which allows me to view the stickers I get on the fly in RAM instead of having to go through the entire level. Like regular items, certain stickers’ spawn chances are also affected by the level one is in, and the particular odds of a given sticker dropping ranges from as high as 20 to as low as 1 out of 10,000 entire item drops. A summarized list of actually used stickers is shown below:
StickerAverage Percentage
Bullfrog (Carry Bob-Omb)0.10%
Skull Kid (+20 Leg Attack)0.19%
Barrel Train (+19 Leg Attack)0.09%
Peach (+23 Weapon)0.08%
Ray Mk III (+18 Weapon)0.17%
Boulder (+32 Body/Spin)0.03%
Solid Snake (+19 Explosive)0.06%
Rusl (+11 Explosive)0.05%
Electrode (+7 Explosive)0.07%
Musketeer Daltania (+21 Slash)0.17%
Daroach (+24 Electric)0.19%
King Dodongo (+38 Flame)0.03%
Ho-Oh (+12 Flame)0.11%
Entei (+20 Flame)0.08%
Boundish (+7 Flame)0.06%
Tractor Trailer (+14 Direct Specials)0.03%
Valoo (+19 Indirect Specials)0.16%
As a result of having a proper sticker route this time around and manipulating many more of them throughout the course of the run, there are slight timelosses in several autoscrollers, thanks to my need to quickly pause and unpause the game to manipulate RNG.

Damage Boosting & Bounce Boosting

The standard in Super Smash Bros. is that getting hit with an attack at higher and higher percents sends you farther and farther off the stage, and that one can manipulate where (Through SDI) and how (Through DI) one is launched. This behavior isn’t changed in Subspace.
To prevent characters from being able to fly all over the place by taking ridiculous amounts of damage, the developers attempted (Key word: attempted) to add safeguards related to the camera. In the free-roaming sections, the blastzones are tied to the camera’s boundaries; moving past them causes a character to be KO’d. When a character takes launching damage, the camera freezes in place for a time, with larger amounts of damage causing the camera to freeze for longer. This leads to an intended mechanic where attempting to damage boost all over the place is met with being KO’d from the camera’s frozen blastzones. However, the developers missed some important details that allowed me to do exactly what they wanted to prevent.
At first, it was believed that only electric attacks were capable of unintended behavior: If one gets hit by an electric attack and SDIs into a wall or ceiling before hitstun ends, the camera doesn’t freeze, and the player flies through the level. I found out later that it works with all attack types, but that every type except for electric attacks has much tighter timing (Getting a wall or ceiling bounce on the first frame of hitstun, instead of any frame of hitstun like electric attacks). Non-electric attacks also have the added requirement of directly moving into a wall or ceiling in the instant the attack hits; the trick doesn’t work if the character is frozen in hitlag or not moving in general.
This occurs because the game only explicitly freezes the camera if the character is in the neutral “taking damage” animation; immediately entering a wall or ceiling bounce (Or in the event of hitting a Jyk, air-dodging on the frame of impact) skips this animation, and doesn’t freeze the camera. Because this technique always (Except the rare Jyk scenario I mentioned) requires you to be in the wall or ceiling bounce animation, I’ve decided to name it “Bounce Boosting”, for a shorthand of “Wall-Bounce Damage Boost” or “Ceiling-Bounce Damage Boost”.

Loading

Brawl’s loading is rather infamous in the community for being quite a bit longer than both previous and future entries, no thanks in part to being the first Wii game released on a dual-layered disk. Even though loading times aren’t meant to be considered as official time losses or improvements, there are several instances in the new version that save time over the old TAS due to legitimate loading strategies, like not having to reload certain characters’ data (Yoshi in The Lake Shore) or loading character data early (ROB in Subspace Bomb Factory 2). In the Timing Table at the bottom of the page, I’ll make note of these instances, and also tally entire level timings alongside individual room timings to weed out random loading variants that aren’t from actual strategies.

Miscellaneous Tricks

  • Pause Glitch: Pausing on the frame one enters an Ambush causes the camera to continue following one around instead of centering on the main fight area. The camera’s own blastzones also still exist, which means moving far away enough from an enemy causes it to despawn. This is used in many areas to completely skip past a fight. As the fight area’s blastzones also still exist, however, I can’t move too far on my own without dying - this is why you often see me perched motionless in an arbitrary location with the glitch active - I’m right at the edge of a blastzone.
  • Ambush End Damage Boost: When an ambush ends, the camera goes back to following you instead of centering around an area. If one takes damage before the transition occurs, the camera doesn’t freeze in place.
  • Wall/Corner Clipping: Under very rare circumstances, it’s possible to clip through certain walls or ceilings as if they didn’t exist, through abusing several moving pieces of collision, one’s own collision points, or simply hitting a wall-ceiling corner at the right speed. This is used a few times to go to unintended areas.
  • Map Menu Skips: When the map shows up after a stage clear and new characters join your party, there’s a one frame window that lets you choose the next level before the map menu pops up, saving 2-3 frames. This was also done in the old TAS, but never documented here.

Stage-By-Stage Comments

I’ll be noting all the improvements over the old TAS, alongside the areas where this TAS loses time to the old one. If the strategy isn’t explicitly mentioned, then it’s the same as the old TAS; you can view my notes for it on the old one’s page.

The Beginning

This movie starts from a fresh bootup with a particular starting time in order to unlock a needed sticker by default (An exclusive +Weapon Attack sticker for Peach, which increases the damage of her Stitch Face turnips). While the old TAS created a save file due to unknowns in routing, this one does not, and saves several seconds at the very start. It also appears to slightly decrease loading times related to the map menu across the rest of the game as seen in the time table at the bottom (Where the time saved across the level is slightly greater than the time saved in the sum of rooms); I assume this is because the game doesn’t have to quietly autosave when it doesn’t have a save file.

The Midair Stadium

  • Room 1: Thanks to the differing RNG from not creating a save file, Mario moves differently this time, and runs to the edge more quickly than the old one. Pauses were still done to manipulate RNG for the next room.
  • Room 2: While the strategy for this room is the same as the first revision, slight differences in RNG at the start (Which led me to pluck a Bob-Omb a few frames later) eventually expanded to saving a number of frames by the end.
  • Room 3: I save Peach here instead of Zelda, due to Peach’s superior mobility in damage boosting. Despite my expectations, the new TAS actually saves a small amount of time over the old TAS.

Skyworld

  • Room 1: Fox Trotting with a Bunny-Hooded Pit reaches the door one frame faster than running.
  • Room 2: I take damage in the first fight more effectively here than the last time, making the Pit-Stop™ at the Cymul to grab a Green Shell unnecessary.
  • Room 3: As it turns out, building up damage and launching with a Bob-Omb Bounce Boost is faster than gliding over the entire level. I delay entering the door for a number of frames to manipulate RNG for the next room.

Sea of Clouds

Peach is used over Kirby (And Zelda/Sheik) due to her aforementioned damage boosting capabilities.
  • Room 1: Peach clears this room over five seconds faster than Sheik through cleverly timed Bob-Omb pulls and launches. I found that by dash dancing into an idle animation, one can use her Down-B and continue moving in the direction she now faces, which greatly improves her mobility when using the move.
  • Room 2: In the old TAS, the game spends extra time loading Zelda back in over Sheik, all for me to switch back to Sheik the frame the rest of the level loads; the new TAS simply uses Peach for both, and doesn’t incur extra loading times. While the first TAS also used Bounce Boosting here, I use it more effectively by exploding myself alongside using the Feyesh present.
  • Room 3: Thanks to Peach’s Float ability, she can use aerial attacks while also holding an item (In comparison to other characters, who are forced to throw most items they’re holding); this allows me to RNG Manipulate a Down-B Bob-Omb alongside a Poppant’s Super Mushroom and take out the Generator before Bounce Boosting over to the next fight. The second fight proceeds more or less the same.

The Jungle

  • Room 1: Alongside manipulating Super Mushrooms, I also build up damage to partially skip the cannon autoscroller.
  • Room 2: With the damage I built up in the first room, I can get through this one more quickly.
  • Room 3: This is the first room of many where I manipulate RNG to get a number of stickers, thanks to having an actual sticker route to follow (Unlike the first attempt).
  • Room 4: RNG differences lead to a very slightly different fight, but is otherwise identical to the old TAS.

The Plains

  • Room 1: This follows the same pattern as the first TAS, but fits in extra Bounce Boosts where the old TAS didn’t.
  • Room 2: I found by accident in a hacked Project M TAS roughly four years ago that one could trigger the level’s ending while still in the final fight, and wrote it off as something to do with the hacks used. As it turns out, one can achieve this legitimately: By reaching an X Coordinate of exactly 1,300, the ending trigger activates. This is precise down to a hundred-thousandth of an in-game unit (Where characters move several units per frame), so the autoscroller’s down-time was used to properly position myself before damage boosting to the end.

The Lake

  • Room 1: I use an attack pattern with very slightly more damage per second (Full-Jump Up Air into Back Air instead of Short Hop Fast Fall Back Airs) and get better attack patterns from Rayquaza itself to finish a bit faster than the old version.
  • Rooms 2-7: By pressing B right before Fox’s Side-B activates, he cancels the full charge; the last possible cancel frame allows him to carry a large amount of momentum instead of effectively nothing, which saves time for a number of rooms. Room 3 has the Koopa’s damage boost done more effectively, as teching right at the edge allows me to fall off and reach the door without needing to use Side-B. Room 4 sees another instance of Fox Trotting, as it’s faster than Fox’s run by one frame. Room 6 has a shortened Side-B to get over the first Towtow, and a damage boost from the Auroros to reach the end more quickly. Room 7 has another shortened Side-B.
  • Room 8: Instead of using a Super Mushroom, I manipulate several Bob-Ombs and Bounce Boost to the end after the fight finishes.
  • Room 12: More stickers are farmed here in comparison to the last time, which still didn’t have a route planned at this point; it slightly loses time as a result of the extra pause buffers needed for RNG.
  • Room 13: I manipulate a Bob-Omb from a Goomba and Bounce Boost to the cannon instead of using the Koopa.
  • Room 14: Bowser was much less cooperative in using his Side-B this time in comparison to before, which made me lose a slight amount of time in this room as well.

The Ruined Zoo

  • Room 1: Unlike the last TAS, I actually reach 999% here. It’s all for show, though.
  • Room 2: I’m able to get on-top of Porky and start Down-Tilt spamming sooner than the old TAS, eventually resulting in a slightly faster time.
  • Room 3: As I mentioned in the last version’s writeup, Bounce Boosting is incredibly useful here; I’m also able to make clever usage of a Soccer Ball and a Motion Sensor bomb to launch me right to the exit after the fight finishes.
  • Room 4: After trying several different things, it turned out that the first revision actually did the fastest thing for once; other time improvements come from having higher damage for the Glunder’s Bounce Boost.
  • Room 5: I manipulate a Bob-Omb from the Primid and use the spring’s wall hitbox as a Bounce Boost launch option to fly right to the end.

The Battlefield Fortress

  • Room 1: Bounce Boosting was used repeatedly to fly through the castle alongside the Shield SDI used in the first TAS.
  • Room 2: This room is slower than the old TAS thanks to the damage I have here preventing me from damage boosting like the old one did. I am able to close the gap by having the Roader ram into me at the start of the level, and the timesave is still positive overall for these two rooms.
  • Room 3: RNG differences and fun with a Cracker Launcher eventually lead to time being saved in this fight.
  • Room 4: Thanks to Bounce Boosting, I’m able to skip the autoscroller on Intense this time (Which I previously thought was impossible), saving a number of seconds. I found that repeatedly Dash Attacking as Meta Knight is faster than his forward roll and run speed, so that’s used in places that are too small for his Infinite Dimensional Cape movement technique.
  • Room 5: By manipulating Bob-Ombs (Done in the first section by spamming Up Airs) and Bounce Boosting at the right moments, I’m able to outright skip several sections of minecart riding.
  • Room 6: I reach the end slightly sooner than before by ending the Infinite Dimensional Cape right at the trigger.
  • Room 7: The gliding in the Old TAS was just slightly unoptimal; I improve it here.

The Forest

  • Room 1: I get through the first fight more quickly by turning sooner and building up damage, get through the rest of the room with the help of Bounce Boosting, and do the second fight more optimally to save even more time.
  • Room 2: The damage I built up in the first room allows me to bounce boost to the fight slightly faster than regular movement; I then manipulate a Bob-Omb from the Puppit to reach the end sooner.
  • Room 3: The increased damage leads to better damage boosts from the Feyesh, alongside a new boost from a Shellpod that forgoes the need to hit the switch and activate the platforms.

Research Facility 1

  • Room 2: Dash Attacking into the elevator saves a few frames over just running into it.
  • Rooms 3 & 4: I’m able to edge-cancel Pikachu’s Quick Attacks more effectively than the last time to save frames here and there.
  • Room 5: This is the only instance where I actually go through a fight that I previously skipped, as it makes the return trip much faster. I go through it more quickly by using the Pause Glitch, and make the return trip faster by Bob-Omb Bounce Boosting.
  • Room 7: The fight at the end of the autoscroller can be done more quickly by killing the first Glunder, getting a Trophy Stand, and throwing it so that the second Glunder dies the instant it spawns in. The return trip is faster thanks to my higher damage.
  • Room 8: Bounce Boosting to Room 7’s door is slightly faster than the movement used before.

The Lake Shore

  • Rooms 1 & 2: I use Yoshi over Link because of his faster mobility. The different characters in comparison to the old TAS is a byproduct of saving Peach instead of Zelda; the first room saves some time, while the second room loses a bit of time, but the time saved from Peach in Sea of Clouds still makes this a timesave as a whole.
  • Room 3: As Yoshi is chosen in this room regardless, the new TAS saves extra time over the old TAS by loading, as Yoshi in this one was already loaded in the Room 2 fight. I’m able to get through this room’s own fight without Yoshi dying this time, which means he also saves time in loading at the start of Room 5.
  • Room 5: Bounce Boosts are used frequently throughout the room, and the last fight is done more optimally.

Path To The Ruins

  • Room 1: I build up more damage and launch earlier than the first one, allowing me to reach the end more quickly.
  • Room 2: By damage boosting at the exact moment the pillar moves down with the third fight’s end, I can clip through the ceiling (By snapping up to the pillar’s floor before it moves too far down) and briefly move out of bounds. This gives me enough time to airdodge without teching on the room’s other ceilings and walls, which now means they can only guide my momentum around instead of changing or outright stopping it. I have enough speed here (And the correct positioning of my collision points, which may or may not be because of the Super Mushroom) to outright clip through the corner of the ceiling and the blockade’s wall. This skips both the final fight and the autoscrolling section before it. I actually found this type of collision clipping in The Great Maze of the old TAS, but couldn’t find any good opportunities to use it. This was the first instance where I did.
  • Room 4: I get Wario to follow me to the blastzone without needing to Up Throw him, which saves time.

The Cave

  • Room 1: RNG differences lead to a different fight than the first time, but the room still ends up having time saved from an earlier Bounce Boost.
  • Room 2: A different SDI pattern saves a few frames over the old TAS.
  • Room 3: The strategy used is the same as the last time, but optimized better in several places (Including the Koopa damage boost and starting the Egg Roll at the Giant Goombas sooner)

The Ruins

  • Room 1: The bounce boosting here is better than the first TAS, allowing me to reach the end sooner.
  • Room 2: I manipulate several more stickers here in comparison to the old TAS, at the cost of finishing it 20 frames slower thanks to pause buffering for correct RNG. The time saved from getting all the necessary stickers more than makes up for this, however. I also build up damage to make the final boost after the autoscroller skip slightly faster.
  • Room 5: I’m able to manipulate a Hothead much sooner in the level, skipping the need to manipulate one from the enemy after the autoscroller is done. I also use a Bounce Boost from the Bytan to move more quickly.
  • Room 6: Different RNG allows Charizard to be dealt with more quickly than the last time.
Thanks to the amount of stickers manipulated for the several more characters here over the old TAS, I spend more time applying them. The time lost here is offset by the numerous extra timesaves further in the run.

The Wilds

  • Room 1: Taking more damage allows me to perform the Mountain Clip much more smoothly than the last time.
  • Room 2: The extra damage prevents me from landing quickly enough like before, however, which means I fly off to my death as the autoscroller begins. I have enough time to get Meta Knight back and re-build up damage again before the fight begins. I also perform better movement after the fight.
  • Room 4: The extra stickers on Meta Knight allow me to finish Galleom off more quickly than the old TAS.

The Ruined Hall

  • Room 1: The stickers I put on Charizard do the same thing here.

The Wilds 2

  • Room 1: I also gave Pit a Bullfrog sticker to make his section slightly faster; Yoshi’s extra stickers allow him to damage boost with his own Bullfrog more effectively than the last time.
  • Room 2: I start the bounce boost sooner than the last time, and also slightly optimize the ladder movement at the end of the autoscroller.

The Swamp

  • Rooms 1 & 2: Fox also has his own Bullfrog, which makes these two rooms much faster.
  • Room 5: RNG differences lead to Diddy following me more slowly, but I still save time overall thanks to not damaging him this time.
  • Room 6: I use the Bob-Omb from the Bullfrog to outright skip the fight present, but the time saved from the fight is slightly lost from non-manipulable cannon cycles.
  • Room 7: I perform a Bounce Boost and continuously SDI on the spikes to reach the end much sooner.
  • Room 8: I manipulate a Bob-Omb earlier in the fight than the last time to finish the fight more quickly.

Research Facility 2

  • Room 1: I shield-drop through the platform instead of getting on the ladder, saving some time.
  • Room 2: Because of the Bob-Omb’s much higher power in comparison to the last time, I actually remain in hitlag from the explosion two frames longer than the old TAS, which makes rooms like this (And Room 7) take a single frame longer than before.
  • Room 3: However, the higher power also allows me to outright kill the two opponents with Thunder instead of needing to stage spike them like the last time, which saves much more than two frames.
  • Room 4: I build up much more damage here, allowing me to save some time.
  • Room 5: The old TAS rides the moving platform to get to the designated Bounce Boost spot; In this one, I Bounce Boost to the Bounce Boost spot and manipulate another Bob-Omb to replace the one I first used.
  • Room 6: Due to the ceiling-floor’s thin structure, damage boosting with the right velocity and hitting the wall-ceiling corner allows me to snap up to the floor above, saving several seconds. I tried to damage boost in the same manner through to the ending door, but couldn’t get it to work.

Path to The Ancient Ruins

  • Room 1: Fox Trotting as Captain Falcon is faster than his run by a much larger amount than the other characters. I use the invincibility downtime from the Roader’s damage boost to manipulate a Bob-Omb from one of the Auroroses, allowing me to bounce boost to the end more quickly.

The Glacial Peak

  • Room 1: The higher damage I take here prevents me from properly landing at the ending door, so I’m forced to grab the ledge and do a slow ledge get-up (Jumping or falling back off still keeps my vertical momentum). It’s still faster, though.
  • Room 2: I manipulate one final Bullfrog sticker here.
  • Room 4: I launch both Lucario and myself with my Bob-Omb here, bait Lucario into air dodging my up air, and finish him off with an Up-B. I start running one frame later than the earliest frame because the earliest frame resulted in me tripping.

The Canyon

  • Room 1: The fight starts off slower than the old TAS due to RNG differences, but I eventually save time overall.

Battleship Halberd Interior

  • Room 2: I use the Sword Primid’s attack to damage boost a bit alongside using Snake’s DACUS like before.
  • Room 3: The more powerful Bob-Omb this time allows me to do the room skip more effectively.
  • Room 4: I manipulate a Soccer Ball to better take care of enemies. It’s used efficiently to kill the Sword Primid instead of using a Trophy Stand like the last time, which lets me more quickly kill the Autolance.
  • Room 6: As it turns out, it is indeed possible to skip the room like I theorized. By SDI-ing into the ceiling slightly (Which is possible to do, for some reason), I can get my final launch vector to be above the retracting barrier’s floor hitbox before it gets below the ceiling, which allows me to fly right through the seam to the ending door.
  • Room 7: Similar to Peach’s float ability, Meta Knight’s glide attack also allows him to keep the item he’s holding. I use one on the two opponents before launching all three of us, finishing them off with the same strategy I used against Lucario.

Battleship Halberd Exterior

  • Room 1: Despite my numerous attempts, my best RNG scenario in this room still required me to wait for over 500 RNG values before the third Bob-Omb spawned, so I make it advance more quickly by repeatedly edge-cancelling turnip pulls. Fortunately, the room still saves time.
  • Room 2: After performing slightly unoptimal SDI from the Quark Mine, I pulled a Bob-Omb by complete accident, which allowed me to reach the end more quickly than before. I later found that if I did the correct SDI the first time, I would have been unable to get the Bob-Omb even with manipulation.
  • Room 4: I get through the fight much more optimally than the first time, using a Stitch Face (And a recently modified bruteforcing script that checks Turnip Face IDs, which were unable to be found previously) and a Super Mushroom to blitz through to the end.
  • Room 5: I found that pulling a Bob-Omb along the second platform was faster than waiting around on the first one, which saved enough time for me to properly manipulate the special Bob-Omb I got the first time around.
This next sticker application is slightly faster than the old TAS in spite of putting on many more stickers, thanks to my error in sticker judgement during the first one that required me to discard a sticker on Meta Knight.

Battleship Halberd Bridge

  • Room 1: I get a better attack pattern than the last time, which lets me finish off the boss more quickly.

Subspace Bomb Factory 1

  • Room 1: The more powerful Bob-Omb lets me clip through the corner and fly over the fight trigger more effectively.
  • Room 2: RNG differences lead to a slightly slower fight, but I still save time in the room through a damage boost the instant the fight ends.

Subspace Bomb Factory 2

  • Room 1: I end the fight more quickly than the last time by knocking the Metal Primid’s corpse into the nearly-dead Autolance.
  • Room 2: I use the Buckot’s own attack to launch me instead of manipulating a Bob-Omb from it (Which would be slightly slower since my Up Air now barely doesn’t kill, thanks to a slightly less powerful Leg sticker), which makes me get caught by the second Buckot’s attack. I’m able to manipulate it to launch me properly, however, and save time in many other areas.
  • Room 4: I intentionally die here to partially load ROB in (So he loads in more quickly in the next room) as ROB is the next character up.
  • Room 7: Purely downwards SDI is faster than Falcon’s falling speed, so I damage boost from a Quark Mine to reach the autoscroller more quickly.
  • Room 8: Thanks to the platform’s properties, I’m able to edge-cancel Falcon’s landing lag from his forward air, which lets me defeat the boss more quickly alongside better timed attacks.

Entrance to Subspace

  • Room 1: The first fight is completed more quickly thanks to more optimized despawning; Meta Knight’s extra stickers let me defeat the Generators more quickly, and the gliding is also slightly faster.

Subspace 1

This level is completely changed from the old TAS, and now uses King Dedede out of all characters because of his Jet Hammer glitch: By mashing the control stick at TAS-level speeds while charging his Jet Hammer, Dedede can move extremely quickly. Captain Falcon is also skipped, because he isn’t needed.
  • Room 1: Dedede takes nine frames longer to load than Luigi, and loses a number of frames in this room in spite of my best efforts.
  • Room 2: He saves that time amount and more in this room, however.
  • Room 3: And I promptly save about three seconds in this room thanks to Dedede’s better vertical movement alongside the Jet Hammer Glitch.
  • Room 4: I delay entering the previous room’s door by several frames to manipulate a Bob-Omb in this one, allowing me to reach the end more quickly.
  • Room 5: A slight amount of time is saved here, most likely due to me not picking Falcon up this time.
  • Room 6: Dedede loses about a second in this room thanks to the several small platforms preventing the Jet Hammer Glitch from being properly utilized; the time saved across all these rooms is more than enough to offset it.

Subspace 2

  • Room 1: I make a very slight detour to save Peach, as she’s used for the entire Great Maze.
  • Room 2: This time, I despawn the Armight with the Pause Glitch instead of despawning it with a Trophy Stand, which saves several seconds thanks to its long spawn animation.
  • Room 3: Killing the Ticken is faster than making it fall to the bottom blastzone by a handful of frames.

The Great Maze

Peach is now used for the entirety of the Maze instead of just the ending boss fight. Her Sticker build maximizes Stitch Face damage and powers up her Bob-Ombs as much as possible, to the point where most shadow characters are killed at 0%. Bosses are defeated much more quickly than before (And in the case of Galleom, over 1,000 frames were saved), the shadow characters are defeated roughly one second faster each (Some saving more, some saving less), and many other rooms were also greatly improved thanks to Peach’s ability to repeatedly spawn Bob-Ombs. There were some rooms where Peach lost notable time in comparison to Falcon (Where damage boosting wasn’t an option) alongside a host of other ones where it goes a few frames in either direction, but these instances are incredibly minor in comparison to all the time she saves as a whole. While it’s possible that a number of the timesaves I found could also be applied to Captain Falcon as well, I think eight thousand frames is just too large of a deviation for Falcon to overcome.

Time Table

The table is separated by levels, and further separated by the level’s individual rooms (Or doors/entrances into rooms that are entered several times in the level, for cases like The Research Facility 1). The time for a particular room always starts on the first non-black or non-loading frame (Or in the case of specially noted loading instances, on the first frame of loading), and ends on the first frame of the door opening/fade to black/hoop bubble disappearing/enemy death/boss’ health reaching zero (Depending on the room). The time for an entire level starts at the finished “Stage Clear!” screen of the preceding level, and ends at the level’s own finished “Stage Clear!” screen. Timing for the two sticker applications starts on the first frame the sticker menu appears, and ends on the frame the map menu is seen again. I also differentiate between the level as a whole and the sum of the individual rooms so that time saves or time losses from uncontrollable loading differences can be seen.
LevelNew TAS TimeOld TAS TimeTime Delta
Bootup to 1-112681838-570
The Midair Stadium
1-1244272-28
1-212471289-42
1-3522538-16
Room Total20132099-86
Level Total37843874-90
Skyworld
2-1481484-3
2-2766891-125
2-315191595-76
Room Total27662970-204
Level Total38254038-213
Sea of Clouds
3-1641938-297
3-2 Character Load118195-77
3-211151627-512
3-313241535-211
Room Total31984295-1097
Level Total40525222-1170
The Jungle
4-112931318-25
4-29871137-150
4-315641546+18
4-49669660
Room Total48104967-157
Level Total58346009-175
The Plain
5-113341501-167
5-220892380-291
Room Total34233881-458
Level Total41724630-458
The Lake
6-110321078-46
6-239390
6-3238276-38
6-4294295-1
6-5282284-2
6-6295316-21
6-7126137-11
6-8631687-56
6-9177189-12
6-109294-2
6-11186187-1
6-1225272520+7
6-13239326-87
6-14203196+7
Room Total63616624-263
Level Total1017710439-262
The Ruined Zoo
7-1370137010
7-213041331-27
7-315391977-438
7-410561118-62
7-53531069-716
Room Total79539196-1243
Level Total990411161-1257
The Battlefield Fortress
8-112281654-426
8-2612596+16
8-315331561-28
8-415922035-443
8-521472808-681
8-6119131-12
8-7485503-18
Room Total77169288-1572
Level Total969111300-1609
The Forest
9-113361576-240
9-215041590-86
9-37951096-301
Room Total36354262-627
Level Total46925332-640
Research Facility 1
10-11321320
10-2490492-2
10-3144145-1
10-4160174-14
10-513181514-196
10-6153157-4
10-719952153-158
10-8207226-19
10-9201212-11
10-10263315-52
Room Total50635520-457
Level Total70397501-462
The Lake Shore
11-1222281-59
11-2307300+7
11-Yoshi Load 1144236-92
11-324152457-42
11-432853286-1
11-Yoshi Load 2155244-89
11-523102529-219
11-680800
Room Total89189413-495
Level Total1113011617-487
Path To The Ruins
12-113911681-290
12-219462846-900
12-32042040
12-4281328-47
Room Total38225059-1237
Level Total52966551-1255
The Cave
13-119191982-63
13-29496-2
13-322392300-61
Room Total42524378-126
Level Total51905321-131
The Ruins
14-116101685-75
14-233113289+22
14-3770
14-4276291-15
14-534273628-201
14-6300334-34
Room Total89319234-303
Level Total1068210998-316
The Wilds 1
Sticker Application 1338122+216
15-115001781-281
15-225452640-95
15-335350
15-410581186-128
Room Total54765764-288
Level Total70867380-294
The Ruined Hall
16-1562639-77
Room Total562639-77
Level Total15201614-94
The Wilds 2
17-112611458-197
17-216721846-174
Room Total29333304-371
Level Total37154076-361
The Swamp
18-1524740-216
18-211441307-163
18-3880
18-46696690
18-5109124-15
18-612291326-97
18-77091017-308
18-823232378-55
Room Total67157569-854
Level Total88819752-871
Research Facility 2
19-1802875-73
19-26564+1
19-3269366-97
19-4525546-21
19-5391529-138
19-6436696-260
19-74342+1
19-811161381-265
Room Total36474499-852
Level Total60506903-853
Path To The Ancient Ruins
20-1953991-38
20-217171722-5
Room Total26702713-43
Level Total36943747-53
The Glacial Peak
21-1683724-41
21-236743671+3
21-3715756-41
21-4166211-45
Room Total52385362-124
Level Total67986934-136
The Canyon
22-126472676-29
Room Total26472676-29
Level Total34523497-45
Battleship Halberd Interior
23-156560
23-2708722-14
23-3614714-100
23-4738811-73
23-5161173-8
23-62071298-1091
23-7209312-103
Room Total26934086-1393
Level Total46976109-1412
Battleship Halberd Exterior
24-1600612-12
24-2291292-1
24-3105108-3
24-410531379-326
24-5490547-57
Room Total25392938-399
Level Total36614063-402
Battleship Halberd Bridge
Sticker Application 2233237-4
25-1225255-30
Room Total458492-34
Level Total15411589-48
Subspace Bomb Factory 1
26-1487548-61
26-2614635-21
Room Total11011183-82
Level Total18541950-96
Subspace Bomb Factory 2
27-116131659-46
27-213161410-94
27-314291446-17
27-424502558-108
27-ROB Load6781-14
27-5824844-20
27-613111351-40
27-76566-1
27-8587694-107
Room Total966210109-447
Level Total1210212531-429
Entrance To Subspace
28-143144533-219
Room Total43144533-219
Level Total48165047-231
Subspace 1
29-1294282+12
29-2220279-59
29-3383553-170
29-4349418-69
29-5168172-4
29-6413339+74
29-71591590
Room Total19862202-216
Level Total35373763-226
Subspace 2
30-1407431-24
30-225252650-125
30-312321310-78
Room Total41644391-227
Level Total51045334-230
The Great Maze
Room Total2397332075-8102
Level Total5419362214-8021
Final Total (Sum of Level Totals)229439252336-22897

Improvements & Final Thoughts

Unlike the last TAS, there’s very little left in the way of guaranteed timesaves. The only absolute thing left I’m actually certain of is in Room 6 of Battleship Halberd Interior, where I could bounce boost to the Bombed and RNG manipulate another Bob-Omb like I did in the old TAS (Which I somehow only realized while TASing six entire levels after this one). By my estimate, it would save roughly 40 frames at best, assuming the extra damage doesn’t interfere with the launch. There’s also the ever-present horde fight differences spawned from single-frame item RNG deviations, alongside getting RNG that doesn’t require me to pause buffer for stickers and items (Which saves two frames per pause). I also apparently made some slight errors in marking the sticker route this time; there should be a tiny bit of timesave in that regard, but nothing like the timesave present between the old TAS and this one.
There’s a glitch that allows you to use unintended characters in levels that don’t have you choose characters from a cutscene (The Plains, The Ruins, Bomb Factory 2, etc), and it was theorized that Yoshi might be faster than Lucas in The Ruins. However, I TASed through that level and all its RNG manipulation while completely forgetting to test that glitch; I suspect that the timesave present isn’t actually that much since The Ruins is almost entirely an autoscroller.
Aside from that, I think I’ve covered all bases with this one and truly pushed Subspace to its limits. It beats the old TAS by over 6.3 minutes (Which now makes it faster than the RTA counterpart by over 100 minutes), which was more than I ever considered when I first started the revised attempt.
Special thanks to all the people who stopped by to say hi during my TAS livestreams, and to those whose suggestions actually helped the TAS move along.

Random Fun Facts

  • I take a grand total of 13,321 visible damage in a game that stops keeping track at 999%. The actual number is much higher thanks to The Great Maze.
  • I pull a grand total of 47 Bob-Ombs from Peach’s Down-B throughout the run, which means that repeating this TAS by chance just from the Bob-Omb pulls alone (Ignoring the copious amounts of other item RNG manipulation and turnip pulls that made the inevitable Bob-Omb pull come more quickly) would have a naive likelihood of 1/240^47 = 1/7.41e111. Adding in the eleven Stitch Faces now makes the naive likelihood equal to (1/240^47)*(1/56^11) = 1/1.26e131
  • The bruteforcing script I used went through a grand total of 110,676 items across the run, which makes the manual rerecord count equal to 165,427. The old TAS also had bruteforcing rerecords used heavily, but I didn’t keep track of them properly and couldn’t assign a number.

Screenshot Recommendations

  • Any potential frames of Peach plucking a Bob-Omb from her Down-B
  • ~141,553 (Where Meta Knight clips through the series of blockades)
  • ~151,416 (Where Captain Falcon is throwing the Cracker Launcher at a horde of enemies as the firecrackers themselves explode)
  • ~208,923 (Where Mario gets hit from another Bob-Omb Peach throws at him)

Memory: SAMSARA ISN'T STEALING THIS FROM ME THIS TIME. Claiming for judging.
Samsara: LOOK,
Memory: Thanks to fsvgm777 for confirming sync.
So Brawl is one of those games that I played a ton as a kid. I liked Melee but Brawl just had so much more content. I try to keep nostalgia to a minimum when I watch TASes but seeing the subspace tases brought all of it back for me. Even with all of Brawl's flaws, it will forever be my favorite smash game.
I loved the original TAS and this TAS is even better. A really great improvement is one where you can visibly notice the author having a stronger mastery of the game and that is definitely the case here. The movement as a whole is much cleaner. Even tho damage boosting is used for more of the TAS, new movement techniques shine through as well. A huge shoutout to the jet hammer glitch with dedede, that made me laugh. The TAS was received extremely well by the audience as well.
Zinfidel: Processing...
Last Edited by adelikat on 11/6/2023 3:34 PM
Page History Latest diff List referrers