Name: [TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s @PurpleSunTAS
Uploaded: 2026-09-18T08:17:00-05:00
Duration: 8 min 17 s
Description: [TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s

[TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s

PurpleSun 2026-09-18 | [TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s

[TAS] SMG2: Flash Black Green Stars

PurpleSun 2024-06-22 | Green Stars TAS WIP
Star order would matter due to Paragoomba and Jack O'Goomba cycles, however use of 2P negates the differences. GS1 was still picked first from the pre-2P route.

Green Star 1 - 16.78s IGT - 2:01:53.73 real-time split (1:13.05 segment)
This is an improvement of 2.97s over my 2020 IL TAS, despite cutscenes: youtu.be/WBvi2DMZO9M

Even against the Blimp boost tech explained later, climbing Yoshiless is faster by a couple of seconds.
A megasideflip is a glitch that still isn't quite understood on a technical level, but requires a certain camera type (XZ_PARA) and angle, and then sideflipping and holding a direction perpendicular to facing angle. See more here: docs.google.com/document/d/1a2wWBhFL7hTVEY9zJmtTk5CEbxiib3lim9hSOCkAWQw/edit#heading=h.54goeomx2quh
This megasideflip route is 1s faster than this zip jump triple jump route: twitter.com/PurpleSunSR/status/1801718542540411149

The megasideflip was made even faster by aiming to grab the ledge. Since this is inside the 2D section, and entering a 2D section from the front slowly forces Luigi into the 2D restriction, grabbing the ledge results in a 1-frame climb.

Grabbing Yoshi at the end would only be 0.5s slower than what was done here.
Usernameispassword suggested starbitting the Paragoomba to bounce off it in mid-air. This strat saves 0.37s over the mega-wallkick route done in the 2020 TAS.

Green Star 2 - 31.15s IGT - 2:03:07.98 real-time split (1:14.25 segment)
This is an improvement of 5.18s over my 2020 IL TAS: youtu.be/XJqTklGuRrg

Thanks to the Paragoomba cycles, it is 3 frames faster to bounce-spin off one to the wallkick up to the Jack O'Goomba, rather than the double wallkick GS1 has to settle with (2P can't help GS1's cycle because it needs the Paragoomba to move to the left).

Taking Yoshi at the end of the climb ends up 0.8s faster than going the rest of the way Yoshiless (which would use a zip jump).
The Blimp boost tech involves freezing a Paragoomba so Blimp Yoshi can go through it, and then spinning it to make it solid inside Blimp, causing it to push Blimp out at high speeds (like a proxy in Spyro games). This one was 121 initial y-speed - Blimp's normal is 7 y-speed.

UPDATE: Improved ending by 0.39s here: youtu.be/emYeiQOWsKY

At the end, backwards infinite flutter from the start is 0.5s faster than going over to the crystal for a zip jump.
I slightly angle the flutter to drift off of a straight line to the star in order to be over a void around the middle part.
After accelerating to top speed with fluttering, you eventually lose speed if you're fluttering close to the ground, but maintain it over void or high above ground. This saved a few frames.
I also turned the camera here - moving to the side of the screen is slightly faster than moving towards or away from it. This saved 0.2s. See more here: docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit

at 150k rerecords now

#supermariogalaxy2 #toolassistedspeedrun

Fluffy Bluff GS3 zip jump

PurpleSun 2024-06-12 | This is tool-assisted.
There's some weird part of this hill that gives Luigi extra speed when running on it. This means it's possible to get zip jump speeds of up to 101hspd, leading to getting all the way across to the final bluffs.
This saves 9 or 10 seconds over my GS3 TAS from 2022: youtu.be/NAdhHZ54ZN0

[TAS] SMG2: Melty Monster Green Stars

PurpleSun 2024-04-27 | Green Stars TAS WIP
Star order used to matter when GS2 used the lava waves, but no longer does.

Green Star 1 - 13.83 IGT - 1:57:44.80 real-time split (1:11.55 segment)

Land cancel doesn't seem to be possible here.

When grabbing pull stars, there's 18 frames from pressing A to the beam grabbing Luigi's bubble - this is the same no matter the distance.
With this in mind, it's generally faster to grab as few pull stars as possible to keep as much pulling time, and also pull from further away to get more speed (which is why I sometimes wait it out until a good one appears).
For the end, I used the IL strat of quickly exiting a pull star (done here by Emperor: youtu.be/8Gq0zWl6Hl0?feature=shared)

Green Star 2 - 33.45 IGT - 1:59:01.58 real-time split (1:16.78 segment)
This is an improvement of 3.01s over my TAS from 2021: youtu.be/ig1Qc9T3mCE?feature=shared

About 0.6s of improvement came from a better pull star section.
2P grabbed the Life Mushroom ready for the lava wave wallboost strat (found by nemi43; see my previous TAS), but as it turned out, that's no longer fastest.

Instead, it is possible to get zip jumps (glitchy slope jumps) off two opaque crystals to soar across the waves and save 2.2s over the lava wave strat.
It is possible to get a single jump with conserved speed when landing from the first zip jump that can get on the second crystal, however, the angle makes the second zip jump worse so it's not faster (from what I've tested).

The final improvements come from a better ending, with a better backflip spin into a higher lava boost gets a higher sandstorm spin, which hits the star around the top half of its collision which is slightly faster than the bottom.

Green Star 3 - 43.90 IGT - 2:00:40.68 real-time split (1:39.10 segment)
This is an improvement of 2.56s over my TAS from 2020, despite cutscenes: youtu.be/MhqGMs8oWSw

The slightly wayward start is because a straight path would have had the first LJ land on the slope, where tilt would make the second LJ higher and would land later for the single jump, losing a couple of frames to what was done here.

The main improvement is another big new skip: the launch star skip, using rock mid-air jumps which you can learn more about here: docs.google.com/document/d/13HenHN3IA2mQih2k8Ve_VZdK-yIoX8fnXB2g2eMaZGM/edit
This skip is tighter than it looks, as there is a 10-second timer on how long you can stay in the air before the game instakills you. This landed 5 frames before that point.
The big downside to the big skip is that it also skips the trigger that blocks you from stopping rolling. This is a problem because it also allows for the automatic stop that happens after rolling for some time. This can be worked around by jumping on each section.
Overall it saves 4 seconds.

Crystals are avoided as hitting them causes lag frames.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Supermassive GS3 in 17.43s

PurpleSun 2024-04-12 | A new improvement of 1.03s (62 frames) over my Greens WIP from last year: youtu.be/Z5AFvSz7VAs?t=74

One to add to known improvements I guess... not sure why I didn't think of trying the Goomba float before. Maybe because I was locked in on the coin zip jump, and didn't think that floating from this far back would work.
But the question of it being faster was brought up in Jhay's Discord server after the A press save strat was shared. So I've tested it now.

This isn't hyper-optimised as well so could be a few more frames to save.

SMG2: Supermassive 0.5xA float to launch star

PurpleSun 2024-04-11 | plus 2P 0.5xA (it lets go of the Luma by spinning)
The knowledge of floating only cutting off when the object is despawned (goes too far off-screen), and that Supermassive generally has a free camera, combined to be able to float off the big Goomba further away from the checkpoint than previously done, saving the A press of backflipping onto it by instead bouncing over from the big Luma.

This overall saves 4xA as it can be used for all three yellow stars & Green Star 3. The only A press left in this galaxy is the star grab for 1 & C (and also ghost textboxes because Luigi).

[ABC TAS] SMG2: Melty Monster 1 in 2xA Presses (Mario)

PurpleSun 2024-04-09 | Mario crosses the Magnificent Magma Sea in 2 A presses.
The required A presses are:
#1: Jumping off the swing
#2: Jumping off the first planetoid of the last section

The key improvement of this video is matching Mario with Luigi; due to Mario not being as fast, it wasn't known if he could do the swinging Luigi does in the end, so he would have needed more health for more fireboosts (and so likely getting coins which costs 1 A press after the level) or more jumps. But this proves he can be just as swingy as Luigi.

Pull stars can be grabbed with B as long as you aren't holding A. I try to show this is the case by having a starbit at the first pull star, so it shoots at the same time I grab.

It was necessary to get the Life Mushroom early on to have enough health to set everything up for the second Life Mush, but, as it turned out, it wasn't necessary to stop and do a second lava wall boost to save 1 more bit of health.
In fact, this run could be much shorter by simply taking the first lava wall boost, detouring to the second Mush, and then going all the way out to the tornadoes with 3 health remaining for the end.

#toolassistedspeedrun #abuttonchallenge #supermariogalaxy2

FASTRL: FighBats The Rig clip for ultra shortcut

PurpleSun 2024-04-06 | When running this game a couple years ago, FighBat ( youtube.com/channel/UCct-bwDAlsaysGzxIiVHAkA ) stumbled on a weird clip in The Rig, which was previously a track with no skip.
A combination of speed and angle going up the little step allows you to clip inside the wall. It's possible to go all the way through without bonking, but if you get a good bonk, you can get lifted up, and with a really good bonk can get up enough to fly over the checkpoint in front of the finish line and turn around to get a lap skip.
This wasn't working at first as I was pushing to get around the finish line and land behind it, but it turns out you need to land into its checkpoint area if you're trying to skip from so close to it.

That's now 4 tracks with a massive ultra, joining Breeze City, Dry Hill and Pyramid Valley. Rookie Beach could probably also have this clip to skip. Flood Area also has a lap skip after completing 1 lap, and Turbine Canyon, Scorpio Circuit, The Chillout, and Storm Creek all have big shortcuts cutting out at least half the track. So the only tracks left without something big are Cove Harbour (the very first track) and Ice Park.

This is tool-assisted.

[TAS] FAST Racing League: Breeze City in 14.56s

PurpleSun 2024-04-05 | A few years ago, I decided to stop working on a TAS of FAST Racing League, after finishing the first 4 of 12 tracks: youtu.be/ItOBCvu-u0Y
But I really wanted to at least do this track as well.

===BONK JUMPPAD===
One of the reasons I stopped the TAS was that I stumbled upon a trick in Turbine Canyon to make the jumppad shorter, which would save quite a bit of time there, but I couldn't replicate it and so skipped over it but later didn't want a TAS that couldn't have that.
Here I found that the same sort of mechanics can work the opposite effect too, getting more height which was helpful to be able to turn in the air without landing too short of the checkpoint, and gained a bit more understanding.
It seems like you either need a certain angle against the wall while jumping, or need to jump a certain time after bonking, or both.
Angle seems more likely, because slightly adjusting control stick angle would change the height a little bit. But it might need the effect of a bonk too.

===LAP SKIP===
This game is a bit like Mario Kart Wii where there are checkpoint areas, mostly used for respawn points. But unlike MKW, you aren't limited to the one before and after.
The checkpoint area for the line actually ends some way in front of it (the bottom of the closest yellow diagonal, so 2-3 cars long), so if you're able to jump and land behind this invisible line, you're counted as also being behind the finish line from a further checkpoint, so you can finish a lap.

Some exceptions though: if you land before the actual line and cross back over it, you can't be facing forwards, even if you're moving backwards. This is why I landed sideways every time, facing ever so slightly backwards but turned around enough to start moving forward immediately after crossing the line.
You also can't be braking or drifting on the ground during this skip for some reason, which is why I stop drift turning as soon as I land and just go neutral.

===BOOST===
Theoretically the boost would be better to use going towards the jumppad for a skip, as it only saves a few frames right at the end. But I haven't managed to get the higher jumppad trick while boosting into the wall. Kinda harder to know your angle or time the bonk right - whichever one actually matters.

#toolassistedspeedrun

[TAS] Bonsai Barber Perfect Day in 1:53.36

PurpleSun 2024-04-01 | Bonsai Barber is a WiiWare barber-sim made by Zoonami, a team led by the director of Goldeneye N64. Inspired by the Japanese art of bonsai and the barbershop experience, it involves giving charming anthropomorphic plants a fresh trim, with styles ranging from simple shapes to bizarre structures.

The game has a real-time mechanic where you can only cut up to 5 plants per real-world day.

This TAS aims to make an entertaining and fast run to complete the Perfect Day achievement on the first day. Due to bottlenecks in the game's reward system, absolute speed isn't of the essence, so there's enough time to mess around with painting, birds, and apples.

10060 rerecords

april fools, lads, the joke is that i spent time on this instead of finishing galaxy

[TAS] SMG: Good Egg 2 in 44.38s (Mario)

PurpleSun 2024-03-12 | Mario lets a Hungry Luma starve in 44.38 seconds (2663 frames).

The triple jump bounce at the start was found by nemi43: youtu.be/wM5Zpot5h7g?feature=shared
While it might be possible to float directly from the bounce, the angle is all wrong for the float needed to get the skip.

The skip had also been theorised by nemi, and, with some help from Dendonodon and BillyWAR, I was able to get it years after the theory: youtu.be/sk68AKjWnyg
But the specific angle and movement used to get the fastest possible skip (i.e. moving far enough to ground pound as early as possible into the final planet's gravity) was figured out by the RTA community (sorry I can't nail down the names; Marioandtrafficlightfan, Subro, Cubic3107, saburoumaru and Nanase all have videos on the same day with the same technique).

#supermariogalaxy #toolassistedspeedrun

[TAS] SMG2: Flip-Swap GS1 in 5.96s

PurpleSun 2024-03-08 | Luigi hyperjumps eastward into a Green star in 5.96 seconds [IGT, 358 frames].

-2 over my 2022 TAS: youtu.be/RikJY-wwPDk

@Joselle_Astrid really had everything figured out long ago, and her research pops back up every now and then, and something within it is then realised for optimisation.
This time it's that aerial movement is ever so slightly faster when going east or west (side-facing the camera) compared to north/south (away from or towards the camera respectively). docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit

This helps both in getting a slightly further long jump and single jump so they can be done earlier, and in a faster flip panel long jump, overall saving 6 frames (note: a smaller bit of improvement may have also been from a better understanding of gaining height from flip panel jumps).

As the camera is so free in this level, it's much easier to get the potential out of the east/west gain, but a lot of levels would have trouble due to either limited movement or forcing itself back to the over-the-shoulder perspective - we don't have SM64's luxury of knowing if we're keeping straight while the camera moves, so it's harder to get the benefits.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Boom Bunker Green Stars

PurpleSun 2024-03-07 | Green Stars TAS WIP
Star order has no real effect - the mine cycles on the 2nd cannon are a little different but not enough to get in the way.

Green Star 2 - 38.08 IGT - 1:55:31.28 real-time split (1:34.42 segment)
This is an improvement of 0.72s over my 2020 TAS: youtu.be/WtCdwYBsCnA

2P freezing the Bullet Bill minimises its big turning circle, as it still turns while frozen, saving 32 frames. The weaving around the coins didn't lose any frames.
I tried to be smart by 2P freezing and spinning the Fizzlet - when hitlag is stacked on the same frame, one of the lags is cancelled. I had hoped the 3-frame spin hitlag would cancel out the 5-frame explosion lag, but it ended up the other way around.

The 2P Boomerang Bro strat saves ~2 seconds. It requires turning the camera to keep 2P on screen - 2P lets go of its target if it goes off-screen.
But having the camera stay side-on for the launch star loses 4 frames as the camera has to move around more in the cutscene, so I move it back in time.

Backflip spin saves 32 frames over running to the cannon.

Green Star 1 - 18.88 IGT - 1:56:33.25 real-time split (1:01.97 segment)
This is an improvement of 1 frame over my 2020 TAS (unreleased).

Buffered crouch acceleration at the start gave enough extra distance to save a frame on spawning the Bullet Bill.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Shiverburn Green Stars

PurpleSun 2024-03-02 | Green Stars TAS WIP
Star order has no real effect - all three stars have the same start so start position difference doesn't matter.

Green Star 3 - 35.45 IGT - 1:51:51.80 real-time split (1:31.94 segment)
This is an improvement of 1.53s over my 2020 TAS, even without land cancel: youtu.be/L1khyJTw79E

No land cancel here.
This route to the Cloud Flower saved 6 frames over starting with the 2nd green platform, and getting a height boost off the blue platform.
The most interesting part was the route from getting clouds and skipping to the launch star, and how many clouds would be best to have at the end.

The route RTA uses - going back to spawn and then long jumping out of bounds from there - is possible with just 1 cloud, leaving 2 at the end.
This route - which is later used for GS2 - is barely possible with 2 clouds (even when entering gravity, I have to slow down for a few frames to avoid swinging out into the deathplane). This route is 92 frames (1.53s) faster than the other.

But on the last planet, having 2 clouds would allow for a double cloud long jump route that saves 0.65s over the standard 1-cloud strat.
So overall, the final route - 2 clouds at the start, 1 at the end - was 0.88s faster.

On the last planet, I saved 3 frames by turning the camera so Luigi was moving to the left in its perspective.
For some reason, you get ever so slightly more speed moving left/right to the camera compared to moving away from it, and much more compared to moving towards it. docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit
This mechanic is difficult to make useful in other parts of the game, as often times the camera either can't move or it's constantly fighting to get back to being behind Luigi, with its back-and-forth swinging making optimising movement an exhausting task without some kind of script or visualiser that we don't yet have.

Green Star 2 - 19.48 IGT - 1:52:55.11 real-time split (1:03.31 segment)
This is an improvement of 0.5s over my 2020 TAS: youtu.be/QdUV1xkWiuo

Not much to note besides the trick itself catching the edge of gravity right above the star into a cool precise spin-cancel ground pound.

Green Star 1 - 17.91 IGT - 1:53:56.86 real-time split (1:01.75 segment)
This is an improvement of 0.58s over my 2020 TAS: youtu.be/thg0ekOZxfg

On the 2nd cloud long jump to the button, I start walking up and then turning and long jumping before completing the turn. This makes it so Luigi turns around quicker when getting up from the ground pound, saving a few frames.
Despite the wayward zig-zag, the movement on the frozen volcano saved 2 frames over going straight over the higher lip to the left.

#supermariogalaxy2 #toolassistedspeedrun

[ABC TAS] SMG1: Melty Molten 2 in 0.5x A presses (Luigi)

PurpleSun 2024-02-03 | Luigi gets through the meteor storm in 0.5 A presses.
The half A press only counts for the level itself, essentially noting that the A press for selecting the star should be held throughout the whole level.
It's required for a higher spin to get the star.

On the Thwomp planet, it's possible to get clipped into the floor. By getting under the edge of a Thwomp as it comes down, Luigi will sidewarp (pushed with a lot of speed in 1 frame) instead of getting crushed.
This sidewarp clips thanks to the cornered floor; running from the top side of the corner means Luigi's tilt doesn't update to the diagonal floor the Thwomp crushes, so instead of going up-right, the sidewarp goes straight right and through the corner.
This was found by Annoying (A Good Duck) who discovered a consistent clip if you die and then hold left on the respawn: youtube.com/watch?v=-nGMh75Dhrg
It's necessary to get the clip facing left in order to get the faster backwards acceleration to get past the big block.
Afterwards, a simple backwards spin knockback got up the small block.

The Life Mushroom isn't necessary - I got that to try a star grab strat using lava boosts.
During the chip section, I did edge grinding to save frames, running along the edge to start falling off but being able to stay on, using precise angles to stay in the 'fall off' (+ ratio) state - this has a slightly faster speed than running.
The really slow grinding... that's definitely RNG manipulation. Trust me, bros.

At the end of the Bullet Bill planet, I entered the launch star while falling off a ledge to get displacement, leading to sliding along the floor and bonking on the next planet which saves some time.

While I avoided coins for most of the ball section, I couldn't find a way to get to the final platform while avoiding the coin ring - it was either too steep an angle or too little speed turning to roll onto the lava platform.
The star grab should not have worked. The weakspin (little twirls after spinning) was frame-perfect. Hitboxes really are crazy. (Sorry for no real technical explanation besides weakspins having a little extended hitbox and star hitboxes being so random).

ty Xander for encode

#supermariogalaxy #abuttonchallenge

[TAS] SMG: Gusty Garden 1 in 1:12.43 (Luigi)

PurpleSun 2023-11-23 | Luigi zooms off flowers to a star in 1:12.43 (real-time, from star select to star get cutscene).

Could be some more improvements but I'm happy enough with what I did to call it a proper TAS.
The floaty fluff boost was found by Tackleshift: twitter.com/tackleshift8/status/1651605488528953344
It requires you to wait a little for the fluff to rise before riding, and then 2P jumping into it. The boost seems to be timing-dependent as waiting too long won't get it.

The first boost was much more vertical than expected, and might actually be able to skip a planet but I didn't try it out enough, and it loses a lot of speed quickly whereas keeping the regular path has a lot of speed overall.
I let go early to conserve as much speed as possible in backwards falling, and then used backwards momentum again on the next planet to swing around and bounce on the Piranha plant.

Momentum was used again to bounce off the Piranhas on the next planet.

The last boost led to two possibilities - 1. get back to the windstream and go all the way to the half-sand planet or 2. hop off early to keep speed over to the Piranhas. The latter turned out to be 40f faster, despite the former being able to immediately set up a fast triple jump from hopping off the fluff and landing in neutral.

#supermariogalaxy #toolassistedspeedrun

SMG: Bowsers Dark Matter Plant (LOTAD)

PurpleSun 2023-11-23 | Just throwing some ideas around.
Start was kinda based off some movement from an ancient DMP collab freerun between Xander and Nim many eons ago. There's probably something even better in those old runs (I have just seen a clip of Nim's unfinished TAS and yes).

Triple jump up to clip spot was found by Tibs I think?

Skip into 2nd gravity section was found by nemi43: youtu.be/51a5mExWWg8?feature=shared

Not sure who discovered the 2P version of the final skip, but here is a slower 1P version from many years ago: youtu.be/J6wmJIeUoj0?feature=shared

#supermariogalaxy #toolassistedspeedrun

[ABC TAS] SMG2: Flip-Swap GS1 in 0xA

PurpleSun 2023-10-06 | Mario floats to a Green Star in 0 A presses.

This saves 0.5xA overall (1xA Mario), improving on a Luigi-only strat involving a triple bounce off some enemies: youtu.be/l9m2kxQx2MI?feature=shared
It also saves 1xA in menuing in a full-game run, as this strat clears up the last Luigi advantage for Flip-Swap, meaning we can avoid the New Ghost Appeared textbox that you get from using Luigi here.

With the knowledge from BillyWAR about crystal floats ending when the crystal despawns (i.e. when it goes off-screen), combined with a freely movable camera I always forget about in Flip-Swap, it's possible to float from the start straight to the Green Star.
Shoutouts to the bunny.

#supermariogalaxy2 #abuttonchallenge

[ABC TAS] SMG1: Good Egg 2 in 0x A presses [Luigi]

PurpleSun 2023-10-03 | Luigi floats away to let a Hungry Luma starve in 0 A presses.
This is tool-assisted.

This is a gameplay improvement of 0.5xA presses, not needing to hold A for higher spins or bounces at any point.
The main improvement is cutting out 6 A presses of dialogue (menuing) from the Hungry Luma.

After getting the new GE2 skip with just a sideflip setup for the crystal float (youtu.be/90QX5QEvo-s), I decided to try with just a Goomba bounce. As Mario, this is much harder and hasn't been done yet, but Luigi is much easier to get floating, and in this case, the flipping animation that usually happens on a crystal decided not to for reasons.

UPDATE (just a few hours later): Mario can do it at least 0.5xA youtu.be/XBWQlODVYKg

The star grab hadn't been proven without holding A (0.5) but was considered most likely possible and here it is.

#supermariogalaxy #abuttonchallenge

Good Egg 2 Skip as Mario

PurpleSun 2023-09-26 | it's a-me
original with luigi : youtu.be/sk68AKjWnyg?feature=shared

[TAS] SMG2: Fleet Glide Green Stars

PurpleSun 2023-09-24 | Green Stars TAS WIP
Star order has no real effect - the Thwomps may be a couple of frames different but nothing else.

Green Star 2 - 54.06 IGT - 1:49:23.00 real-time split (1:49.85 segment)
Compared to my 2020 TAS (youtu.be/DA8-_G1KBlo), the flying section was improved by 2 frames (42.23 on timer vs 42.26).

I struggled to keep up with the 2020 TAS without copying inputs, and even adjustments to try to improve only got to being 0.2 units ahead by the tunnel entrance (Fluzzard Hspeed is generally 47 so that's a very tiny difference).
The main issue is there is a lot of downhill - the holes at the start, the tunnel, especially the Banzai Bill part, so alternating between boost and lift to get more distance like in Wild Glide is not as beneficial here as it also ends up higher, so you lose that extra distance with sharper downtilts later.

The framesaves came right at the end of the tunnel, where the bottleneck is no longer downhill but an uptilt right at the end.
A light uptilt followed by 1-frame sharp tilt under the ceiling saved a frame on that last stretch of tunnel compared to the gentle curve of tilt from before, and optimising that + general better movement to the star saved another.

Green Star 1 - 14.18 IGT - 1:50:19.86 real-time split (56.86 segment)
IL TAS is 13.98 (youtu.be/vDHeMTBJWAo).

I would usually delay the cutscene skip to optimise in-game time without losing real time, however, there's a quirk with the Jibberjay's dialogue where - after skipping cutscenes - it has some cooldown or inactive state before the dialogue box opens, so you could press A 0.3s later than possible and still be as fast as pressing A as early as possible.
Skipping cutscenes earlier than I usually would manages to cut down a bit (but not all) of the waiting time, so it's faster RTA by 13 frames.

Because this quirk gives freedom for suboptimal movement, it means there is space for another optimisation quirk - camera cutscene movement. For dialogue or other mini-cutscenes, the camera needs to reach a certain point to rest before the box appears or the thing happens, so frames can be saved by getting the camera closer to its point before activating the scene/dialogue.
If heading straight to the Jibberjay, the camera really slowly moves to its point, so instead I run and slide around the right, saving 7 frames on camera movement (which seems to not be limited by the cooldown quirk I guess?).

Also 7 frames difference was bigger than the IL of 6 frames difference. Maybe 1 more frame to save in IL!

As nemi43 found, tilting up and slightly to the side grabs the star 1f earlier because hitboxes.

#supermariogalaxy2 #toolassistedspeedrun

Good Egg 2 skip

PurpleSun 2023-09-04 | Originally theorised by nemi43: youtu.be/aDkOvUhw6I8
This is tool-assisted.

I had tried this crystal float a while ago but could not even control Luigi enough to get out of the gravity. See here: youtu.be/1C4g-3myaJA

Dendonodon (@dendonodon2858) picked it up and pretty quickly worked his way out, but found the direction I had been going in wasn't going to work. I wouldn't have tried this again and wouldn't have got it without his findings.

So here it is, going in a new direction and barely keeping the crystal within the 'spawn' boundaries of the camera (it kinda goes offscreen for a bit but not too far) - floats get cut off if the object you float off gets despawned (discovered by @billywar1507 when doing Flip-Out A Button Challenge).
They can also die off quickly if you move forward too early, but moving forward later on helps to slow Luigi down and keep the crystal on screen.

Float setups usually involve spinning onto the crystal because it makes it much easier, but for this skip, I needed to be able to spin once changing gravity. The workaround was to be able to land directly on the crystal and jump, as this meant there didn't have to be as much movement to get a float.

Wonder isn't the only trippy game, you know.

#supermariogalaxy

[TAS] SMG2: Slipsand Green Stars

PurpleSun 2023-08-12 | Green Stars TAS WIP
Star order matters due to the cycles of the moving blocks on the 2nd planet, making an affect on GS2 & 3 but not 1.
I've potentially lost some time here; my 2020 tests found the blocks to be slightly better with cutscenes so GS3 goes first, but during this TAS I improved that section to then mean the no-cutscene cycles ended up much better than before, but I didn't want to redo everything up to that point so I've left it as future improvement.

Green Star 3 - 1:04.65 IGT - 1:44:39.70 real-time split (2:00.75 segment)
An improvement of 4.31s (+ cutscenes) over my 2020 TAS: youtu.be/xsd_7sIqS90

It's actually possible to long jump all the way over the first stone ring, but the route that follows is 0.2s slower.
Like with star order, there's a little bit of future improvement in the last segment of sandcurrents that was found in the latter stars.

And another case in the 2nd planet, where later stars get a very precise edge long jump after going over the big curve to avoid losing speed from the sand. I really did not have patience with this galaxy.

The Romp launch works because the ends of the Romp don't immediately crush you but instead try to push you to the side (see nemi's IL for Romp pushing strats: youtu.be/Wy2RYwIa-Ck?t=185 ).
This means you can jump in front of the Romp and clip inside enough to cause a big push in 1 frame, and if you can get a slope jump during that, the clip speed is conserved and fixed to the whole jump.
I had to go to the second row of Romps because there was no good angle jumping from the first row that would go straight to the star.

Green Star 2 - 49.35 IGT - 1:46:12.38 real-time split (1:32.68 segment)
2020 tests had a 51.20 here.

The improved moving block cycle is obvious here; in the previous tests, the blocks started moving left before long jumping.

Getting the star, a straight beeline to the star was 1f slower than a little bit of extra drifting to the left.
Also, the starbit collection with IR was necessary. If you collect a starbit and you're still in "collecting" state (glowing white) when grabbing a star, that starbit will be duplicated. Picking it up on the way, I would grab the star just barely within this window, so instead I use IR to pick it up earlier and shoot it out, avoiding the starbit textbox after the level.

Green Star 1 - 37.43 IGT - 1:47:33.15 real-time split (1:20.77 segment)
2020 tests had a 38.48 here.

Much the same as GS2. For the star grab, I was able to get another edge long jump avoiding the sandy speed loss, saving a couple of frames by getting enough speed to grab the star on top of its hitbox rather than the side.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Upside Dizzy Green Stars

PurpleSun 2023-07-29 | Green Stars TAS WIP
Star order matters due to the cycles of the moving platforms and the starting Paragoombas. GS1 first is at least 1 second faster.

Green Star 1 - 13.18 IGT - 1:41:39.88 real-time split (1:08.98 segment)
See here for a 12.96 IL TAS with better cycles: youtu.be/rw64kFfrQNg

Bonking and going through a gravity flip causes Luigi to twirl around, which leads to a backwards wallkick.
The height of the bonk affects the facing angle by the time you need to wallkick, which slightly affects speed.

A starbit shot was used to bring the Paragoomba further left and bounce off it earlier. This saved 0.25s (15 frames) over 2P freezing despite the 3f hitlag.
Ground pounding at the end saved 0.32s (19f).

Green Star 2 - 16.38 IGT - 1:42:38.95 (59.07 segment)

Neutral wallkick strain (neutral on joystick after jumping to bypass the speed cap) and a slightly different bonk to change wallkick angle helped to bounce off the higher Paragoomba.

I initially had trouble with backwards falling, as it doesn't regain momentum after getting slowed down by gravity flipping.
Also, gravity flip vertical momentum is affected by holding A as well as how you move into the changing gravity (e.g. from the long red chute, the long jump bounce in GS1 that goes through the blue gravity flips higher than double jumping from red platform despite the latter falling for longer).

Mizu first suggested spinning once at the top. Spinning in the red gravity just before the very top kept enough speed and height through the final gravity flip to make it to the final platform.
The setup for backwards falling is scuffed, involving a frame-perfect jump from getting thrown off a ledge-grab, and getting stopped by leaving the red moving gravity through the side of it to then spin up.
But overall it saves 1.2 seconds.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Boo Moon Green Stars

PurpleSun 2023-07-16 | Green Stars TAS WIP
Star order doesn't matter here.

Green Star 3 - 1:02.88 IGT - 1:37:16.28 real-time split (2:02.45 segment)

As the first planet is a sphere, long jumping generally isn't faster. The only exceptions are at the start to quickly get to top speed, and through the doors - jumping squeezes through earlier to save 8 frames.

2P freezing the Boo to get a squeezed backflip saved 11 frames.
I waited until landing to spin, which initially loses a couple of frames, but turning on floor while entering the launch star changes Luigi's angle when landing on the next planet, allowing for a quick backwards land cancel and spin to the snake block, saving 14 frames overall.

Snake blocks seemingly don't count as moving platforms, so there's no cool boost to gain.
Also funny reverse gravity float off the sign.

As the moon is a moving platform, it can cause jumps to accelerate after their peak. I experimented with a mega triple jump conserving all the extra speed to get the star chips, but it was 45 frames slower than the RTA double backflip strat. UPDATE: A new approach to the MTJ using an instant turnaround for the double jump was closer but still slower by 4f.

Grabbing the star further left saved a frame due to hitboxes being confusing.

Green Star 2 - 45.00 IGT - 1:38:44.61 real-time split (1:28.33 segment)

Avoiding the coin isn't necessary due to maxed file, but it didn't lose any frames and is a good showcase for 242.

Squeezed backflip wasn't faster here due to the Boo's bad positioning.

Double long jump + floaty airspin cancel was 8 frames faster than a long jump + single jump. It's also possible (27f slower) to ground pound the moon and uncrouch spin to grab the star without falling off: twitter.com/PurpleSunSR/status/1680632811764568065

Green Star 1 - 1:02.95 IGT - 1:40:30.90 real-time split (1:46.29 segment)

This time it's important to get on the third snake block as soon as possible, as that's the next autoscroller bottleneck.

For the star, a couple of skip ideas were tested:
#1. Long jump, 2P freeze an Octoboo and 'bounce' off it - off by about 100 units (5-ish LJ frames).
#2. 2P squeeze backflip from an Octoboo - 5f slower replacing the LJ at the end

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Space Storm Green Stars

PurpleSun 2023-06-15 | Green Stars TAS WIP
Star order matters due to bubble cycles for GS3, where cutscene cycles are better by 9 frames, but also pull star cycles means GS2 would save 3 frames. So overall doing GS3 first is faster by 6 frames. Quick maths.

Green Star 1 - 5.88 IGT - 1:33:04.05 real-time split (1:03.57 segment)

The cycle of the ring of pull stars means grabbing a pull star immediately isn't necessarily faster.
After much testing, it was found to be marginally faster to do a land cancel before grabbing the pull star planet, as this leads to a slightly higher and faster first pull where the speed adds up later on, while also catching the cycle at a perfect time.

The rotation of Luigi in the bubble matters for star grab, and can make up to 4 frames of difference. In this case, it saved 3 frames to twist him around using the IR cursor.

Green Star 3 - 16.86 IGT - 1:34:16.10 real-time split (1:12.05 segment)
This is an improvement of 1.6s over my 2020 TAS: youtu.be/-KnxWlYubGA

The bubble cycle is either global or the shooter is simply close enough to stay active (you can see the bubbles firing from the first planet). This means hyper-optimisation of the whole level isn't necessary provided you get the fastest cycle.
Despite this, I did optimise the first section as best I could, including an early release from the bubble to save a couple of frames on the launch star.

The main improvement to get an earlier cycle was going around the wall and setting up a boosted triple jump.
Jumping over a void (or moving floor) from a moving floor will make you accelerate in the air even when falling forward. This - plus landing in neutral to conserve more speed - led to the double jump being as fast as a long jump, and the triple jump then almost reaching the speed cap of 32, all despite going the opposite way of the moving floor.

Green Star 2 - 14.80 IGT - 1:35:13.83 real-time split (57.73s segment)
This is an improvement of 19 frames over my 2020 TAS, despite no-cutscenes losing 3 frames on the pull stars: youtu.be/tT7OmwDp0vg

Interestingly for the second planet, I managed to get a 6f faster route by reloading the wrong savestate during playback. A setup for a different route idea ended up making the main route idea faster, thanks to landing on the edge of the wall in such a way to be able to long jump without a stuttering run, or just immediately falling off.
Backflip saved 1 frame.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Gravity Gauntlet Green Stars

PurpleSun 2023-06-10 | Green Stars TAS WIP
Star order would matter with the fireball cycles for GS1 if it weren't for the 2P freeze strat levelling out cutscenes and no-cutscenes. Otherwise GS2 first would be 3 or 4 frames faster.

Green Star 2 - 19.86s IGT - 1:31:06.06 real-time split (1:16.11 segment)
This is a 0.69s improvement over my 2020 TAS (0.69s plus cutscene difference): youtu.be/5PdldVGADec

One of the key improvements was being able to long jump into the sideways lava wall without having to spin cancel, which saves 9 frames. Usually, getting a lava-wall boost here would send Luigi downwards (As seen here by nemi43: youtu.be/EWGiL2oEMRg?&t=218) but there is a precise spot in the corner that sends the boost sideways.

The brief 180 in the boost sequence helps to line up the third boost to get an earlier gravity switch that can stay close enough to the left to be able to get the extra boost, while also managing to go further up the new gravity field to end up closer to the star. This saves 18 frames over trying to go straight right the whole time.
The other key improvement is in the final boost, where it is 21 frames faster to go around the Whomp rather than under it, which was only possible with the earlier gravity switch.

A couple of frames were saved by going neutral before double jumping which affects jump height and speed, and another frame was saved by holding back for a frame during the double jump to have a lower speed triple jump, keeping it closer to the star.

Green Star 1 - 11.33 IGT - 1:32:00.48 real-time split (54.42s segment)
This is a 0.7s improvement over my TAS made in 2020: youtu.be/y-k1jR0HPbs
The IL WR by Ethanrocks (youtu.be/m759Lysh9IU) showed that a single jump was enough height for a fireball boost to reach the crate, which made up the bulk of improvement over my TAS.
The issue with this boost is that it won't let you long jump until the end of the "fire run", but this run can be cut short by going onto an edge.

Some slowing down of the long jump around the curve resulted in a better long jump off the step that overall saved 2 frames.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Chompworks Green Stars

PurpleSun 2023-05-14 | Green Stars TAS WIP
Movement to this galaxy is a cursor click, 9f faster than analog stick.
Star order matters due to Chomps being further along with cutscenes, making GS2 a little over 3s faster to do first.

Green Star 2 - 33.71 IGT - 1:27:24.08 real-time split (1:29.65 segment)
This is an improvement over my 2020 TAS (youtu.be/HgSvHzpwEuE) although cutscenes makes it difficult to know how much by.

Waiting for the Chomp, I have some fun with slopes and the Gearmos.
Looking back over this, I've realised I forgot to optimise the launch star cutscene by minimising its camera movement, so there's probably a couple of frames to save there.

Bouncing on the Gearmo is 2f faster as it improves the long jump sequence making the last long jump need less turning and straining. A single jump at the end was 1f faster.

I originally had a 33.95 with a lower flip panel long jump, which had less speed and jumped later, but Shadow64 suggested that a faster jump was possible.
I had tried faster before, but they also went too high and over the star, but after the suggestion was able to optimise one low enough for a spin to hit the star, saving 14 frames.

Green Star 1 - 7.70 IGT - 1:28:15.11 real-time split (51.03 segment)

The speed of a zip jump depends on your angle compared to the slope's angle - the more perpendicular you are, the faster the zip. The problem for this star is that faster zips would go way off the side.
However, I was able to optimise a jump to get 31spd with a good angle to set up a stretched-out triple jump (conserving speed with neutral in the single jump, no neutral is faster for double jumps for some reason).

If you run on the pressure pads for at least 1 frame, you're able to get moving platform momentum boost, so the triple jump was delayed a frame.

Overall the zip jump route was 0.5s faster than long jumping plus momentum'd triple jump.

Green Star 3 - 39.65 IGT - 1:29:49.95 real-time split (1:34.84 segment)
I used the current IL WR to help with strat ideas (youtu.be/WM5o2eKxPVE).

Zip jump route was 0.5s faster than long jump route (IL's lj route was the fastest one by 1 frame, over forward slope jumping up).

While instawalk is possible by entering while crouching, and can be faster by entering in neutral crouch, it is marginally slower (at least 2 frames) due to the time it takes to set up crouching into the launch star.

While long jumping on 2nd planet, I take a frame to slightly adjust the angle, but I do this on the very edge where Luigi gets thrown off the edge. This is called edge grinding, which has slightly more speed than normal running (18 compared to 15) and even adds a bit of initial speed to the long jump (19 compared to 16.5).
Triple jump up was 0.3s better than anything else I could try with the Chomp right there.

There was a potential ending strat involving a spring wallbounce off the back wall ( twitter.com/PurpleSunSR/status/1657853806787403778 ) which could save 0.5s, but the screw cycle doesn't work out.

birthday upload :)
#toolassistedspeedrun #supermariogalaxy2

[TAS] SMG2: Starshine Beach Green Stars

PurpleSun 2023-04-29 | Green Stars TAS WIP
Star order doesn't really matter here - the Lakitu for GS1 is the same with or without cycles.
I showed the movement on the world map just because it's one of the faster cursor-click movements (11 frames over joystick).

Green Star 2 - 23.65s IGT - 1:23:09.11 real-time split (1:20.00 segment)

While carrying the land cancel out to the water enters water faster, it's facing the wrong way so the movement in water is much slower than for long jump ground pound.

Long jump on the slope works by landing on the very edge.
While the trajectory of the long jump looks off, this was the perfect setup for a zip jump that went off a small part of slope angled towards the Yoshi egg at 60+spd.
This zip jump (plus conserving speed for the single jump by landing in neutral) saved 3 seconds over long jumping across the gap.

With backwards Yoshi flutters, you get higher acceleration but your speed is fixed once you reach a certain height above ground.
By keeping low to ground, acceleration is even higher and can get above the usual cap of 32spd up to 37.
Eventually it decelerates back down to 32. There is also a short burst of speed after dismounting.

Green Star 1 - 24.51s IGT - 1:24:16.96 real-time split (1:07.85 segment)

After cutscene skips, Luigi is spawned slightly down-left of centre which is why the start goes in a different direction.

I tested three routes for this star: #1: Same as GS2, including varied flutter routes trying to bounce off the Pianta and instant triple bounce off the fruit. #2: jda7's Lakitu bounce ( youtu.be/2uiiaGeGeWY ). #3: An alternative Lakitu bounce or straight flutter going under the tunnel.
The results were: #1 24.80, #3 24.61, #2 24.51.

Green Star 3 - 43.43s IGT - 1:25:55.58 real-time split (1:38.62 segment)
This is an improvement over my 2021 TAS demo: youtu.be/hg8Tyo10qE8

UPDATE: This has been improved, a 42.28s just about managing to get a boosted leaf jump to work instead of waiting for another cloud. youtu.be/1fOPdsSK5ic

A Blooper launch is possible (seen here: youtu.be/PTWmPQlxNIw) but it takes too long to lure the Blooper to the surface.
Other launching ideas couldn't get the balance of horizontal and vertical speed necessary for where they are.

Zip jump on the outskirts of the starting area was 0.6s faster.
It is possible to zip jump on the grassy slope above the tunnel, but the left side can only be done high up which takes a while to get up, and the right side only gets jumps going way off to the left. Ultimately the best was 10 frames slower than long jumping.
It's possible to shore launch up the tower, but it takes too long to set up, making it 1.4s slower.
Cloud cycles are manipulated by keeping them off-screen.
The walljump spin was able to snap directly onto the cloud, allowing for a backflip which saves 0.5s.

The issue with the cloud cycles is where the other cloud is when starting the cloud long jumps.
You get much more speed (34spd compared to 20) if you jump from a moving object over a void or very high above ground, so the ideal is to have no clouds under Luigi when starting the long jumps, which is difficult with the fastest cloud cycle for climbing the tower.
I managed this by slightly diverting where the first long jump starts but still having a trajectory towards the star, which 'edged around' the cloud below.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Honeyhop Green Stars

PurpleSun 2023-04-18 | Green Stars TAS WIP
Star order doesn't matter here since no affected cycles change movement between either star.

Green Star 2 - 44.35s IGT - 1:20:39.13 real-time split (1:40.58 segment)
This is an improvement over my 2020 TAS: youtu.be/0YeVYsjCiX8

Left side beeless climb (nemi's IL route: youtu.be/iBfqJ6XjF6s ) was initially 20 frames faster than right side, and then the zip jump was found.
This zip jump is extremely precise - too close to the edge and it counts as an edge jump which has normal properties (speed capped to 32, decelerates when moving forward), but too close to safe ground and it stops dead. It's a very thin line to get the uncapped 59spd jump.
This speed carries into the swing and remaining momentum was conserved in the single jump.

Facing left on the swing (and the setup required) ends up grabbing the wall 9 frames later than facing right (which has a better setup), however - because wallgrabs are more lenient while facing away from the wall - it is only possible to get this glitchy 90+ degree wallkick facing that way, and this wallkick is 21 frames faster than a normal one, making this 12 frames faster overall.

The instant sideflip conserved running speed into the backflip, and then landing on the top part got higher acceleration to long jump early enough to graze past the bubble.
For the floaty fluff, the angle coming off the earlier swing is usually the set angle coming off the floaty fluff, but this angle can be altered by turning (after neutral) while jumping into the fluff.
This meant I could face sideways off the fluff, gaining more speed while falling.

An alternative route for the climb that skips the flower flinger was only 3 frames slower getting the megawallkick off the Queen Bee.
Speaking of the Queen, she starts to talk if you stay on the ground for 5 frames.

Green Star 1 - 26.88s IGT - 1:21:49.11 real-time split (1:09.98 segment)
This is an improvement over my 2020 TAS: youtu.be/KG3QC0CJGzQ

Much the same as GS2 until the top of the 2D section.
Because it can keep enough speed to long jump, it's 0.65s faster to do this triple jump and jump on the cloud to get a long jump to the star than to use the floaty fluff.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Sweet Mystery Green Stars

PurpleSun 2023-04-08 | Green Stars TAS WIP
Star order matters a very little bit due to the Choppah cycles explained more below. GS1 first is 1 frame faster.

Green Star 1 - 7.30 IGT - 1:18:04.36 real-time split (1:02.70 segment)
Even with cutscenes, this is a 1f IGT improvement over my 2021 TAS (uses Mario as his acceleration is better): youtu.be/xrkFOBb8Qe4

2P egg break is tied fastest with a backwards double jump that spins at the end.

I had to delay jumping and fluttering on Yoshi because of a cooldown on spinning. Backwards fluttering stays at the same speed after a short time (the third flutter in this TAS).
The flutter acceleration - and subsequent speed (about 29 horizontal speed) - gained from the spin makes this strat a few frames faster than fluttering as early as possible.

Bouncing off the berry (possible by having Yoshi stick its tongue out with B) is a strat I lifted from Jerome's GS2 IL: youtu.be/Yz8QEhgTgY8
Bouncing gets a burst of speed and brings flutters straight to the cap of 32hspd.
It caps at 32hspd when ascending (y speed above 0) but can go a little higher when descending. The burst of speed is maximised by keeping a flutter through bouncing, as this makes Yoshi 'fall' earlier, being able to get up to 38hspd rather than the 34 when actually bouncing.
The bounce saved 0.2s.
An extra frame was saved thanks to the Choppah, which is where star order came in. With cutscene cycles it ends up right next to the berry, and with a 2P freeze I brushed by it so it nudged Yoshi's trajectory more towards the star. This saved having to strain harder during the flutters, which would have lowered the burst of speed.
Dismounting gets another burst of speed including a mini-warp of about 67spd.

Green Star 2 - 11.20 IGT - 1:18:58.55 real-time split (54.19s segment)

Backwards jump with a spin at the end was fastest egg break by 9 frames.
Same things apply as above with the spin-start flutter and berry flutter-bouncing.
Eventually the burst of speed runs out and there isn't much bonus falling speed when going (mostly) backwards, so I go for height-optimal flutters near the end to be able to dismount as early as possible.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG1: Dusty Dune 3 in 1:50.65 (Mario)

PurpleSun 2023-04-04 | Mario spin cancel jump box break reverse wallkicks through the sunbaked sand castle in 1 minute 50.65 seconds (6639 frames).

This is part of the second session of the SMG Random Star TAS where a bunch of SMG TASers get together and TAS random stars (hence the name) in an attempt to help fill up the TAS sheet.
Having already done this level as Luigi in 2020 ( youtu.be/saMCD7QtOAQ ) it was easier for me to take this one.

Others in the Random Star TAS Session:
BillyWAR - Beach Bowl P in 2:30.92: youtu.be/ohfUCNhEykw
Shadow64 - Melty Molten P: [WIP]
ToogsX - Freezeflame P: [WIP]
Xander - Sea Slide 3: [WIP]

SMG TAS Server: discord.gg/h2YSCZm

Land cancel isn't possible in the start. Buffered crouch acceleration allowed for an earlier long jump, a 2P long jump (single jump and Z+2P A within 4 frames) to get enough height to land onto the switch.

In the megasideflip, I hold left and then switch to holding right to stay in the "mega" area/state for longer, saving a few more frames, and then 2P jump to save even more frames.

This strat I found to improve Luigi's TAS ( twitter.com/PurpleSunSR/status/1326318044549959680 ) doesn't work out as well for Mario due to his lower speed and height making both the bouncing and long jump slower.
But a new strat was found with the quicksand route by getting a backwards-sideways slope jump carrying a lot of previous backwards speed, missing the secret planet and getting all the way to the Pokey planet.
It is possible to skip the Pokey planet but it involves a lot of gravity traps so it's much slower. twitter.com/PurpleSunSR/status/1643384872696684544

Although the glass tower is already an instawalk (doesn't landstun), I turn around in the launch star to be able to walk and long jump a few frames earlier, also doing buffered crouch accel to save a couple more frames.

The movement inside the tower is so scuffed, but there is cool new tech with a catchy name coined by Xander - spin cancel jump box break reverse wallkick (SSJBBRWK). This wallkick is frame-perfect and I'm still not certain what affects the wallkick angle, but rubbing up against different walls seemed to do something. Both times I used this saved 0.1s each.

It is possible to break boxes without a bonk by slowing down as you spin or standing far enough away. No-bonks were faster for the last two boxes, particularly the last one being able to get a backflip soon after the box break.

I missed a lot of starbits just not paying attention. 33-34 in view with the secret planet at 0:44, and then the 30 around the glass tower. Should be 93-94 total in this level.

#toolassistedspeedrun #supermariogalaxy

[TAS] SMG2: Flipsville Green Stars

PurpleSun 2023-03-31 | Green Stars TAS WIP
Star order matters slightly - the cycles of the moving platforms means GS3 is faster to do without cutscenes by 0.25s, while GS1 and GS2 differ because of the effect of land cancel on their routes. GS2 is 4 frames faster to do first with cutscenes.

Green Star 2 - 10.78 IGT - 1:15:11.96 real-time split (1:06.43 segment)
This is an improvement of 0.1s IGT over my 2020 TAS (youtu.be/pUWDsUgGFPw) however, this is with cutscenes, so the improvement is more like 0.6s.

Top running speed is gained at the start by doing the second long jump immediately on safe ground. While I don't make use of having top speed, this tech also gives higher acceleration, making the quick-turn for the third long jump 5 frames faster.
Getting onto the roof bracket to the star was already around 2 seconds faster than the RTA strat going to the side and backflipping up, but there is now also a zip jump that saves 0.3s.

Green Star 3 - 17.11 IGT - 1:16:12.41 real-time split (1:00.45 segment)
This is an improvement of 0.25s over my 2020 TAS (youtu.be/oAaXH5qgCjk).

I improved the Toad bounce by lining up a more sideways angle. Going more sideways in the air is a little faster than straight backwards (known from infinite fluttering), getting up to 34hspd compared to 32.
This led to ground pounding the grate 7 frames earlier.
The remaining improvement was general optimisation of the final section. This section includes getting the top running speed long jump tech, and doing a leap of faith into the star by using a mechanic of jumping off a moving platform over a void. This causes Luigi to accelerate when falling, even when moving forwards.
This leap also clips through the fence, getting ejected at 66hspd to save an extra frame.

Green Star 1 - 5.91 IGT - 1:17:01.66 real-time split (49.25s segment)
This is an improvement of 0.25s over my 2020 TAS (youtu.be/H2jkvp5i-1w).

The main improvement is neutral walljump strain, where going neutral on the joystick after walljumping can bypass the restricted movement walljumps usually have.
This was particularly useful for the second walljump, where I held Luigi back enough to avoid ledgegrabbing, and then forward to maintain speed when landing.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Supermassive Green Stars

PurpleSun 2023-03-26 | Green Stars TAS WIP
Star order matters due to the Thwomp cycles as will be explained later. Camera movement can manipulate the coin's cycle to work with both cutscenes and no-cutscenes. As all three stars use the coin, land cancel has no difference.

Green Star 2 - 8.56 IGT - 1:12:12.75 real-time split (1:06.69 segment)

At the start, the awkward-looking long jumps are necessary to get to top running speed (15hspd). The second long jump has to happen immediately when landing, and must be on safe ground away from the edge of the ledge.
Camera is turned to keep the coin off screen enough for it to be in the perfect position for the zip jump.
The zip jump works here because of both the shape of the slope and the running direction when approaching the slope, coming more across than down. The subsequent slide zips sideways from this, at 64hspd. Overall this saves 2.34s.
The weakspin at the end grabbed the star 1 frame earlier.

Green Star 3 - 18.46 IGT - 1:13:14.55 real-time split (1:01.80 segment)

This managed to be an IGT improvement of my previous TAS that used an IGT-pausing checkpoint abuse strat: youtu.be/0k-eeiC9gss
Zip jumping is used again to get a tight wallkick strat that was previously done with a long jump (youtu.be/KwHCkBmDdJU).

The reason for star order mattering here is that the Thwomp cycle without cutscenes, when manipulated with the same camera movement for the coin cycle, is perfect for the fastest way to the launch star.
With cutscenes, the cycle has the Thwomp high up in the air as you pass under it. However, it is actually 7 frames faster to have the Thwomp push Luigi down to backflip earlier, and the cycle without cutscenes achieves this making it faster to do GS3 after one of the other two.

At the end, the backwards bouncing off the Koopa saved 2 frames over the previous strat bouncing from a long jump.

Green Star 1 - 7.65 IGT - 1:14:05.53 real-time split (50.98 segment)

The coin cycle is manipulated to gain a momentum boost for the backflip to the star.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Supermassive GS3 in 18.75s

PurpleSun 2023-03-21 | Not part of the Green Stars TAS but an IL improvement of 0.83s (50 frames) over my previous 2020 TAS: youtu.be/rSjkTOOWvjc

The reason this won't be part of Greens is because the checkpoint abuse strat here is 0.26s slower in real time than a very tight wallkick strat under the Thwomps (which has also been improved since 2020).
It is only faster for in-game time because the timer stops for the life counter, being paused for roughly 2 seconds.

At the start, the first long jump needs to be at an angle to land on safe ground, and then the second long jump needs to be done immediately to get top running speed when next landing (15spd). This doesn't work if on an edge of a ledge.
This running speed tech makes the rest of the starting section much faster.

The big new trick is a new route to the checkpoint abuse. The camera is turned to manipulate the cycle of the big coin so that part of it is moving toward the checkpoint by the time I reach it.
The momentum of the moving coin is then used in a backflip to quickly get close enough to the checkpoint for P2 to grab it.
I couldn't manage to get a long jump spin to work from the coin, it's just barely too low.

At the end, P2 is used to freeze the Koopa to bounce to the star. Releasing the Koopa makes its hitbox expand slightly for an earlier and higher bounce.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Bowser Jr.s Fearsome Fleet Green Stars

PurpleSun 2023-03-18 | Green Stars TAS WIP
Star order matters only a bit, as the position of the Fizzlit is a little different between cutscenes and no cutscenes, but this overall changes the optimal fluttering line by a few frames.

Green Star 2 - 20.45s IGT - 1:10:10.16 real-time split (1:21.00 segment)

This is a 0.5s improvement over my TAS tests in 2020, meaning this IL could be improved: youtu.be/3R1MdYG6iDM
There's improvement in the route to Yoshi, getting a much earlier long jump off the statue at the cost of a less direct long jump route, saving 5 frames overall. This is 6 frames faster than long jumping around the normal way.
I tried for a zip slope jump, but the slipperiness is inconsistent around the statue and the only zippy slides went straight off the edge, unable to jump with the added speed.

The Fizzlit bounce strat is from nemi43 way back in 2012: youtu.be/dBFVlublQTc
After 10 seconds of infinite fluttering, your speed gets cut to 0. Bouncing off an enemy resets this 'airtime timer', so I take a curved flutter line to bounce off the Fizzlit.
I flutter as early as possible, taking advantage of higher acceleration when fluttering close to ground, as well as higher acceleration from a more backwards-moving flutter before shifting sideways for a higher top speed.
These advantages - as well as a cooldown on being able to spin (head-shake) after mounting Yoshi - outweighed the acceleration gained by shaking before starting to flutter. Shaking would have been slower by 10 frames.

When fluttering, the top horizontal speed is 32 while rising. But while falling, speed is dependent on your angle, with 34.7 the maximum when as closest as possible to a right-angle.
The rapid fluttering towards the end maintains this higher falling speed.

Green Star 1 - 12.83s IGT - 1:11:06.06 real-time split (55.90s segment)
An improvement of 9 frames over my 2020 TAS: youtu.be/ubBjFiyfrek

As a straight line is better here, the flutter start is different, since backwards acceleration would cause a slow curve. This means the acceleration gained from shaking is actually faster here.

When dismounting, I jump into the Gunner to get a slight push from it to grab the star 1 frame earlier.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Beat Block Green Stars

PurpleSun 2023-03-15 | Green Stars TAS WIP
Star order matters here due to the beat block cycle. Since GS1 mostly doesn't touch them, it's GS2's route that matters, and the no-cutscene cycle has slower blocks midway and near the end, making it 0.5s faster to do GS2 first with cutscenes.

Green Star 2 - 13.58s IGT - 1:07:56.55 real-time split (1:09.44 segment)

Due to how close it is to the edge, the crystal at the start has no way to setup an extremely fast zipped slope jump, so it is faster to be left alone.
In the long jump sequence, running speed can be kept to speed up the single jump-double jump up the block, and then keep speed for the crystal.

This first crystal slope jump is a little slower than the others due to the limits of a quick setup, plus aiming the right trajectory with a good facing angle, which would minimise turning time and speed loss to jump straight to the next crystal.

With the next crystal jump, I get 60hspd and a great trajectory to land safely on the yellow block. Landing safely is important so that landing with neutral joystick can conserve speed into the following single jump (capped at 25spd). Landing on the edge in neutral causes 0spd.
The facing angle is also important as I can easily strain the single and double jump to set up the last crystal jump.

This last jump gets 73hspd, finishing with a single jump straight onto the green block and instant backflip spin to the star.

Green Star 1 - 9.51s IGT - 1:08:49.16 real-time split (0:52.51 segment)

My hope with this crystal slope jump was a facing angle and trajectory that could lead to a conserved speed single jump onto the side of the last crystal to conserved speed double jump into the star.
But I couldn't get that to line up. Good angles led to bad trajectories off the map, and good trajectories had not so good angles.
So I had to make do with a 71spd slope jump and quick turn to long jump onto the side of the crystal for a long jump spin to the star.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Haunty Halls Green Stars

PurpleSun 2023-03-09 | Green Stars TAS WIP
Decided to start the video early for simple IR art

Star order matters due to the cycles for the green moving platform and the Octoboos. It is 0.7s faster to do GS1 before GS2 because of these.

Green Star 1 - 6.76s IGT - 1:04:11.06 real-time split (1:02.40 segment)

If you jump when (safely) on a moving floor, your jump accelerates after its peak rather than decelerating. However, you won't get this speed boost if you jump from the edge.

Also, top running speed is gained and conserved by long jumping as soon as you land from the first long jump.

Small hop to line up the final boosted long jump, saving a few frames over running due to the jump speed boost and less drag from the floor.

Known improvement: after finishing GS3, it was found that neutral speed conservation (landing with neutral joystick) with the jump speed boost could make a triple jump to the end 5 frames faster than long jumping. Want to move on so won't go back.

Green Star 2 - 11.83s IGT - 1:05:06.25 real-time split (55.19s segment)

4f improvement over my 2020 TAS: youtu.be/Sy6O893-okk

Camera turn twice at the start to manipulate the green moving platform later on.
2P saves a frame hitting the Octoboos at the end, as this allows Luigi to spin and enter the launch star earlier without hitting them and getting stopped by the impact.

It is 10 frames faster to slow down the long jump to land onto the bump and backflip from there.

Green Star 3 - 45.70s IGT - 1:06:47.11 real-time split (1:40.86 segment)

Getting behind the castle wall is 1 frame faster than going around.

The lips on the side of the moving platforms are sloped, and with this particular slope, the fastest jump speeds happen at the edge.
Edge slope jumps don't have the fixed speed and angle of regular ones which means you'll lose speed unless moving backwards.
Fortunately, it's possible to turn around just as the jump happens, and edge slope jumps are also able to spin, making this small skip possible and (initially) saving 19 frames over a normal route, and 11 frames over a slope jump from the second moving platform.

"(Initially)" because of the sling star. Normally this will force Luigi into an angle where he stops when landing, but by starting to turn (from standstill) before entering the sling star, Luigi's angle will change.
With this, I have Luigi face somewhat left so he skids when landing, allowing for a much earlier long jump.
This is why I delay the long jumps to the sling star, so that I can land next to it and turn, and the long jump reroute is why the timesave of the earlier trick is different overall, being 20 frames against long jumping around, and 13 over a later slope jump.
Also there's some vectoring or low ceiling that gets a really low long jump.

On the final stretch, the broken bricks almost led to a cool slope jump trick, but can't make it all the way across.
Instead, jump speed boosts are used again, picking up even more speed by bouncing off the Mattermouths. I shoot a starbit at the last Mattermouth so it turns around into a better bouncing spot.
Speed is then conserved into the jumps at the end by landing in neutral.

massive wall of text Chatting

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Freezy Flake Green Stars

PurpleSun 2023-03-03 | Green Stars TAS WIP
Star order doesn't matter here, although it almost did with a slower GS2 strat involving the flying ? block.

Green Star 2 - 37.95 IGT - 1:00:59.31 real-time split (1:34.05 segment)
A longer description of this star can be found here: youtu.be/j_J_UWSEdpM

The first launch star is a balance of optimisation for the earliest statue break while still being able to get to the launch star on its first active frame. The strat used was fastest by 29 frames, and also is good enough to enter the launch star at 0spd which somehow makes the instawalk 5 frames faster.

With certain movement, conical or jagged slopes can cause some very buggy fast slides for a frame or two, leading to glitchy slope jumps. These slope jumps have a fixed speed and angle trajectory.

As with Tall Trunk's slide, the path you take doesn't matter, until the end where exiting upside down causes Luigi to face backwards, leading to a smooth pipe strat 35 frames faster than a forward exit.

Another glitchy slope jump on the Goomba statue, but couldn't get one at that speed with a better trajectory than hitting the lava near the end. This strat is 1.4s faster than a normal long jump route.

Green Star 1 - 20.43 IGT - 1:02:03.06 real-time split (1:03.75 segment)
Because of the different set up not having to detour to the Fire Flower, the glitchy slope jump gets a straighter trajectory. Speed is conserved in the single and continues into the double jump after breaking the Bowser statue, being fast and high enough to land on the ledge to the slide.

Green Star 3 - 10.43 IGT - 1:03:08.66 real-time split (1:05.60 segment)
Adjusting for cutscene difference, this is an improvement of roughly 3 seconds to my TAS from 2021: youtu.be/K5xMO-TfyCI

For details on the rock mid-air jumps, here's Joselle's doc: docs.google.com/document/d/13HenHN3IA2mQih2k8Ve_VZdK-yIoX8fnXB2g2eMaZGM/edit

Most of the timesave comes from a completely direct route to the star rather than the mid-air turn from the previous TAS.

Remaining timesave comes from understanding how horizontal speed works with rock jumps: it's a simple case of waiting between each jump, lower vertical speed causes higher horizontal speed. With this in mind, I steady out at 59-60 hspd jumping when yspd drops to -20, which hovers just above the DeathArea.

SMG TASing server: discord.gg/h2YSCZm
#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Freezy Flake GS2 in 37.95s

PurpleSun 2023-02-18 | Green Stars TAS WIP

37.95s IGT - 60:59.31 real-time split (1:34.05 segment)

Freezy GS1/2 were two of the few stars I hadn't tested prior to starting this TAS, as star order didn't seem like it'd matter (more about that later).

The first launch star is a balance of optimisation for the earliest statue break (not waiting so long for the launch star to activate) while still being able to get to the launch star on its first active frame.
While other routes looked cooler (e.g. getting turned backwards in mid-air to have more speed), the fastest by 29 frames was how RTA runners do it (used Pottoww's IL as a blueprint: youtu.be/QWP7saYcBaw?t=10).

An instawalk from the launch star is common here, but interestingly, entering the launch star while turning ends up getting landstun, and entering while standing completely still gets faster and earlier acceleration.

1-flower cycle was attempted but failed (youtu.be/6GEkxIaaQ-k), missing out by around 0.6s.
With certain movement, conical slopes (e.g. the mountain, some crystals) can cause some very buggy fast slides for a frame or two, leading to glitchy slope jumps.
The consequence of slope jumps being patched in SMG2 is that they now stay at a constant speed and trajectory.
It is difficult to control the trajectory, and the jump in the 1-flower attempt has a much more wayward and slow setup after getting the 2nd flower.

As with the Tall Trunk slide, you don't lose any time with whatever movement you do provided you don't jump a lot.
However, exiting the tunnel without jumping is a little slower and doesn't let you spin/throw fireballs. Also, the gravity you leave the tunnel at will have an affect on your angle, sometimes changing angle mid-air.

Backwards exit with a mid-air fireball was 35 frames faster than the RTA method: forward exit, jumping on the lower brick block and fireballing around the corner.

Star order almost mattered at the end. The fastest regular route (by 6 frames) was to long jump onto the flying ? block, and its cycle changes with cutscene skips.
But another glitchy slope jump was found on the Goomba statue, getting a 67hspd jump that barely misses the lava geyser. Unfortunately this kind of jump couldn't make it onto safe ground, but the lava boost and movement on the sloped boundary was 20 frames faster than holding it back onto the ice platform and long jumping.

#supermariogalaxy2 #toolassistedspeedrun

So close to one-flower cycle

PurpleSun 2023-02-14 | Just showing Greens TAS is still alive :)
Tried to get to the pipe for Freezy Flake GS2 in one single Fire Flower usage. Just under a second off where I think you can start fireballing (you can't inside the tunnel).

[TAS] SMG2: Rolling Masterpiece Green Stars

PurpleSun 2023-01-02 | Green Stars TAS WIP
I included the World movement because it is mind-blowing movement with the small IR click saving an astonishing 2 frames and breezing past Freezy Flake.
Star order doesn't matter for this galaxy.

Green Star 2 - 58.83 IGT - 58:07.10 real-time split (1:55.59 segment)

1.8s improved from 2020 despite starting from cutscenes: youtu.be/mTI_hX9rb5o

Backflip was 25 frames faster than long jump spin - a spinless long jump can't land directly onto the ball without a little more height as well as slowing down significantly, otherwise it just gets the forced jumping mount. Spinning from backflip was 21 frames slower.
The triple jump chain into the launcher was 2 frames faster than rolling and doing 1 jump.

Optimising launcher entry involves more than just aiming for the middle as too much speed will cause some slow swirls, but also having too little swirling causes the ball to vibrate in the middle, taking longer to lock in and be launched.
Because of this, it's faster to aim slightly off-centre and slow down in the air to avoid the vibrating while still minimising the time spent swirling.

In the 2020 TAS, I had thought that jumping through the red brushes was a little bit slower, but what I missed was getting the trajectory just right to both minimise the speed lost from hitting the brush and to hit the side of the yellow brushes when landing.
It turns out that this hit avoids the usual landing slowdown, making the route much faster, about 1s faster in fact.

2P can freeze the starball. This is used to quickly turn around from the key's cutscene.
Trajectory of the next launcher is too fast to jump straight out and land safely, but some speed is retained with the movement done.
The jump straight onto the top disc requires all that run-up and a jump on the last possible frame.

Like the yellow brushes, hitting the arrow sign when landing avoided the landing slowdown to have more speed for ruler bouncing. In fact, it kept too much speed so I weaved in the air to slow down enough to catch the edge of the ruler to get a faster fall towards the star.

Green Star 1 - 35.08 IGT - 59:25.26 real-time split (1:18.16 segment)

1.1s improved from 2020.
Skipping cutscenes means a different starting position. A direct backflip wasn't possible, so long jump spin was fastest by 6 frames over a backflip spin.

For the star grab, there's no jump height difference that depends on horizontal speed, so jumping directly underneath would have worked just as well as the big run-up.

Happy new year!

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG1: Gusty Garden Secret in 46.95s

PurpleSun 2022-12-30 | Rainbow Mario destroys the Golden Chomp in 46.95 seconds (2817 frames).

In an effort to give every star in Super Mario Galaxy 1 a TAS, a bunch of TASers got together, spun a big wheel with the level names, and made some LOTADs (low-optimisation tool-assisted demonstration. Bars.). I bagged Gusty Garden Secret, spending only 3 and a half hours on it.

Check out the other TASes:
- Toogs: Gusty Garden 3 (youtu.be/HZbwMH7MKgY)
- BillyWAR: Melty Molten S (youtu.be/-lpnaCX_Rpw)
- Xander: Space Junk S (youtu.be/q5qmgP59UiE)
- Shadow64: Sand Spiral (youtu.be/VDaQDqlj46I)
And the TAS server: discord.gg/h2YSCZm

I missed out on the big VC, but the next day I had Billy and then SaladBowl for support while they researched bonkwarps and saved A presses respectively. Fan favourite Toogs made a cameo appearance.

With Toogs having done Gusty 3 with cutscenes, I skipped them as this star is in the same level, emulating a full-game run.

Buffered crouch accel (holding Z before you gain control and letting go soon after) allowed for a 1f earlier long jump.

I made sure to stay in the gust tunnel as much as possible while still grabbing the ? coins, using Toogs' Gusty 3 and Tibs' IL record (youtu.be/cRsxCKsn1qc) as blueprints.
This could still be optimised, though, as I only really had a few glances at how much actual speed I was keeping.

My initial strat with the last two coins and Piranha Plant was to hop off the floaty fluff facing forward so that I entered the planet facing backwards, using that momentum to get the first coin and swing straight around either to the Piranha or the other coin.
The issue with that was I had to wait to hop off to be far enough along to swing right around.
My best effort with this was still 1s slower than the final strat.

The final strat landed forwards, landing much earlier, and gathered a lot of speed before landing so I could single jump with all that speed (about 23hspd, above the top speed of a long jump) and get all the way around the planet to get near the Piranha.
The Piranha kill could be improved by a few frames with a slightly earlier spin or a better jump line, however, the real bottleneck is vine RNG.
This RNG can be manipulated with different timed jumps and spins, so it could end up the same whether you kill the Piranha earlier or later. It could also end up much much faster. More playing around needed here.

The last planet was disappointing. I had hoped for some gravity shenanigans, maybe some momentum conservation to have super fast single jumps and ground 2P jumps, but long jumps won in the end by a handful of frames.

#supermariogalaxy #toolassistedspeedrun

[TAS] SMG2: Cloudy Court Green Stars

PurpleSun 2022-12-11 | Green Stars TAS WIP
Star order matters a bit here, as the land cancel affects the different starting routes, but is no different between GS3 and GS2 (once getting to their proper split at the drums), but 8 frames slower in GS1.

Green Star 3 - 48.73 IGT - 54:20.85 real-time split (1:44.65 segment)

2.43s saved over a rough 2020 TAS

While the starting route is a little later bouncing on the first drum than the route used in GS2, it makes up for it with the backwards facing bounces going further to the right to make the big skip a little faster.

This version of the skip is 1.1s faster than the IL route starting cloud long jumps from the cymbal.
An alternative version that starts with a cloud sideflip was only 2 frames slower.

From the sloped tower in the 2D section, I landed on the very edge where Luigi would long jump off a slightly lower floor and get a low long jump.
The effect of this compared to a normal long jump then carries on through the rest of the star making for different jumps up the crusher. Overall, the route saved 4 frames.

Green Star 2 - 18.65 IGT - 55:22.83 split (1:01.98 segment)

0.45s saved over my 2020 TAS: youtu.be/_K_sa1GU4DY

As explained above, this starting route bounces on the first drum a little earlier, but is faster here.

Frame-perfect cloud spin cancels and long jumps get a much higher boost than most mid-air cloud ljs (about 23-24 initial yspd compared to anywhere between 13 and 18 - and those 18spd jumps have to be very delayed).
They're enough to start jumping from almost under the last drum, and save 1 second over a single cloud lj from the cymbal.

Green Star 1 - 5.35 IGT - 56:11.51 split (48.68 segment)

2f saved over my 2020 TAS: youtu.be/55cF7iXMaFA

Landing in neutral after the first cloud spin cancel conserved a little bit of speed, enough to be able to instantly long jump. However, much more height and a better angle was gained in the long jump by delaying it a few frames.

Originally I had two routes tied at 5.36 - one was the 2020 route (lj + 2 cloud sideflips), another was instant backflip + backwards single jump + cloud sideflip.
However, @EpicRandomGamingChannel (Xander) then challenged me to untie one of those routes, helping me out in VC.
We ultimately found a slightly better 2020 route by doing an instant backflip from the long jump - this has less y speed but more horizontal speed than the first sideflip (which goes away from the star), and this meant an earlier cloud sideflip (that goes toward the star) could get under the star to hit it at a low weirdly extended point and save the frame.

#supermariogalaxy2 #toolassistedspeedrun

Luigi climbs a tree

PurpleSun 2022-11-27 | 8.51 IGT - 50:28.16 real-time split (51.85 segment)
Improves on 2020 TAS by 0.32s (19 frames)

Mega triple jump strat found by Vellopia (original vid deleted)

[TAS] SMG2: Tall Trunk GS2 in 21.70s

PurpleSun 2022-11-26 | Green Stars TAS WIP
Plan is to upload this now, GS1 as a Short to test that out, then both with GS3 which might take a while as I have not tested anything prior for sliding mechanics.

21.70s IGT - 49:36.31 real-time split (1:19.78 segment)

An improvement over this 2020 TAS by -1.08s: youtu.be/JawBPjpTBXU

Star order between this and GS1 comes down to how their first planet routes work with a land cancel.
GS2 benefits much more as the land cancel can combine with an instant sideflip for a faster start.

This sideflip strat is 0.3s faster than a mega double jump walljump spin up to that ledge, which is also faster than a triple jump.
The run-up to the long jump had to be a bit wavy in order to have a good facing angle for the instant sideflip to be able to walljump.

Instant sideflips are done by holding Z and backwards to facing angle, and pressing A on the frame of landing. Sometimes this doesn't work, particularly when landing on an edge; haven't quite figured out why and how to reliably fix it.
At the top, neutral landing allowed for an instant long jump into the launch star.

This whole route was more than a second faster than doing a mega triple jump.

On the second planet, it isn't possible to jump land cancel, which would have made it even faster to do a small jump to set up the two backflips. With camera cancel, this is only 2 frames faster than a backflip spin at the start into another backflip.

As found in the 2020 TAS, there is a little bit of wiggling about required after ground pounding to set up a backflip that gets an unusual speed boost off the ceiling.
Yoshiless was only 6 frames faster than using Yoshi. You can see what Yoshi strats look like here: twitter.com/PurpleSunSR/status/1596573885084102656

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Bowsers Lava Lair Green Stars

PurpleSun 2022-10-31 | Green Stars TAS WIP
Star order matters due to the Flomp (flippers) cycles for lava boosts and the Mattermouth (Dry Bones head) cycles for GS2.
GS1 first is faster.

Green Star 1 - 35.18 IGT - 46:46.73 real-time split (1:31.05 segment)

Despite not having running speed out of the ground pound, the squeezed backflip from the Magikoopa saved 15 frames.
Due to the shape and gravity of the long planet, long jumping early ends up getting lower hspd and a lot of airtime with it too - running to the edge and long jumping saved 3 frames because of better gravity.

The big strat is the lava boost. There's a small lava wall in the darker jaggedy lava starting the Whomp section, and hitting the wall causes a horizontal boost which doesn't stop accelerating, and this state is conserved (although speed is always cut down) when hitting lava again, and also with the little bounce on normal ground.
Hitting the rock at a certain point causes a fall state with 0 speed, which, like falling off in neutral (as shown here by Joselle: youtu.be/W28X4lLB2nY), allows Luigi to briefly run across lava.
From this, a single jump and slide down the rock was 1 frame faster than running around and long jumping.

Also, due to this Greens run having a file with 9999 coins, there is no textbox at the end.

Green Star 2 - 46.70 IGT - 48:16.53 real-time split (1:29.80 segment)

With the 2nd Flomp out of the way, and P2 breaking the question block, the lava boost strat can make it out to the rock named finger and bounce to the Luma's steps.
Attempts to skip the gate were either too slow (using the well-known strat with jumping up the front of the gate and over) or didn't work with everything I tried (hyper slope jump off a falling Whomp couldn't get angled right).

Glitchy slope jump off the crashed meteor saved 3 frames. Going directly forwards would get the glitchy speed, but jumping would immediately stop to 0, whereas jumping further off the side kept all the speed, although it goes off to the opposite side too.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Honeybloom Green Stars

PurpleSun 2022-10-25 | Green Stars TAS WIP

I had a longer description describing the slower strats compared against but it ended up with like 4k characters so trimmed it down.

Star order matters in this galaxy as the cycles of the Choppahs are changed by cutscenes. This changes the strats for GS2 in the bouncy fruit and pole sections, and cutscene cycles end up a little bit faster (8 frames), so GS2 goes first.

Green Star 2 - 57.01 IGT - 43:50.98 real-time split (1:52.88 segment)

At the rainy wallkicks, backflip spin wallkick was 14f faster than the nearest strat involving a Goomba bounce. Neutral was used when landing to be able to long jump immediately
Right after, long jump spin wallkick spin to land higher and then backflip to the top was 8 frames faster.
I held Z when landing and let go to get buffered crouch acceleration, saving a frame around the corner.

The waterfall section: As the final Choppah was high enough for 2P to freeze and Luigi to squeeze under, the main goal was the fastest route to get to that squeeze point grabbing the ledge.
Backflipping onto the tower was the fastest way to get to the cloud & flower used to get to the ledgegrab point, with the next closest being a long jump under the tower into a triple jump spin (27 frames slower).

Due to the squeezing, Luigi automatically climbs the ledge, and also gets a little proxy boost on the way. What is done here affects the route to the vine: a backwards jump with long jumps into the Beeshroom + a frame of hovering was 8 frames faster than the best route jumping forwards.

When Bee hovering, you get 8hspd if you're hovering upwards, and 10.25 when going down. Mashing A when just barely dropping keeps good height while maintaining the highest speed.
The bounce into the frozen Choppah gained a bit of extra speed skimming (clipping) off the tail, making it the fastest way onto the cloud.
Spinning towards the Choppah while letting go of 2P A ended up getting the damage boost a little higher, so I was able to bounce earlier and land the damage boost 1 frame earlier.

Spin-cancel ground pound grabbed the pole 1f earlier than a spin.

Another squeeze through a Choppah. 2P freeze changes its collision slightly, but starbit stunning during the freeze restores it. This was used to get another small proxy jump into a double jump, saving 12 frames.

Green Star 1 - 42.00 IGT - 45:15.68 real-time split (1:24.70 segment)

As explained above, the Choppah cycles are different without cutscenes, so the waterfall route has a new look, as you can't squeeze under the last Choppah. Along with this, the first Choppah gets in the way of GS2's faster backflip tower route.
In the end, this new triple jump route with a skim across the top of a 2P frozen Choppah ended up 14 frames faster than the nearest, which was an altered version of the backflip tower route.

For the bouncy fruit, taking the higher road with an extra coin refill was 25 frames faster than the low route I took in this 2020 TAS: youtu.be/3cR-YPmpx9Q
For some reason, there was a huge boost off the last fruit, more than double the initial speed of a regular bounce. The starbits had no effect.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Cosmic Cove Green Stars

PurpleSun 2022-10-15 | Green Stars TAS WIP

Green Star 1 - 20.58 IGT - 38:03.86 real-time split (1:16.51 segment)

For a more in-depth description, see the individual video here: youtu.be/XMDBlu3_u48
Brief notes:
- The start doesn't have any jumps in order to preserve hard landing state, which allows
the penguin launches to have uncapped horizontal speed.
- The launches work by clipping into the penguin, and then getting clipped out while spinning. If the clip-out reaches the surface of water, you get a spinny dolphin jump with all the clipping speed.

GS2/GS3 star order used to matter before new GS3 strats, as the cycles of the moving water blocks are different between cutscenes and no cutscenes. Neither of the two have water blocks sitting in the way of the shore launch, so there is no difference now.

Green Star 2 - 1:04.71 IGT - 40:03.76 real-time split (1:59.90 segment)

Ice-skating mechanics:
- Skating coasts at 18hspd, and spinning adds a boost of 9.25 to whatever the current speed is, but doesn't boost if current speed is above 18.75
- Long jumps have a fixed speed, while standard jumps are based on ground speed, while adding a small boost on top. Ground speed is conserved through jumps.
- There are two different deceleration rates of spin-skates. No full list of what causes what, but the lesser rate kicks in when you've been on ground for a short time. This is why I wait to jump at the start.
- Also for some reason uncrouch spin gained more speed than a regular spin for that initial one.

In the cave, 2P brick break + backwards slope jump + landing on the slope
just before entering water (so that holding Z will give a water long jump) ended up 0.4s faster than the nearest challenger (backwards sliding).

In the last section, double jump would be faster but this would prevent being able to long jump later on (even if Z is held during the jump), so a long jump spin is much better.
For the long jump, you can barely clip into the corner of the ceiling, and this causes a weird 'push' out of the water's surface that lets you instantly long jump. This saved 0.8s over the normal water lj method, and 0.9s over a really cool backwards long jump.

For the star grab, a similar corner clip wasn't useful as water walls don't act like surfaces. Auto-ledgegrab is faster by 8 frames over backwards spinning out of the water.

Green Star 3 - 1:11.00 IGT - 41:58.10 real-time split (1:54.34 segment)

I was able to compare to a 2020 IL TAS I did with this one: youtu.be/FB2KuYX9OZo Saved 0.53s in the first section,
and another 0.91s in the cave.

The corner clip somehow conserves double jump state, giving a triple jump 'out of water'. This saves 0.3s over the strat done in 2020 (jump & spin onto the curved ground, backflip into the water).

The Goomba launch works the same as penguin launches.
To get it high enough for the clip-out to reach the surface, some A Button Challenge tech came in handy: Goomba raising. When stunned by a spin or starbit, Goombas recoil up into the air. Spamming them with starbits can gradually lift them up.
While this causes a lot of hitlag, it was faster than the strat used in 2020, and also meant the Paragoomba could be 2P frozen higher up to make the next trick quicker.
The only downside to the launch is that when it hits the water above, Luigi immediately dives down. Fortunately not too much for 2P to let itself go of the Paragoomba.

Because Goombas immediately explode upon touching water, a Goomba launch isn't possible with the Paragoomba (it only clips Luigi down), but it can clip Luigi into the wall, where a shore launch is possible.
This is enough to reach the star; holding neutral during the launch allows for a spin once Luigi has reached the peak.

#supermariogalaxy2 #toolassistedspeedrun

[TAS] SMG2: Cosmic Cove GS1 in 20.58s

PurpleSun 2022-09-30 | Green Stars TAS WIP
Yet another improvement, by 4.07s over this one here: youtu.be/YvSP9w-LQcY

20.58s IGT - 38:03.86 real-time split (1:16.51 segment)

Improvement - hard landing launch:
The strangest change from the previous one is that there is no land cancel or any jumping at all at the start. It looks like an ABC run. There's a short history to this.

From what we had understood before (described in this megashorelaunch video by username: youtu.be/bcCLTZ1MykQ?t=103 ), shore launches couldn't maintain a horizontal speed past 32 (the general speed cap) even though they exit at much higher speeds.
However, when making the ABC run (youtu.be/nHFTmux0Ae4), I was able to keep the initial horizontal speed of the launch to go flying across the planet.
As it turns out, this is because of something known in the ABC crew as hard or soft landing state.
Hard landing state is active after any stunned landing (intro fly-in, launch star, big fall) or exiting a pipe. Soft landing state comes from jumping, bouncing, or a small fall.
This state was only known to affect spin knockbacks, where you can only control the movement of knockback if you're in soft landing.
But it looks like it also affects this, as hard landing would get uncapped launches while soft landing capped them to 32, even with the same line-up to the launch.

And this is why there are no jumps, in order to keep hard landing and get the super fast launches (the second launch gets around 100 hspd).

Penguin launch:
There are two known tricks, one getting rebounded off stationary penguins when swimming or shell-riding into them, and another that is based on shore launching where you spin as the 'diving tutorial' penguin rockets up to the surface of water, which causes a huge 'dolphin spin' jump due to the added speed exiting the water: youtu.be/wkZYyZsjYG4

Putting these two together, you can get a launch off a stationary penguin, as spin swimming into it causes clipping which means a lot of speed in 1 frame (swimming in general can do this but spinning is necessary for the launch to work).

An alternative route was getting a huge launch off the 2nd penguin after a slow swim over (1st penguin launch would take just as long), which then lands back in water near the Cheep Cheeps to launch off one straight into the star. This was 2s slower than the double penguin route used.

#supermariogalaxy2 #toolassistedspeedrun

[ABC TAS] SMG2: Cosmic Cove GS1 in 0.5xA (+ P2 0.5xA)

PurpleSun 2022-09-22 | Mario and P2 attack a penguin and a fish for some flying miles to collect a Green Star in 0.5 A presses.

Both Mario and P2's 0.5 are only required for the Cheep Cheep, with P2 using it to freeze and spin the Cheep, and Mario needing it to clip inside without hitting or being hit.

This saves 4 A presses from the previous count, avoiding the ice waterfall walljumps by launching off the Cheep into the star.

Also in this video is a launch worthy of the title 'mega', somehow getting 129 XYZ initial speed (67 yspd, 110 hspd) off the penguin to go rocketing across the first stretch of the planet.
You can find more mega launches here: youtu.be/bcCLTZ1MykQ

#SuperMarioGalaxy #abuttonchallenge