![[TAS] SMG2: Supermassive GS3 in 17.43s @PurpleSunTAS](https://i.ytimg.com/vi/REXqfA8tcDY/maxresdefault.jpg)
Uploaded April 2024 | Updated September 2026, 2 weeks ago
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.
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.
![[TAS] SMG2: Supermassive GS3 in 17.43s](https://i.ytimg.com/vi/REXqfA8tcDY/mqdefault.jpg "[TAS] SMG2: Supermassive GS3 in 17.43s
A new improvement of 1.03s (62 frames) over my Greens WIP from last year: https://youtu.be/Z5AFvSz7VAs?t=74
One to add to known improvements I guess... not sure why I didnt think of trying the Goomba float before. Maybe because I was locked in on the coin zip jump, and didnt think that floating from this far back would work.
But the question of it being faster was brought up in Jhays Discord server after the A press save strat was shared. So Ive tested it now.
This isnt hyper-optimised as well so could be a few more frames to save.")
![[TAS] SMG2: Haunty Halls Green Stars](https://i.ytimg.com/vi/RWiF9dD1RKQ/mqdefault.jpg "[TAS] SMG2: Haunty Halls Green Stars
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 wont 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 wont go back.
Green Star 2 - 11.83s IGT - 1:05:06.25 real-time split (55.19s segment)
4f improvement over my 2020 TAS: https://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 dont have the fixed speed and angle of regular ones which means youll lose speed unless moving backwards.
Fortunately, its 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, Luigis 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 theres 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 cant 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")
![SMG2: Twisty Trials Comet using only the analog stick [TAS]](https://i.ytimg.com/vi/RhiznriYgg4/mqdefault.jpg "SMG2: Twisty Trials Comet using only the analog stick [TAS]
This is tool-assisted.
Turning Turning Double Time completed without pressing a button or shaking the controllers, only using the Nunchuks joystick.
This is currently the only regular (yellow) star to be proven pure joystick only. There are a few Greens that are possible, and there are other yellow stars, such as Hot-Stepping Dash Pepper, that can be beaten joystick only if you consider shaking for launch stars as a meta control.
I had tried the normal Twisty star before, and got to the end before finding that the star was out of reach. But this star is nicely placed above the final block.
There were a couple of hurdles along the way where I needed a certain cycle to be able to fall onto a turning block and run up it before Luigi would start slipping down.
But the major hurdle was right at the end, and the solution came by accident - if you run up the very edge of a block, you can get off it and quickly land back on, and being off it means youre not getting dragged down with the turning surface, so you get a bit more height, and a bit more speed.
This was necessary to get up the final block, and particularly right at the end of the climb, I used this to get in the air and curve over the turning corner onto safe ground.
also this doubles up as a 0xA vid and a 0 spins vid.
smh purple sun u pressed a to sleect starrr!!.
#SuperMarioGalaxy2")
![[TAS, outdated] SMG2: Flip-Swap GS1 in 6.06s](https://i.ytimg.com/vi/RikJY-wwPDk/mqdefault.jpg "[TAS, outdated] SMG2: Flip-Swap GS1 in 6.06s
obsoleted by 6 frames in this TAS: https://youtu.be/fnTWLzyNF_M
Luigi hyperjumps into a Green star in 6.06 seconds [IGT, 364 frames].
-1f from sonicpackers 2012 TAS: https://youtu.be/WgZGmVK3I3I
Crouching at the start moves slightly forward and accelerates faster, saving 2 frames on the starting run-up.
The frame of improvement comes from long jumping later to spin cancel jump immediately, jumping 1f earlier than sonicpacker and still getting the best flip panel jump all the way into the star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Boo Moon Green Stars](https://i.ytimg.com/vi/Rk994rMgs0A/mqdefault.jpg "[TAS] SMG2: Boo Moon Green Stars
Green Stars TAS WIP
Star order doesnt 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 isnt 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 Luigis 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 dont count as moving platforms, so theres 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 isnt necessary due to maxed file, but it didnt lose any frames and is a good showcase for 242.
Squeezed backflip wasnt faster here due to the Boos bad positioning.
Double long jump + floaty airspin cancel was 8 frames faster than a long jump + single jump. Its also possible (27f slower) to ground pound the moon and uncrouch spin to grab the star without falling off: https://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 its important to get on the third snake block as soon as possible, as thats 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: Lava Lair Green Star 2 in 0xA presses [Mario]](https://i.ytimg.com/vi/RkDFGqmr6K4/mqdefault.jpg "[TAS] SMG2: Lava Lair Green Star 2 in 0xA presses [Mario]
Mario waits for an opening to grab a Green star in 0 A presses.
I had P2 on as this might be part of a future World 2 run, following the World 1 run made by Billy, Moosh and I early last year: https://youtu.be/HqSIIo-RUV0
In SMG2, unlike SMG1, you can grab pull stars with B, but only if youre not holding A.
With Flomps, you have some control over your velocity when thrown off and can go huge distances when moving the same way you get thrown. Its a shame they cant throw in any direction, couldve been useful for the Speed Run Comet.
In something Im calling edge grinding (or tail grinding), its faster to get that stumbling to fall state when running on the edge of a floor rather than simply running, although in the part I mostly used it, the shorter distance of running across and around the firebars would likely have been faster.
#SuperMarioGalaxy2 #AButtonChallenge")
![[TAS] SMG2: Bowser Jr.s Fiery Flotilla Green Stars](https://i.ytimg.com/vi/RlxbpD0l4wQ/mqdefault.jpg "[TAS] SMG2: Bowser Jr.s Fiery Flotilla Green Stars
Green Stars TAS WIP
Star order is based on the Thwomp cycle. After cutscenes, its cycle is pretty good, but without cutscenes, its cycle is slower and alternative strats to the top are faster for both stars.
For GS2 with cutscenes, the Thwomp jump is faster than the alt strat; GS1 benefits much more from its own strat. The difference is 0.8s. Here is GS2s alternative (and the comment that gave me GS1s strat): https://youtu.be/qmK9BPKPkCs
Green Star 2 - 16.58 IGT - 21:19.88 real-time split (1:12.72 segment)
The movement to the Thwomp didnt need to be optimised due to the cycle. I tried some cool ideas to fill the time but they took a little too long. Jumping on the Thwomps eyebrow was neat, though.
There seemed to be an improvement with a better long jump angle to reach the next planets gravity 1 frame earlier, but it didnt translate into overall timesave.
For the star grab, the goal was to lavaboost up into the star rather than loop over it. The long jump was delayed right to the edge, and for a bit of extra distance, I did an airspin cancel. This involves pressing Z while spinning close to the ground, which prevents a ground pound and results in either a low and fast spin (frame-perfect Z press) or a slow and floaty fall state.
Getting this lavaboost up into the star saved 8 frames over the RTA strat of long jumping earlier and dropping onto the star.
Green Star 1 - 14.68 IGT - 22:17.90 real-time split (58.02 segment)
The alternative strat involves a backflip wallkick from the Bowser Jr. statue.
The long jump to the purple wall needed to get tilt off the railing and needed to be low to account for the gravity turn, as going lower into the turn ends up faster.
The Magikoopa was hit by a starbit so it would skip its attack, preventing it from shooting a spell at the statue which would have exploded and caused lag.
I had wanted to get the wallkick to go at a steeper angle, but it wouldve bumped against the small overhang of the castle, pushing Luigi off to the side.
Its hardly noticeable, but the first long jump lands on the edge and starts an edge grind, which had slightly better hspd and a better turning circle than landing safely, and also resulted in running speed long jumps.
22:17.90 World 1 segment
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Honeyhop Green Stars](https://i.ytimg.com/vi/Ro_DJLzG_Rs/mqdefault.jpg "[TAS] SMG2: Honeyhop Green Stars
Green Stars TAS WIP
Star order doesnt 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: https://youtu.be/0YeVYsjCiX8
Left side beeless climb (nemis IL route: https://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. Its 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: https://youtu.be/KG3QC0CJGzQ
Much the same as GS2 until the top of the 2D section.
Because it can keep enough speed to long jump, its 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: Flip-Swap Green Star 2 in 0xA Presses](https://i.ytimg.com/vi/S1yB5rzsb3k/mqdefault.jpg "[TAS] SMG2: Flip-Swap Green Star 2 in 0xA Presses
Mario floats to a Green Star without jumping.
This saves 1xA over the previous count, doing a crystal float to reach the star.
I had managed this in 0.5xA with Luigi, but it is possible 0x by using the electric beams to recharge spins in mid-air, enough to be able to get onto the crystal.
Crystal floats are confusing but the theory is that Mario interacts with the push/NPC sphere inside a crystal while on top of it. This sphere messes around with gravity - usually causing Marios falling speed to be very slow - and by going backwards over a certain area of the crystal, the axis of the gravity is reversed, so Mario ends up falling upwards.
On flat gravity, floats are very powerful, and for some reason going straight backwards causes them to cut off quickly, but holding a sideways angle makes them go further - interestingly, infinite flutters go faster sideways. Maybe something related?
Ground pounds are also affected, causing Mario to go forward at a fixed speed and only actually pounding down once the effect of the float is finished.
I selected GS1 because I wanted to try some stuff for that too.
P2 was chilling, making it easier for BillyWAR to put this in an upcoming project...
#SuperMarioGalaxy2 #AButtonChallenge")
![[ABC TAS] SMG: Freezeflame Comet in 0xA presses [Luigi]](https://i.ytimg.com/vi/SQI6vXVXT0I/mqdefault.jpg "[ABC TAS] SMG: Freezeflame Comet in 0xA presses [Luigi]
Luigi outskates the Cosmic clone in 0xA presses.
Shoutouts to BillyWAR for suggesting this. Hes currently going through SMG1 making videos for every star and saving more A presses along the way (including the coolest flipbug raising strat in Gusty Garden 3): https://www.youtube.com/channel/UCBCqPFqPz2OUFOFdUZ9Bkyg
The layout of the level is fine for not jumping (besides a trick needed for the last gravity flip) but the issue is being fast enough to beat the clone.
Spinning immediately in the air tended to get the fastest landing, although late spin cancels were very close when testing.
Spin skating has a trick where you can spin once just before the Luma fully recharges and then you dont have to wait as long for another spin (but you do have to wait a little bit).
This can be further improved by cancelling this spin by cutting corners, as then the next spin can be almost immediate.
I couldnt replicate it anywhere else, but one landing was on the edge and some jerky analog movements let me fall off and land back on in an instant, which let me spin skate early.
For the last gravity flip, it isnt possible to reach the gravity with a normal spin, so edge lift is required.
This is where you land on the edge of a ledge so that you dont get snapped to the floor, and you can then spin from this higher landing, which can barely reach the gravity.
These tricks meant I could be fast enough to catch up to the clone at the end and bounce off it into the star.
oh hey also its 10 years since i joined youtube
#supermariogalaxy #abuttonchallenge")
![[TAS] Super Mario Galaxy 2: Haunty Halls Green Star 2 in 11.90s](https://i.ytimg.com/vi/Sy6O893-okk/mqdefault.jpg "[TAS] Super Mario Galaxy 2: Haunty Halls Green Star 2 in 11.90s
Luigi backflips off a bump in the path to a Green Star in 11.90 seconds [IGT, 714 frames]
The camera pan at the start moves the second planet off-screen, stopping the paths cycles for a moment. This was helpful as it meant the path to the star needed less meandering, so I could backflip closer to the star.
The camera cancel on the second planet is frame-perfect, pressing up on the D-pad on the first landing (dust) frame. A jump cancel is also possible here, which is also frame-perfect.
It was 10 frames faster to land on the bump and backflip, rather than continue the long jump closer to the star but landing on lower ground.
I had had a cooler idea for the star grab involving the broken walls but it ended up over 1 second slower. Here it is on my second channel: https://youtu.be/TccadovLC7c
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
