PurpleSun | [TAS] SMG2: Tall Trunk GS2 in 21.70s @PurpleSunTAS | Uploaded 1 year ago | Updated 5 minutes ago
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
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: Tall Trunk GS2 in 21.70s](https://i.ytimg.com/vi/A_dVxMJYqYo/hqdefault.jpg "[TAS] SMG2: Tall Trunk GS2 in 21.70s
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: https://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 doesnt work, particularly when landing on an edge; havent 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 isnt 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: https://twitter.com/PurpleSunSR/status/1596573885084102656
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] Super Mario Galaxy 2: Battle Belt Green Star 3 in 1:42.25](https://i.ytimg.com/vi/C1RfkOBAQ8o/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Battle Belt Green Star 3 in 1:42.25
Luigi and P2 beat up groups of enemies to collect a Green Star in 1 minute 42.25 seconds [IGT, 6135 frames].
I actually made this a little while ago, but irl work and the TAS competition and the ABC World 1 project shoved this to the side.
There is already a confirmed improvement to this, which is better shell movement for the leftwards Boos, saving at least 4 frames.
Also using starbits on the Goombas would be significantly better. I forgot about that.
At the start, it is actually slightly faster to go over the top of the planet to hit the mini Goomba, however, that route leaves the starbit out of the camera shot, so it cant be used for the next planet. Therefore, bottom route is faster in the long run.
On the second planet, I was able to P2 spin the farthest Lil Cinder and starbit hit it on the same frame, instantly breaking it.
For the next three planets (plus Shroomship), nemis crabber HGP (https://youtu.be/_hEkLhRqQyA?t=22) is a few seconds faster than taking the Fire Flower all the way to the pumpkin Goombas.
This is helped by being able to freeze the crabber in place rather than having to spin its guard up.
Getting a swish straight into the sling star was a bit faster than ground pounding once in the ships gravity.
The pumpkin Goomba route seems like it can be improved: the positions of the Goombas are affected by how quickly you do the first few planets.
Having P2 break the pumpkins of two Goombas was useful to throw the Goombas closer to each other, but it may actually be less useful for defeating them.
With the remaining pumpkin Goomba, I was able to perform a falling ground pound, where Luigi starts a ground pound at the same time as interacting with an NPC sphere; this then causes the push effect (seen with crystal floats) to force Luigi to fall while in the ground pound animation. [Disclaimer: This is just a theory]
This technique could potentially be faster on the other pumpkin Goombas, rather than dealing with them as regular Goombas, however I did completely forget there were starbits I couldve used.
I couldnt manage to long jump onto the edge of the pipe without missing it or bonking into it, so a single jump was faster.
As stated before, the Boo section could be slightly faster. Its a struggle to get this done quickly as, while turning, the shells light goes into 3D so any Boo that seems like its crossing the beam is actually off to the side on the Z-axis.
The improvement comes from braking earlier *after* turning around to have a tighter turning cycle to finish off the remaining Boos faster.
The first and third Lil Brrs were extinguished by using P2 spins, and then smashed by Luigi.
The second Lil Brr was extinguished with an invisible skating spin - I dont know why Luigi didnt twirl, but it was definitely a spin from Luigi that got the Lil Brr. P2 then smashed it just before it went out of shot.
Recently looking for improvements, I attempted to get the Undergrunts with a backwards bounce off the dirt and a second bounce into a ground pound - this hit them at exactly the same time as the normal strat that I used in the video.
I tried backwards strats while getting the key but they didnt turn out faster than a simple run, and then a long jump onto the edge of the pipe (this one worked out, why couldnt the other one? aaaaaaa)
P2 couldnt do anything useful with the Swaphoppers. Instead, I bounced on the second one as gravity changed, spun to have less airtime, and then long jumped around the top to the third. Long jumping around the bottom was only slower by a few frames.
I tried to backwards spin directly to the pipe after bouncing off the third hopper, but the spin always carried forwards momentum. It ended up being faster to ground pound and then backwards spin out of a jump.
I tried to do a falling ground pound too, but it didnt work out.
As speedrunners know, a backwards bounce to the sling star is fastest.
woooo 4000 character description
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Rightside Down GS1 in 45.28s](https://i.ytimg.com/vi/CdOdK9QOn4U/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Rightside Down GS1 in 45.28s
Luigi fireballs his way to a Green Star in 45.28 seconds [IGT, 2717 frames].
Many of the strats were adapted and altered from nemis IL run: https://youtu.be/OK u7dJL0Y
UPDATE: Yaasd suggested a backwards bounce at 0:28 that continues down below the lifting platforms, and its an improvement of 0.5s: https://youtu.be/3t_9nZJqyJk
For the fall down to the flower, I performed an instant wall release by grabbing the very bottom of the wall, so far down that the game thinks Luigi should slide off it instantly. This lets me keep the same backwards angle but resets falling speed so I can make it straight to the flower. This technique is also used in Honeyhive Cosmic races: https://youtu.be/tOT9CZDjyzc
Spinning after the backflip might improve the next part as, from what I remember, spinning has slightly faster backwards speed as a backflip, although too much speed would overshoot the double Goomba bouncing which is done for even more speed overall.
P2 is useful for holding the Paragoomba in place to bounce off it without having to spin.
The moving platform is really awkward because of the small gaps to fall through - holding neutral will automatically grab a ledge (but I dont remember testing up or down, believing it would have the same result because 2D)
So I decided to fall and hold left to wallslide off the platform and be turned around at the bottom. This meant I could fall backwards to the Thwomp section, using the other moving platform as a ceiling that stopped a spin going so high and landing later.
To get to the Fire Flower, I found a quick spin after a walljump on the left was slightly faster than walljumping from the right.
The fireball section itself could use some improvement; looking back at some of the movement, its very clunky.
However, I am satisfied with what happened between 0:44 and 0:47, with a sideflip and fireball angle change that interacted with the falling crate in such a way that Luigi grabbed and instantly climbed the ledge on the right. This allowed for a quick backflip to the gravity flip, although this might not be the best strat as it meant I had to throw two fireballs at the next pile of crates instead of just one, due to the lack of forward momentum with the fireball angle change that wasnt enough to long jump and escape the falling crates.
Finally, I triple jump off a crate and walljump off-screen, holding neutral as soon as Luigi jumps to bypass the restricted walljump speed (a trick known as neutral walljump strain), so I could get more forward speed by holding right, enough to be able to triple fireball into the star.
This strat saved 8 frames over the normal strat of entering 3D and long jumping back in. You might not be able to see whats going on but sometimes a little mystery is nice in a TAS.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s](https://i.ytimg.com/vi/D5V8ibOL_tc/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Rolling Coaster Green Star 1 in 22.95s
Luigi coasts to a Green Star in 22.95 seconds [1377 frames, IGT].
The backflip at the start saved 0.15s over long jumping and spinning.
Jumping over the railings ended up being a little bit faster, and jumping further along the rolling rail is also faster as it gets to the high-speed point earlier.
Down the slide, swinging between the sides ended up being ~0.2s faster than taking a more central route.
There isnt much to say about the last stretch other than a light left turn around the Galoomba to avoid slipping down, and then light right turning towards the star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Fleet Glide Green Star 2 in 53.75s](https://i.ytimg.com/vi/DA8-_G1KBlo/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Fleet Glide Green Star 2 in 53.75s
[TAS] Super Mario Galaxy 2: Fleet Glide Green Star 2 in 53.75s")
![[TAS] SMG2: Sky Station Green Stars](https://i.ytimg.com/vi/DFBO_RjseEA/hqdefault.jpg "[TAS] SMG2: Sky Station Green Stars
Green Stars TAS WIP
Started off by showing the pre-galaxy stuff, including going fast enough to the helm to beat the 2P textbox by 6 frames. Unfortunately you cant cancel the textbox by pressing the helm button at the same time.
GS1 - 7.28 IGT - 1:13.90 real-time split
With Japanese text and without the 2P textbox, this would be 5.93 (textbox loses 0.75s, English loses 0.6s).
Its faster to take this star with cutscenes than GS2, as skipping cutscenes for GS2 gives a better cycle on the moving platforms.
Taking the roof route was only 7 frames faster than going around.
The low long jump only saves 1 frame over a fruit bounce.
GS2 - 19.61 IGT - 2:17.10 real-time split (1:03.20 segment)
The IGT is slower than my IL TAS because, in the IL, the first cutscene skip is delayed to perfect the cycles for in-game time; this doesnt matter for real-time.
To be faster in real-time, the platforms are further manipulated by having slightly suboptimal lines (and a swerve) at the start of the 2nd planet to keep the platforms on camera so they keep moving along the cycle, the main platform rising up ~20f earlier than with optimal movement.
Delaying the first cutscene enough to make use of the first yellow platform (skipped with a spin here) is barely not enough to make this earlier real-time cycle, even with more optimal lines.
GS3 - 36.08 IGT - 3:48.63 real-time split (1:31.53 segment)
Backflip over the hill was 0.4s faster than long jumping around the side.
The camera is turned just before the launch star appears in order to minimise its movement during the mini-cutscene, which saved 2 frames.
Originally on the glass planet, I landed on the checkpoint, backflipped up, and long jumped through and over the Octoomba to get the same long jump from 1st to 3rd platform. But the cycle wasnt great and I lost speed on the 3rd platform.
So this other route (which had been thought up a couple years ago) that seemed slower at the start ended up being faster by 0.41s, as it kept running speed after the long jump from platform to platform.
Also, while Im using a pre-made 9999 coins file by bloodtv_ that means collecting coins doesnt matter, I decided to grab the coin with P2 just because.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Sky Station Green Star 1 in 3.06s](https://i.ytimg.com/vi/DFcbN5cvWLw/hqdefault.jpg "[TAS] SMG2: Sky Station Green Star 1 in 3.06s
Luigi nabs the fastest Green Star in 3.06 seconds [IGT, 184 frames].
This is a 1f improvement over the previous 3.08 by myself in 2019: https://youtu.be/b-yFYZLkmps
The frame save is embarrassingly simple. By keeping enough speed while straining the second long jump further to the right (and left once the camera has turned), Luigi hits a higher point of the stars hitbox.
I had tried various lines in my previous TAS but none had improved; the difference was not losing too much speed so it doesnt drop in front of the star before hitting the top spot. Its all spheres. Tears for Spheres.
The low long jump is caused by long jumping off a ledge where the tilt hasnt been calculated properly, reducing the yspd (9 units initial speed, versus the normal 15). Unsure if its to do with edges in general, or specifically a slanted edge.
This strat saves 1 frame over the fruit bounce strat.
Could there be more frames to save? Yeah, I think so. Theres a way to get an even lower long jump, although its horizontal speed is low too, so a following long jump hasnt been able to get over the edge without a wayward line. Currently sits at 3.13s despite a 4f earlier 2nd long jump.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Sky Station GS3 1xA full game route](https://i.ytimg.com/vi/Ey9akCDygYo/hqdefault.jpg "[TAS] SMG2: Sky Station GS3 1xA full game route
The A press is used to start the Lumas cutscene, and is done in such a way to instantly end the cutscene, resulting in an unnecessary jump from a necessary A press.
At the start, I used Mooshs strat from his run of 0x gameplay A presses (i.e. excluding dialogue): https://youtu.be/-zCuJPmgL_A
The only difference is the angle and trajectory of the triple bounce, as Moosh had a better angle that bounced off the rock and landed closer to the Chomp, whereas I found that my angle could bounce off a green bulb that a Chomp pushed out. Probably slower but I like the variety.
In theory, the textbox skip works because you stand on the edge of the Lumas trigger area, and then get pushed off without touching the joystick (in this case crouching and uncrouching with Z) and press A just before leaving the trigger area. This A press starts the dialogue cutscene (evident by the camera zoom), but as Mario then leaves the area, it counts as ending the cutscene. In this case, ending the cutscene triggers the Luma to transform into a launch star.
There are other NPCs such as the Gold and Silver Gearmos from the Cmere levels where you can do this technique, but nothing happens aside from the camera zoom. The theory there is that because the progression from talking to the Gearmo (spawning the cage) happens *during* the cutscene, instantly ending the cutscene skips that and so it has no result.
As for the final planet, while landing on the higher platform before the Octoomba reaches the Octoomba 20 frames earlier than the lower platform - and also is fast enough to backwards bounce past the first elevator - it ends up getting a slower cycle on the third elevator, ultimately being 0.3s (24f) slower.
Taking the low route, its necessary to get an electric shock in mid-air to get up to the Octoomba, as spinning from the ledge is barely too short and I couldnt manage to get a ledgegrab.
The second electric shock wasnt necessary as I could have just gone down at the second elevator and spun across, but I did it because of the accidental bouncy slide across the second elevator.
On the final elevator, I turned around and got clipped into the ceiling before getting pushed out with a walking animation, and finally a spin to the star.
#SuperMarioGalaxy2 #AButtonChallenge")
![[TAS] Super Mario Galaxy 2: Cosmic Cove Green Star 3 in 1:21.61](https://i.ytimg.com/vi/FB2KuYX9OZo/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Cosmic Cove Green Star 3 in 1:21.61
Luigi and P2 jump out of water like those fish that jump out of water to collect a Green Star in 1 minute 21.61 seconds [IGT, 4897 frames].
The improvement here was suggested by Really_Tall, going around the water circle in the final section instead of swimming through it.
A regular water jump barely cant make it to the water above (going to where the 1-up is) and water long jumping further right takes too long overall.
But jumping back onto the solid part of the circle and backflipping up gets to the Goombas platform very quickly, and a backflip from there reaches the trigger for the water blocks as early as possible, earlier than bouncing off the Goomba or water jumping from above.
This saved 1.33s over the previous run, and saves ~2.3s over the normal RTA method of going through the circle and backflipping off the top of the Goombas platform.
The water here is strange - aside from the bottom part, entering with a jump or spin will not interact with the water until you have reached the peak of the jump/spin, so you can travel quite far into it. A jump + a spin before the peak goes even further.
[from previous run]
The first area could use some improvement. At a certain point, single skate jumps and double skate jumps have more speed than a skate long jump (ljs have an average speed of ~26 units per frame, jumps have ~27 units per frame over a shorter distance). Not sure at what point this extra speed occurs as trying skate jumps from the beginning had nowhere near that speed, and even after one lj, a single jump didnt get that speed.
The second area is where P2 shines. Breaking the bricks as P2 saves 0.3s over ground pounding, bursting the bubble saves 0.4s over trying to avoid it, breaking the box and shrinking the spiny at the start of the underwater section saved a few frames from a better line and no break stutter (Luigi stops when breaking a box underwater), and hitting the lever saves 2.5s.
This is where the first water long jump is; by holding Z when entering water and then picking up enough speed at the surface, you can long jump out of water.
Long jumping early and following with a single jump onto the used block saved 0.5s over long jumping later straight onto the used block.
The final section gets interesting. The moving water blocks at the top only start moving once you pass a point somewhere just above the big blue arrow, so optimisation is only really required in the first half - the second half doesnt have to be perfect as long as you make the cycle to get the earliest possible long jump to the star (when the third block is slightly raised).
[old ideas removed, see here: https://youtu.be/6-UTeD07tqI ]
The rest of the section was simple, as noted above, and involved setting up another Z-held water entry to get two water long jumps, one to the second block, and the final one to the star.
A weak spin was the only way to grab the star from that early of a long jump, so it saved a couple of frames.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[ABC TAS] SMG1: Good Egg 2 in 0x A presses [Luigi]](https://i.ytimg.com/vi/FNFdeW4mYBM/hqdefault.jpg "[ABC TAS] SMG1: Good Egg 2 in 0x A presses [Luigi]
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 (https://youtu.be/90QX5QEvo-s), I decided to try with just a Goomba bounce. As Mario, this is much harder and hasnt 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 https://youtu.be/XBWQlODVYKg
The star grab hadnt been proven without holding A (0.5) but was considered most likely possible and here it is.
#supermariogalaxy #abuttonchallenge")
![[TAS] Super Mario Galaxy 2: Slipsand Green Star 3 in 1:14.10 (outdated)](https://i.ytimg.com/vi/FQ7dW_1L-vU/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Slipsand Green Star 3 in 1:14.10 (outdated)
Luigi launches off a Romp to collect a Green Star in 1 minute 14.10 seconds [IGT, 4446 frames].
So this isnt exactly optimised but I had to show off this crazy new strat that happened completely accidentally. UPDATE: Romp launch has been optimised some more but still needs work: https://youtu.be/xsd_7sIqS90
Firstly, I mostly followed nemis IL (https://youtu.be/Wy2RYwIa-Ck?t=139) throughout the level, improving the first planet with some tighter long jumps and an earlier upwards strain towards the launch star.
I also used P2 to freeze a couple of troublesome fireballs.
For the second planet, I long jumped from the edges at 0:37 and 0:42, as edges of moving/scrolling floors do not have any friction, so long jumps are normal instead of being pulled back slightly.
This could save time for other long jumps here but the long jump route played out perfectly to bounce off one of the Pokey heads onto the block.
I tried long jumping earlier off the block to get horizontal displacement momentum (which happens sometimes when long jumping off a moving platform) with a forward angle, but this only went sideways, and walljumping off the closest wall is slower than angling to the right and jumping off the left wall.
For the slide, I started out doing the known strat of getting the Romp to push Luigi by sliding up against the edge, which wont get him crushed.
Whenever Luigi would start getting pushed to the side or start clipping into the floor, I tried jumping at different times with different angles and time spent holding A in order to stay in front of the Romp but above ground.
One time, though, this ended up with jumping at a certain point where Luigi had clipped slightly inside the Romp and floor, so the game had ejected him by ~170 units. The jump then conserved most of this speed, going at a hspd of ~130 units/frame.
Assuming units are centimetres, this would be at least 174 miles per hour. This speed was constant as speed cant be changed during a slope jump in SMG2.
Angle also cant be changed from slope jumps, so I had to deal with going to the left. I timed a ground pound to land up-left of the star, so I could move down and right at a faster slide speed to reach the star.
This strat was 3 seconds faster than getting Romp pushed all the way.
This could obviously be improved by just getting the launch earlier, but could also be improved if we can figure out how to jump to the right, which may simply be a case of getting pushed on the left edge of a Romp so that we have to angle to the right to keep getting pushed.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
