PurpleSun | [TAS] SMG2: Cosmic Cove GS1 in 20.58s @PurpleSunTAS | Uploaded 2 years ago | Updated 1 hour ago
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
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
![[TAS] SMG2: Cosmic Cove GS1 in 20.58s](https://i.ytimg.com/vi/XMDBlu3_u48/hqdefault.jpg "[TAS] SMG2: Cosmic Cove GS1 in 20.58s
Green Stars TAS WIP
Yet another improvement, by 4.07s over this one here: https://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. Theres a short history to this.
From what we had understood before (described in this megashorelaunch video by username: https://youtu.be/bcCLTZ1MykQ?t=103 ), shore launches couldnt maintain a horizontal speed past 32 (the general speed cap) even though they exit at much higher speeds.
However, when making the ABC run (https://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 youre 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: https://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")
![[TAS] SMG2: Boom Bunker Green Stars](https://i.ytimg.com/vi/YABedH0_foE/hqdefault.jpg "[TAS] SMG2: Boom Bunker Green Stars
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: https://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 didnt 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] Perfect Run final section in 20.80s](https://i.ytimg.com/vi/YFjtFG5vy2o/hqdefault.jpg "[TAS] Perfect Run final section in 20.80s
20.80s from landing from the previous section, to exiting the launch star (camera shake frame).
Shadow64 asked about a theory that jumping from the sling star would get a lot of speed due to the tilt when landing, so I tested it, and it did, but there ended up being a lot more to it than that, making the entire section revolve around whether you fast single jump, double jump, or long jump at the start.
This is because of the long jump chain from the second set of steps onwards, as whatever you do before puts you in a different position and thus your chain will end up in different spots, requiring different levels of strain and even single jumps to make it flow.
Added on to that, the chain would need to be as far left as possible going towards the Boomerang Bros to manipulate all of them into a bunch on the right-hand side, while also having a rightwards angle to go straight into the Boomers.
After a lot of testing, it seems like double jumping with the speed gets the best possible long jump chain. It needs to strain the following two long jumps to land on the edge of the second square platform, and also needs to briefly go neutral from the Thwomp platform to avoid the extra momentum (which isnt enough to reach the next square platform). It also needs to wait for the rightwards long jump so that the Boomers have time to jump over.
Overall, this saves ~0.25s over sonicpackers TAS,
However the Boomer strat at the end is also faster than sonicpackers strat, and the long jump route benefits the strat with a long jump straight into the first Boomer. Overall, this section saved ~1.35s.
P2 could also save a bit more time if it can hit the Boomer at the back at the right angle to knock it off the edge - its hitting angle is based on Luigis position from the enemy.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG: Honeyhive Cosmic Mario Race: High Score - 8 Coins](https://i.ytimg.com/vi/YXx3Lw-YzEM/hqdefault.jpg "[TAS] SMG: Honeyhive Cosmic Mario Race: High Score - 8 Coins
HEYimHeroic (https://twitter.com/HEYimHeroic) asked if I could get as many coins as possible in Honeyhives cosmic clone race with both bros.
7 is the record, attained by 5 people on Cyberscore and a few more not on that site including Heroic herself.
But there is an eighth coin: a starbit plant atop the waterfall.
With the power of TAS strats, it is possible to get this coin and still win with some time to spare.
This spare time was spent slowly sliding down a wall, and moving about in honey.
Interestingly, if you touch the star on the frame the cosmic clone sits on it (5 frames after it has touched it), it counts as a successful star grab and prevents the clones victory cutscene. The game is literally rigged in your favour... by 5 frames.
Side note: because Cosmic Luigi is fast, the current best Luigi count is 7. When testing, I barely had enough time to even grab the 8th coin before Cosmic Luigi won.
#SuperMarioGalaxy")
![[TAS] Super Mario Galaxy 2: Beat Block Green Star 1 in 12.90s](https://i.ytimg.com/vi/YmJuFbggTjY/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Beat Block Green Star 1 in 12.90s
Luigi leaps off a crystal to collect a Green Star in 12.90 seconds [IGT, 774 frames].
This one had been done before GS2, but I had wanted to go back to see if I could get to a better point in the block cycle, as I had to slow the long jump onto the green block at 0:16.
Being able to go to the yellow block instead and then having the blocks change from there would be a tiny bit faster, however, this is not possible to get in a fast run.
The cycle always starts on fly-in cutscene (when the levels music starts), so delaying the intro wouldnt help, and in terms of skipping the fly-in cutscene, the earliest skip is 1.8s too late for the cycle, and the latest skip is 1.3s too early.
But this is the best possible outcome of what is available, and the strat of long jumping off the crystal (which is only possible by having enough horizontal speed to not end up with a slope jump) is 6 frames faster than an instant backflip off the ground (again, because of speed, trying to get a backflip off the crystal would get a slope jump instead).
Also, cutscene tests are done - it is 3 frames faster to do Green Star 1 first, as the cycle with the fly-in cutscene benefits it more than it does Green Star 2. Also, while it is slower than the IL in terms of in-game time, skipping the cutscenes straight away for Green Star 2 is much faster in real time.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Clockwork Ruins Green Star 3 in 1:24.51](https://i.ytimg.com/vi/Yp_QK_nZUa8/hqdefault.jpg "[TAS] SMG2: Clockwork Ruins Green Star 3 in 1:24.51
Luigi and P2 go on a temple run through the ruins to collect a Green Star in 1 minute 24.51 seconds [IGT, 5071 frames].
Apologies for the choppy framerate at parts - the framerate kept fluctuating when dumping frames. Need to learn to limit FPS to 15 or something.
Input display: https://youtu.be/lQepNs1deL0
This TAS isnt the greatest in optimisation as it is - the main issue of optimisation is in the first section. Both intro cutscenes affect the wheel cycles in different ways, the first one mostly changing the second half of wheels.
There is much to keep testing with judging the best balance of cutscene skips along with understanding movement across these wheels when on different cycles.
I started my testing by copying Nemis ideas (https://youtu.be/OzivXwVp3Rc?t=262) and then adjusting from there.
I also tried to adjust the skips and movement to reach the rising arm of the final wheel when it was level with the ? block, rather than having to jump to the middle, but barely missed out on it.
Interestingly, in terms of real time, GS1 is faster if you delay cutscenes compared to instant cutscene skips, whereas GS2 (and presumably GS3) is faster the other way around.
On the second planet, the backflip wallkick from the cage was 0.3s faster than backflip ledgegrab. Interestingly, the backflip ledgegrab up to the gravity arrow was 4 frames faster than backflip wallkick, because the latter strat would have to either land late with a weakspin to hit the arrow or wait for Master Luma to replenish the normal spin.
P2 cannot hit the arrow.
Long jumping and pressing the button midway through the ground pound animation ended up being 0.3s faster than a single jump with a normal ground pound hit, and still faster than a long jump-spin-ground pound, even if the spin is cancelled immediately to avoid a homing ground pound.
A textbox skip may be possible using a proxy off a Goomba/Jack OGoomba (a set up known to work elsewhere), but I have had no luck with it yet, and it would take a lot of starbit hitlag or waiting for a Goomba to run along to the Luma, potentially diminishing the timesave to 0.
On the third planet, delays between long jumps were necessary to align the long jumps past the final wheel.
Neutral wallkick strain was used to help get up to the top from a lower backflipping position.
I held forward (left) on the second-to-last pole to go straight into the last, saving a few frames over continuing to the right and wallkicking.
Long jump-backflip-wallkick was 0.4s faster than triple jump-wallkick.
P2 saved a small bit of time (0.25s) clearing the crates so Luigi could land right next to the elevator and activate it earlier.
The elevator trick is another point of low optimisation. From what I tested, the fastest method was getting a double wallkick off the extended wall (done by instant spin knockback back into the same wall, seen commonly in Honeybloom runs) as it meant the wallkicks were closer together compared to the spaced out angled wallkicks in speedruns.
But there could be better movement between wallkicks that changes that conclusion.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Cosmic Cove GS1 in 24.65s](https://i.ytimg.com/vi/YvSP9w-LQcY/hqdefault.jpg "[TAS] SMG2: Cosmic Cove GS1 in 24.65s
[TAS] SMG2: Cosmic Cove GS1 in 24.65s")
![[TAS] SMG2: Supermassive Green Stars](https://i.ytimg.com/vi/Z5AFvSz7VAs/hqdefault.jpg "[TAS] SMG2: Supermassive Green Stars
Green Stars TAS WIP
Star order matters due to the Thwomp cycles as will be explained later. Camera movement can manipulate the coins 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: https://youtu.be/0k-eeiC9gss
Zip jumping is used again to get a tight wallkick strat that was previously done with a long jump (https://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] Super Mario Galaxy: Gusty Garden Comet in 1:05.88](https://i.ytimg.com/vi/_Ivqaw5BtdI/hqdefault.jpg "[TAS] Super Mario Galaxy: Gusty Garden Comet in 1:05.88
Mario ruins Major Burrows career in a Daredevil run in 1 minutes 5.88 seconds [3953 frames].
This emulates a full-game run, watching the intro cutscenes that means it has a deficit to IL runs of ~11s.
The land cancel at the start had an interesting effect: the cutscene started 2 frames later but I gained control on the same frame as if I didnt land cancel. Doesnt save or lose time but its a thing. Allows for earlier movement but that doesnt make a difference overall.
As the RNG aspect of this fight is easily dealt with by finding where Burrows pops out and reloading a state to go straight to that spot, the key part of this run is getting the fastest hits.
(Also, if youre familiar with the RNG of when Burrows first pops out, which was a big thing in Boss Blitz TASing, in this star it is not really random since you always land on the planet at the same time)
The first two hits are done instantly by doing a spin cancel ground pound (Wiimote shake for 2 frames, press Z on the second frame) right next to Burrows as he pops out. This ground pound carries the hit properties of the spin, so as Mario touches Burrows when he pops out, it registers as a hit.
The third hit could do this, but you always have to do a second ground pound, so its slower to set up that instant hit.
Instead, I get the earliest second ground pound and starbit stun Burrows so he doesnt run away, while also jumping backwards and bouncing off the dirt mound to fling back into Burrows for the spin. This saved 8 or 9 frames over long jumping.
Potential improvement: For the first ground pound of the third phase, Burrows jumps directly away from you, no matter what path he was on, so we could try ground pounding to the side of him so he jumps into the tree and is closer for the hit.
Interestingly, ground pounding after the third hit and jumping just before the final cutscene was faster than doing nothing, as it let me move as soon as possible.
Spinning into the star saved 2 frames.
#ToolAssistedSpeedrun #SuperMarioGalaxy")
![[TAS] Super Mario Galaxy 2: Cloudy Court Green Star 2 in 19.10s](https://i.ytimg.com/vi/_K_sa1GU4DY/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Cloudy Court Green Star 2 in 19.10s
Luigi cloud long jumps from the drum to a Green Star in 19.10 seconds [IGT, 1146 frames].
The single jump at the start looked slow, but it sets up a long jump route that is a couple of frames faster than long jumping all the time, due to how long jumps become slightly faster and longer when wearing Cloud Suit. Less straining is needed to get to the edge of the square platform where the first cloud long jump is performed.
This cloud long jump is set up with a ground spin that is quickly cancelled by falling off the edge, which regains running speed sooner than letting a ground spin continue.
Forming a cloud on top of another was faster than jumping from two separate clouds as cloud long jumps stay at top speed after the peak, so there is higher speed for longer instead of cutting it short in the middle.
Normally, you get up to the hi-hat cymbal and cloud jump from there. However, you are able to do 3 cloud long jumps all the way to the Green Star from beside the topmost drum if you can get the highest long jump speed (accelerating up to 34 hspd) from the first cloud lj, which is something that continues to confuse me, as it tends not to work without some sort of movement (e.g. in Shiverburn GS3, I couldnt get it until single jumping beforehand, or holding a specific direction, and when testing jumps from the hi-hat for this star, I also had to single jump or long jump beforehand).
In this case, it seems that holding up from where I formed the cloud will always get the highest speed. Holding right doesnt seem to get it, which is strange as I am facing right when forming the cloud. Holding down doesnt either.
From there, getting the first frame of both the cloud spin cancel and the long jump for the last two cloud ljs results in much higher jumps than later inputs, enough to get up to the Green Star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG1: Hurry-Scurry in 49.43 [Mario]](https://i.ytimg.com/vi/_pNG6OP_DTs/hqdefault.jpg "[TAS] SMG1: Hurry-Scurry in 49.43 [Mario]
Mario skips the launch star to complete Shrinking Satellite in 49.43s [2966 frames].
This emulates a full-game run by watching cutscenes. This is a deficit of ~10s compared to ILs.
Yet another level showcasing Mario doing a skip that was previously thought to be Luigi-only. Although, in this case, Mario could do the skip, but it was thought to be 2P-only and slower than the normal route.
However, the launch star skip can be done without a P2 mid-air jump, and despite Mario landing later than Luigi does, it still turns out to be 0.3s faster than the normal route for the red plumber.
There might be a slightly better route with this later landing, but Luigis route still worked out.
There is room for improvement in the skip itself with earlier possible HGPs and better angles available, but Ill let The Boss optimise that :) (TAS playlist here: https://www.youtube.com/watch?v=CHIywL2labQ&list=PLX-UHl4lUQzuLyU9SC6AolZW4odODr5Wf )
#SuperMarioGalaxy #ToolAssistedSpeedrun")
![[TAS] SMG2: Boo Moon 1 in 0.5xA presses [Mario]](https://i.ytimg.com/vi/_wULtm4-R_g/hqdefault.jpg "[TAS] SMG2: Boo Moon 1 in 0.5xA presses [Mario]
This run hasnt improved the levels A press count, but it confirms that Mario is equal to Luigi here, meaning we can avoid the menuing A press for unlocking the Luigi ghost.
Potentially possible 0x, but getting below 1x is the important thing for now.
#SuperMarioGalaxy2 #AButtonChallenge")
