PurpleSun | [TAS] SMG2: Chompworks Green Stars @PurpleSunTAS | Uploaded 1 year ago | Updated 1 hour ago
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
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: Chompworks Green Stars](https://i.ytimg.com/vi/c0NwVag22VU/hqdefault.jpg "[TAS] SMG2: Chompworks Green Stars
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 (https://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, Ive realised I forgot to optimise the launch star cutscene by minimising its camera movement, so theres 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 slopes 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, youre 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 momentumd 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 (https://youtu.be/WM5o2eKxPVE).
Zip jump route was 0.5s faster than long jump route (ILs 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 ( https://twitter.com/PurpleSunSR/status/1657853806787403778 ) which could save 0.5s, but the screw cycle doesnt work out.
birthday upload :)
#toolassistedspeedrun #supermariogalaxy2")
![[TAS demo] SMG2: Sky Station Comet with 40 Coins](https://i.ytimg.com/vi/cGb6WovjNVo/hqdefault.jpg "[TAS demo] SMG2: Sky Station Comet with 40 Coins
Suggested by Vallu, it is possible to get 40 coins in Sky Station Comet while beating the time limit.
This isnt optimised to a high level.
The bottleneck here is the first planet. 24 coins needed to be collected in roughly 31 seconds (it takes about 9 seconds from entering the launch star to grabbing the first clock on the next planet).
Without 2P, I managed to get to being only 3 seconds away from grabbing the next clock, but I couldnt see further optimisation doing enough to make up the deficit.
After the top side, I did a long jump around. Sometimes when jumping through a transition to upside-down gravity, Mario/Luigi will suddenly change facing angle mid-air. This is commonly seen in speedruns of Flipsville.
I used the now backwards speed to get a faster bounce up to the clock.
I found when long jumping off the crystal, Luigi did another slight angle change in mid-air. Its not as noticeable at full speed, but it did make a difference to speed in the air.
The long slow tunnel has coins and you cant ground pound to make it faster. Bad.
The level is free once youre on Peewees planet, with so many clocks around and only 3 coins plus Peewee hits to do.
#SuperMarioGalaxy2")
![[TAS] SMG1: Honeyhive Cosmic Mario Race in 37.56s](https://i.ytimg.com/vi/cePg_oAbkm0/hqdefault.jpg "[TAS] SMG1: Honeyhive Cosmic Mario Race in 37.56s
Mario picks up Luigis strats to beat Cosmic Mario in 37.56 seconds [2254 frames].
This emulates a full-game run, so it has cutscenes - a deficit of 10.45s to IL runs.
This is sort of a LOTAD but isnt really a low standard of optimisation so a MOTAD I guess. This level as Mario has been in the works for some time, mainly trying to make Luigis skip work.
The double jump strat at the start was found by FireBreather and saves 2 frames over walking (which is faster than long jumping as the trigger is on the ground so floaty long jumps arent useful).
The race boost conservation involves pressing A on the frame of GO! for the fastest boost, letting go of A and Z 1 frame and pressing A the next. This speed is almost twice as fast as a long jump, and high speeds continue through the following jumps.
The fourth jump would be roughly the same as a long jump, however, I landed on the very edge of the ledge, which meant Mario got warped forward slightly (known as an edge boost, found by Xander), giving an extra burst of speed to conserve with a jump.
This jump might not be incredibly useful in a more optimised setting, though, as it doesnt get a good strain towards the eventual spot for the big skip. A long jump helps along the way.
The skip is done by falling backwards and grabbing the very bottom of a wall. This causes Mario to slip straight off, but he doesnt change his original facing angle as usual. This wallgrab cuts off horizontal speed but more importantly vertical falling speed.
The horizontal speed is regained faster than the falling speed, which makes it barely possible to spin and land on the ledge below (note: trying to spin without the wallgrab hasnt even managed to grab the wall below yet, so the wallgrab, as far as we know, is necessary. and its also faster anyways).
A 2P jump on the ledge into a quick spin was the fastest method for grabbing the star, due to the quick vertical speed from the jump and the quick backwards speed from the spin.
#SuperMarioGalaxy #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: The Perfect Run in 1:44.30](https://i.ytimg.com/vi/dDCz6NtXlg8/hqdefault.jpg "[TAS] Super Mario Galaxy 2: The Perfect Run in 1:44.30
Mario, Yoshi and P2 make the perfect run in 1 minute 44.30 seconds [IGT, 6258 frames].
This is dedicated to diamond mentos and nemi.
I was planning to go back to this at some point for sub 2, but Nemi recently put out a 1:50.68 IL run - going way beyond my expectations - finding new fluttering tech for the first two sections to gain high speed while going backwards:
1:50.68 IL by nemi43: https://youtu.be/5fDqJ5cHMNQ?t=132
This motivated me to come back sooner, using Nemis IL as the groundwork for improving much of my previous demo which you can find here (2:00.38 time): https://youtu.be/iB0gigyiIo0
Mario is faster as he can break the egg and mount Yoshi 9 frames earlier than Luigi. Luigis dismount speed may be faster but it makes very little difference compared to this.
The new tech involves Yoshi spinning and then fluttering close to the ground, which picks up lots of speed, and it seems the more ground you cover, the faster you go (up to a certain point).
I start by going left a little bit. I couldnt find a way to spin early so I had to wait for the jump to dip down and then spin.
I could spin early for the second flutter though, and gained up to 32 hspd - nearly the top speed of a sideways flutter.
P2 and the single starbit helped to clear the way and flutter to the launch star, saving 13 frames over going for the last flower swing.
To set up the remount, I aimed the first flutter just in front of the back row of flipswitches to cover more ground for higher speed, and also off to the right as the target launch star is to the right.
Unlike sideways flutter speed, backwards flutter speed is consistent, so I could focus on getting the most height for the first set of backwards flutters, and then use lowering rapid flutters for more speed in the second sideways set.
The dismount was on the frame before hspd gets set to 0 for spending too long in the air. As the flutter speed was too fast for Mario to catch up to Yoshi in the air, I held down for 5 frames to slow Yoshi down before dismounting, at a speed where I could ground pound quickly and not lose too much height.
The interesting thing about optimising the second flutter is that the launch star is in a 2D section: when entering 2D from the front, the game slowly drags you into the middle no matter your speed. This means you have to flutter a few times before you can enter the launch star.
Because of this, optimisation becomes more about entering the section as soon as possible, but still able to reach the launch star at the earliest point.
I aimed for the last corner of the electric beams, as this was the straightest possible flutter angle without going too far away from the launch star.
In the flip-swap section, bouncing off the Octopuses let me change flutter direction so I could optimise for straighter angles, with a short rightwards flutter between the second and third Octopuses followed by holding straight up while falling.
I fluttered just before bouncing to get the bounce while still continuing to flutter, keeping more speed than bouncing (bouncing stays at 32, fluttering could quickly go up to 34 by resetting the flutter).
Fluttering into the pull star section, I bounce off Toad and hold the bounce all the way, as this delays the air timer that eventually sets hspd to 0.
Going under the obstacles was 2 frames faster than going over, using rapid flutters that keep the 34spd throughout and then two longer 32spd flutters to get up to the launch star.
In the Bros section, I set up a very slight downwards angle before fluttering to get a small curve throughout the flutters that lined up with the first and last Hammer Bros.
Flutter bouncing off the first Hammer Bro wasnt any faster as, at that point, Yoshi is just accelerating and acceleration isnt different between fluttering and bouncing. The bounce was also useful for getting higher up to be able to do faster lower flutters later on.
Reaching the last Hammer Bro, I had P2 hit the far Boomerang Bro off the platform, and Yoshi tongue the closest Bro to stun it.
Just before bouncing off the first Bro, Yoshi could tongue the second one.
I flutter bounced off the first Bro to avoid getting all the height of a triple bounce, so I could quickly bounce off the second Bro.
I spun when bouncing off the second Bro to avoid getting loads of backwards momentum away from the launch star.
Finally, bringing Yoshi to the end is 1 frame faster than ditching him beforehand, due to a fast backwards flutter and dismount, leaving Yoshi with Rosalina.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Flip-Swap Comet in 0xA [Mario]](https://i.ytimg.com/vi/dHaEupwsm0k/hqdefault.jpg "[TAS] SMG2: Flip-Swap Comet in 0xA [Mario]
Purple Coin Flip n Sprint without pressing A.
This saves 1xA over the previous count, avoiding jumping into the star by using flip panel displacement to reach the star.
Most of this route and these tricks come from Mooshs 1xA TAS: https://youtu.be/_wF89JD2zqU
The flip panel flings in particular were something I looked to implement, theyre just so cool.
Also in this run are spin cancel skids, which are skids that cancel the spin animation but carry spin properties, so flip panels still activate.
As a brief explanation, flip panel displacement happens by sliding off a panel (before it reaches the top), storing momentum and a link to the panel, and then spinning to activate the panel and grabbing a wall while it is moving. The wallgrab releases the stored momentum, while the movement of the stored panel vertically displaces Mario.
I first saw flip panel displacement from LancerDeltarune; random people on Twitter have said they found it many years ago but I cant properly credit that as a source: https://clips.twitch.tv/ScaryTacitWitchPanicBasket
P2 is hanging out ready for BillyWAR and Mooshs World 1 project. I couldnt use P2 to get purple coins as I would have to let go of A, so then Id have to press A again for P2 to be useful again. I couldve used P2 to freeze the electric fields (it would have let go automatically when off-screen) but I wasnt too bothered about that.
Its just a theory but the ledgegrab on the panel at the end seemed to reset stored displacement so I was able to store it from the slide afterwards.
This idea came about because I had tried many times to get displacement from sliding off the panel with no result. I even slope jumped off the panel, which is a sure-fire way to get displacement, but that didnt work.
I tried going back through the run to find out if anything I did would affect how displacement was stored, savestating from certain points and then jumping through the level.
I couldnt pinpoint anything, but found that none of the slides I had done through the run seemed to affect displacement at the end, and also werent what was stopping it from happening the first time around.
I eventually gave up on that research and just messed around at the end.
I got the sliding ledgegrab (with the weird tilted walk after climbing) and decided to try to activate displacement with a sliver of hope, and got a tiny upwarp. And then I found I could get displacement from a slide-off. Just because of a sliding ledgegrab?
Its bizarre and Im definitely missing something but, hey, it worked in the end so whatever.
#SuperMarioGalaxy2 #AButtonChallenge")
![[ABC TAS] SMG2: Clockwork Ruins 1 in 1xA (4xA) presses](https://i.ytimg.com/vi/dK0Dxuk48bA/hqdefault.jpg "[ABC TAS] SMG2: Clockwork Ruins 1 in 1xA (4xA) presses
Mario has adventure time for an adventure time for in 1 A press (4 including dialogue).
The A press is required to long jump across the gap between the spikes and the walljump wheels.
A lot of this is messing around waiting for cycles. I waited bouncing on the pumpkin goomba on the second planet because, while I would have made the second safe zone in time, I wouldnt have made the next cycle in time to save any time (for adventure).
The Luma takes up 3 dialogue A presses. There is a theory to save 2xA here by using a Goomba to push Mario out of the Lumas trigger area while pressing A, which should get a textbox skip, but it hasnt been successful yet.
A theory to save the final gameplay A press is to get slope displacement off a moving wheel at the start of the final planet, and then, before it disappears, grab a wall to either get displaced upwards to the top or sideways to the walljump wheels.
Mario can let go of poles by shaking the controller.
P.S. thanks for 1k subs!
#supermariogalaxy2 #AButtonChallenge")
![[TAS] SMG: Gold Leaf Comet in 2.5x A Presses [Luigi]](https://i.ytimg.com/vi/dkFnjZSy8vk/hqdefault.jpg "[TAS] SMG: Gold Leaf Comet in 2.5x A Presses [Luigi]
Luigi beats Cosmic Luigi in 2.5 A Presses.
This saves 0.5x over the previous count, avoiding having to jump up to the big tree by getting on the wooden fence, spinning onto the mushrooms, and spinning up to the ledge.
Technically, the star is possible to reach in at least 0.5xA just by spinning up the intended path, but the cosmic clone is too fast to beat that way, and is still too fast to beat with a couple of jumps along that way, so instead we do the speedrun strat in as few A presses as possible.
On the way to the wooden fence, I ran up the side of the log to fall off it and remove hard landing state, which then allows me to control the direction of spin knockbacks.
The 0.5x was needed to get the height to spin onto the mushrooms.
When on the mushrooms, I slid off in such a way that it bumped Luigi up, which helped get enough height to spin to the ledge.
#SuperMarioGalaxy #AButtonChallenge")
![[TAS] SMG2: Spin-Dig Green Stars](https://i.ytimg.com/vi/dqkDv7LoOC0/hqdefault.jpg "[TAS] SMG2: Spin-Dig Green Stars
Green Stars TAS WIP
Star order has no time difference; the yellow platforms on the 3rd planet might be slightly different but are easily manipulated using the camera.
Green Star 2 - 37.75 IGT - 11:19.80 real-time split (1:34.12 segment)
Crouching next to the orange pole when running towards the edge causes a strangely fast slide for 1 frame. Jumping from this into a double jump-triple jump saved 0.6s over a regular triple jump.
The triple jump sequence also had small adjustments to get the tilt right for the triple jump to curve over the edge and land directly into the Lumas talking area.
Camera turn saved 1 frame of cutscene time.
Drill clip from underneath is an IL strat that saves 1.16s over topside.
The yellow platforms are manipulated with a camera turn so that Luigi lands on the middle ones while they spawn. Their collision moves upwards as they spawn, adding momentum to a backflip to save 10 frames.
Green Star 1 - 15.81 IGT - 12:19.23 real-time split (59.43 segment)
Unlike in GS2, the digger is on a good cycle, so the faster starting movement is a backwards slope jump (which also gets unusually fast speed for a couple of frames) into a bounce and spin; the slope speed is somehow conserved so landing on the sloped edge of the Lumas platform causes an immediate fast slide.
P2 is used on the digger, and then also to spin the Luma. This prevents Luigis spin from making the Luma dizzy which would make it take longer to be able to talk. This saved 8 frames, with the whole digger strat saving 0.32s over the strat used on GS2.
Green Star 3 - 21.38 IGT - 13:36.08 real-time split (1:16.85 segment)
Vine RNG is so bad. Different timing on Piranha kill and even seemingly different timing grabbing the vine has an affect on the speed of its growth.
Luigis speed on the vine is dependant on how fast it grows; this is why its faster to not get the earliest Piranha kill, so Luigi is on the vine as its still growing. His speed is also affected by the rate (and possibly even strength) of spin inputs - holding off on shaking for a couple of frames ended up saving a couple of frames.
In the last section, neutral walljump strain was used to get to the star a little bit earlier.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![World 2 Greens M route [tool-assisted demo]](https://i.ytimg.com/vi/eALHtTRY4wc/hqdefault.jpg "World 2 Greens M route [tool-assisted demo]
This route saves 8 frames over the current W route (Plank → Boulder → Hightail → Cove → Glide → Honeybloom → Lava Lair), when both use their optimal combination of IR cursor and joystick movement.
When both use only joystick, W route is 2 frames faster.
When M route uses optimal movement and W route uses only joystick, M route is 25 frames faster.
Path: optimal movement [gain over other movement]
Puzzle Plank → Hightail Falls: cursor [7 frames]
Hightail Falls → Boulder Bowl: joystick [3 frames]
Boulder Bowl → Wild Glide: cursor [11 frames]
Wild Glide → Cosmic Cove: cursor [9 frames]
Cosmic Cove → Honeybloom: joystick [1 frame]
Honeybloom → Lava Lair: joystick [6 frames]
Interestingly, despite the 4 frames of hitlag from the crystal, Puzzle Plank → Hightail Falls is faster than Puzzle Plank → Boulder Bowl by 4 frames.")
![[TAS] SMG2: Hightail Falls GS3 in 57.23s](https://i.ytimg.com/vi/eFt2U7PtYkU/hqdefault.jpg "[TAS] SMG2: Hightail Falls GS3 in 57.23s
WARNING: Some weird flashing (its strobey but dark) when eating or single jumping with a Dash Pepper (or first person cameras). This is a graphical bug in the older version of Dolphin Im using (5.0 stable).
Green Stars TAS WIP
I was planning to put all 3 on here but GS2 had a desync. It was good in the end as I had found some improvements for GS3 - beforehand I didnt want to have to go back and redo everything but the desync motivated me to go back.
This is roughly a 2.9s improvement over this demo from last year, which has a lot of detail on some tricks: https://youtu.be/IlklCoR-V2A
57.23s IGT - 27:34.13 real-time split (1:53.28 segment)
The single jump into Yoshi was a little faster than another long jump, as another long jump would float in the air too long.
The main tech is Dash Pepper jumping.
Bunnyhopping along the ground is a little over 1 frame faster per jump, as running speed peaks at 32.4 units/frame, whereas jumping speed has an initial speed boost between 60-70, a speed of 32 while going up, and a speed of up to 34.7 while falling. Theres a frame with about 28 when landing.
Jumping on edges when horizontal will cause a lower jump, but on edges of slopes and ramps, it causes a much higher jump, which can then get up ramps faster.
The jump up the first big curve saved ~0.5s.
Double jumps can be constantly done by fluttering each jump, even if only for a frame.
The jump up the big wall saved ~0.8s. This was where I had found improvement in GS2 later on, 13 frames improvement with a much better jumping angle.
The big flutter at the end is the jda flutter. To keep all the speed going forwards, Yoshi needs to jump off the stone block as its falling, and also needs to be *on* the block; jumping off the very edge doesnt work.
After some time, a sideways slightly backwards angle needs to be held to keep speed, as is usually done for infinite flutters.
Flutter height is maxed to be as close to the first flower swing as possible.
This flower swing route is where much of the timesave comes over the 2021 demo. I must have done really badly with this route back then to find it as 0.45s slower than some Dash jumping, when in fact its 0.6s faster with what I had known then, and now its 2s faster with a new technique.
This technique is grabbing a flower swing at a specific spot so Yoshi ends up facing opposite to the way the swing goes. This spot is directly under the swing, with only 1 unit of margin on each side.
This gives Yoshi a lot of backwards momentum to add to a dismount that flies back into the star.
The swing animation is cancelled with a frame-perfect flutter, but this loses a bit of height so it only saved 1 frame needing a weakspin to grab the star.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG: Freezeflame 3 final section without using Ice Flower](https://i.ytimg.com/vi/fUmJLqo8kWg/hqdefault.jpg "[TAS] SMG: Freezeflame 3 final section without using Ice Flower
Now making this public to coincide with Supper Mario Broths Twitter post detailing the water boxes: https://twitter.com/MarioBrothBlog/status/1365716700520460289
[Written from the past; early February]
A few days ago, HEYimHeroic (https://twitter.com/HEYimHeroic) told me that they and Supper had been discussing about water boxes placed in Freezeflames lava platforms, placed there as a workaround for Ice Mario instead of having to programme Ice to run on lava. They wondered if it could be possible to swim below the platforms.
This section had long been considered (but not proven) as possible when the idea of jumping on edges of lava came about: https://youtu.be/Qj2hvqlstp4?t=315 (video was just last year, but the clips were recorded sometime 2019. There are many clips of speedrunners jumping on lava too, as it seems quite easy to do if you overshoot jumping into the lava for Freezeflame 2 skip).
Around the same time I TASed this, Usernameispassword also completed this challenge by solely using these jumps (and long jumps too); no full video yet but there is a clip of their star grab in the TAS server. You should join that, by the way: https://discord.gg/h2YSCZm
But after I had made a showcase for Heroic of accessing and swimming in the waterboxes, I wanted to include the trick while completing the challenge.
It was unnecessary for the first gravity section, as I could have just long jumped across the large gap and got the same 2 lavaboosts to the snowy platforms. But it looks cool, right? Right?
I wanted something different for the 2nd section, so I long jumped over and lavaboosted to the Ice Flower platform, getting a small bounce off the rocky side of a lava platform, and then showcasing the lava running tech that you can learn about here: https://youtu.be/W28X4lLB2nY
For the final section, I did a lava edge jump and angled wallkick, and then swam most of the way up until a point where the next lava platform wasnt connected to the one I was in. I did some slow wallkick and then got a long jump off the edges of two platforms. I ended with a sideflip done by holding back and pressing A just before landing.
So this is one less transformation for SMG1. I understand some define Life Mushroom as a transformation, but with the tech involved, it isnt actually necessary.
ice flower date of death: 8th feb 2021 rip f in chat sob emoji
#SuperMarioGalaxy #Transformationless")
![[TAS] SMG2: Flip-Swap GS1 in 5.96s](https://i.ytimg.com/vi/fnTWLzyNF_M/hqdefault.jpg "[TAS] SMG2: Flip-Swap GS1 in 5.96s
Luigi hyperjumps eastward into a Green star in 5.96 seconds [IGT, 358 frames].
-2 over my 2022 TAS: https://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 its 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). https://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, its 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 dont have SM64s luxury of knowing if were keeping straight while the camera moves, so its harder to get the benefits.
#supermariogalaxy2 #toolassistedspeedrun")
