Uploaded November 2022 | Updated September 2026, 2 weeks ago
8.51 IGT - 50:28.16 real-time split (51.85 segment)
Improves on 2020 TAS by 0.32s (19 frames)
Mega triple jump strat found by Vellopia (original vid deleted)
8.51 IGT - 50:28.16 real-time split (51.85 segment)
Improves on 2020 TAS by 0.32s (19 frames)
Mega triple jump strat found by Vellopia (original vid deleted)
![[TAS] SMG2: Bowser Jr.s Fearsome Fleet Green Stars](https://i.ytimg.com/vi/MumvKuWTB3k/mqdefault.jpg "[TAS] SMG2: Bowser Jr.s Fearsome Fleet Green Stars
Green Stars TAS WIP
Star order matters only a bit, as the position of the Fizzlit is a little different between cutscenes and no cutscenes, but this overall changes the optimal fluttering line by a few frames.
Green Star 2 - 20.45s IGT - 1:10:10.16 real-time split (1:21.00 segment)
This is a 0.5s improvement over my TAS tests in 2020, meaning this IL could be improved: https://youtu.be/3R1MdYG6iDM
Theres improvement in the route to Yoshi, getting a much earlier long jump off the statue at the cost of a less direct long jump route, saving 5 frames overall. This is 6 frames faster than long jumping around the normal way.
I tried for a zip slope jump, but the slipperiness is inconsistent around the statue and the only zippy slides went straight off the edge, unable to jump with the added speed.
The Fizzlit bounce strat is from nemi43 way back in 2012: https://youtu.be/dBFVlublQTc
After 10 seconds of infinite fluttering, your speed gets cut to 0. Bouncing off an enemy resets this airtime timer, so I take a curved flutter line to bounce off the Fizzlit.
I flutter as early as possible, taking advantage of higher acceleration when fluttering close to ground, as well as higher acceleration from a more backwards-moving flutter before shifting sideways for a higher top speed.
These advantages - as well as a cooldown on being able to spin (head-shake) after mounting Yoshi - outweighed the acceleration gained by shaking before starting to flutter. Shaking would have been slower by 10 frames.
When fluttering, the top horizontal speed is 32 while rising. But while falling, speed is dependent on your angle, with 34.7 the maximum when as closest as possible to a right-angle.
The rapid fluttering towards the end maintains this higher falling speed.
Green Star 1 - 12.83s IGT - 1:11:06.06 real-time split (55.90s segment)
An improvement of 9 frames over my 2020 TAS: https://youtu.be/ubBjFiyfrek
As a straight line is better here, the flutter start is different, since backwards acceleration would cause a slow curve. This means the acceleration gained from shaking is actually faster here.
When dismounting, I jump into the Gunner to get a slight push from it to grab the star 1 frame earlier.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG2: Fluffy Bluff Greens](https://i.ytimg.com/vi/NAdhHZ54ZN0/mqdefault.jpg "[TAS] SMG2: Fluffy Bluff Greens
Green Stars TAS WIP
Star order doesnt matter, as there are no affected cycles and the land cancel doesnt matter as all the stars take the same path to a single point (Cloud Flower). Star order is 3 2 1 to keep with RTA runs and make the selects look a little nicer.
Green Star 3 - 31.85 IGT - 18:19.33 real-time split (1:29.42 segment)
Because of weird edge interactions when standing up, the Cloud Flower was instead picked up with an uncrouch so that Luigi could stop earlier to backflip earlier.
Going up and cloud long jumping down to the next planets flower is 3 seconds faster than an alternative strat: 3 cloud long jumps from the bottom of the tree, landing at the start of the planet and cloud jumping up to the flower.
Ending strat was frustrating to test, as there was hope for cloud sideflips to be useful but they only seemed to pick up speed when going off to the left rather than straight forward. The best sideflip strats ended up 0.4s slower than this final strat, while other ideas (like sideflipping off the slope) lost more.
The final strat uses a grounded cloud long jump to get the bulk of the height, and another one in the air.
Forward slope jump speed works off the most recent running speed, so the third cloud jump was delayed to pick up speed to use later, saving more time than the extra height of an instant cloud lj would have.
Each slope jump is frame-perfect, and picked up a little bit of extra speed too.
Green Star 2 - 9.80 IGT - 19:12.75 real-time split (53.42 segment)
Here is where cloud sideflips shine. When the speed gain actually works (GS3 grrrr) you can pick up a similar high speed from them as cloud long jumps (about 29-30 top speed compared to 34), while also getting a lot of height, and its an instant jump.
The first cloud sideflip didnt get the speed for some reason.
The second needed to go backwards, so the movement in the air was mostly leftwards before switching to the right on the last frame to minimise horizontal movement.
The third gained speed straight into the star.
This strat was 2 frames faster than its closest rival: two backwards single jumps on clouds into a cloud sideflip mostly used for its vertical.
Green Star 1 - 11.05 IGT - 20:07.41 real-time split (54.66 segment)
UPDATE: Retested this some more with the camera idea below, improved to 10.80 (15f): https://youtu.be/cAsWuTeXKvg
The backwards single jumps worked better in this star due to its farther distance needing that extra speed where two starting cloud sideflips would have less.
Speed is a weird thing. These jumps seemed slower if moving straight backwards towards the camera. They gained more speed if the camera started turning. Straight backwards was 0.4s slower than this camera-turning sequence.
Need to test more with deliberate camera turns, but with this setup, several different angles and adjustments either got 11.05 or slower so I cant squeeze any more frames.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] SMG2: Melty Monster Green Stars](https://i.ytimg.com/vi/OQKR7Ms7_LQ/mqdefault.jpg "[TAS] SMG2: Melty Monster Green Stars
Green Stars TAS WIP
Star order used to matter when GS2 used the lava waves, but no longer does.
Green Star 1 - 13.83 IGT - 1:57:44.80 real-time split (1:11.55 segment)
Land cancel doesnt seem to be possible here.
When grabbing pull stars, theres 18 frames from pressing A to the beam grabbing Luigis bubble - this is the same no matter the distance.
With this in mind, its generally faster to grab as few pull stars as possible to keep as much pulling time, and also pull from further away to get more speed (which is why I sometimes wait it out until a good one appears).
For the end, I used the IL strat of quickly exiting a pull star (done here by Emperor: https://youtu.be/8Gq0zWl6Hl0?feature=shared)
Green Star 2 - 33.45 IGT - 1:59:01.58 real-time split (1:16.78 segment)
This is an improvement of 3.01s over my TAS from 2021: https://youtu.be/ig1Qc9T3mCE?feature=shared
About 0.6s of improvement came from a better pull star section.
2P grabbed the Life Mushroom ready for the lava wave wallboost strat (found by nemi43; see my previous TAS), but as it turned out, thats no longer fastest.
Instead, it is possible to get zip jumps (glitchy slope jumps) off two opaque crystals to soar across the waves and save 2.2s over the lava wave strat.
It is possible to get a single jump with conserved speed when landing from the first zip jump that can get on the second crystal, however, the angle makes the second zip jump worse so its not faster (from what Ive tested).
The final improvements come from a better ending, with a better backflip spin into a higher lava boost gets a higher sandstorm spin, which hits the star around the top half of its collision which is slightly faster than the bottom.
Green Star 3 - 43.90 IGT - 2:00:40.68 real-time split (1:39.10 segment)
This is an improvement of 2.56s over my TAS from 2020, despite cutscenes: https://youtu.be/MhqGMs8oWSw
The slightly wayward start is because a straight path would have had the first LJ land on the slope, where tilt would make the second LJ higher and would land later for the single jump, losing a couple of frames to what was done here.
The main improvement is another big new skip: the launch star skip, using rock mid-air jumps which you can learn more about here: https://docs.google.com/document/d/13HenHN3IA2mQih2k8Ve_VZdK-yIoX8fnXB2g2eMaZGM/edit
This skip is tighter than it looks, as there is a 10-second timer on how long you can stay in the air before the game instakills you. This landed 5 frames before that point.
The big downside to the big skip is that it also skips the trigger that blocks you from stopping rolling. This is a problem because it also allows for the automatic stop that happens after rolling for some time. This can be worked around by jumping on each section.
Overall it saves 4 seconds.
Crystals are avoided as hitting them causes lag frames.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG2: Honeybloom Green Stars](https://i.ytimg.com/vi/P3fjZ-mpwn8/mqdefault.jpg "[TAS] SMG2: Honeybloom Green Stars
Green Stars TAS WIP
I had a longer description describing the slower strats compared against but it ended up with like 4k characters so trimmed it down.
Star order matters in this galaxy as the cycles of the Choppahs are changed by cutscenes. This changes the strats for GS2 in the bouncy fruit and pole sections, and cutscene cycles end up a little bit faster (8 frames), so GS2 goes first.
Green Star 2 - 57.01 IGT - 43:50.98 real-time split (1:52.88 segment)
At the rainy wallkicks, backflip spin wallkick was 14f faster than the nearest strat involving a Goomba bounce. Neutral was used when landing to be able to long jump immediately
Right after, long jump spin wallkick spin to land higher and then backflip to the top was 8 frames faster.
I held Z when landing and let go to get buffered crouch acceleration, saving a frame around the corner.
The waterfall section: As the final Choppah was high enough for 2P to freeze and Luigi to squeeze under, the main goal was the fastest route to get to that squeeze point grabbing the ledge.
Backflipping onto the tower was the fastest way to get to the cloud & flower used to get to the ledgegrab point, with the next closest being a long jump under the tower into a triple jump spin (27 frames slower).
Due to the squeezing, Luigi automatically climbs the ledge, and also gets a little proxy boost on the way. What is done here affects the route to the vine: a backwards jump with long jumps into the Beeshroom + a frame of hovering was 8 frames faster than the best route jumping forwards.
When Bee hovering, you get 8hspd if youre hovering upwards, and 10.25 when going down. Mashing A when just barely dropping keeps good height while maintaining the highest speed.
The bounce into the frozen Choppah gained a bit of extra speed skimming (clipping) off the tail, making it the fastest way onto the cloud.
Spinning towards the Choppah while letting go of 2P A ended up getting the damage boost a little higher, so I was able to bounce earlier and land the damage boost 1 frame earlier.
Spin-cancel ground pound grabbed the pole 1f earlier than a spin.
Another squeeze through a Choppah. 2P freeze changes its collision slightly, but starbit stunning during the freeze restores it. This was used to get another small proxy jump into a double jump, saving 12 frames.
Green Star 1 - 42.00 IGT - 45:15.68 real-time split (1:24.70 segment)
As explained above, the Choppah cycles are different without cutscenes, so the waterfall route has a new look, as you cant squeeze under the last Choppah. Along with this, the first Choppah gets in the way of GS2s faster backflip tower route.
In the end, this new triple jump route with a skim across the top of a 2P frozen Choppah ended up 14 frames faster than the nearest, which was an altered version of the backflip tower route.
For the bouncy fruit, taking the higher road with an extra coin refill was 25 frames faster than the low route I took in this 2020 TAS: https://youtu.be/3cR-YPmpx9Q
For some reason, there was a huge boost off the last fruit, more than double the initial speed of a regular bounce. The starbits had no effect.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG2: Flipsville Green Stars](https://i.ytimg.com/vi/PntYuh_2Q3w/mqdefault.jpg "[TAS] SMG2: Flipsville Green Stars
Green Stars TAS WIP
Star order matters slightly - the cycles of the moving platforms means GS3 is faster to do without cutscenes by 0.25s, while GS1 and GS2 differ because of the effect of land cancel on their routes. GS2 is 4 frames faster to do first with cutscenes.
Green Star 2 - 10.78 IGT - 1:15:11.96 real-time split (1:06.43 segment)
This is an improvement of 0.1s IGT over my 2020 TAS (https://youtu.be/pUWDsUgGFPw) however, this is with cutscenes, so the improvement is more like 0.6s.
Top running speed is gained at the start by doing the second long jump immediately on safe ground. While I dont make use of having top speed, this tech also gives higher acceleration, making the quick-turn for the third long jump 5 frames faster.
Getting onto the roof bracket to the star was already around 2 seconds faster than the RTA strat going to the side and backflipping up, but there is now also a zip jump that saves 0.3s.
Green Star 3 - 17.11 IGT - 1:16:12.41 real-time split (1:00.45 segment)
This is an improvement of 0.25s over my 2020 TAS (https://youtu.be/oAaXH5qgCjk).
I improved the Toad bounce by lining up a more sideways angle. Going more sideways in the air is a little faster than straight backwards (known from infinite fluttering), getting up to 34hspd compared to 32.
This led to ground pounding the grate 7 frames earlier.
The remaining improvement was general optimisation of the final section. This section includes getting the top running speed long jump tech, and doing a leap of faith into the star by using a mechanic of jumping off a moving platform over a void. This causes Luigi to accelerate when falling, even when moving forwards.
This leap also clips through the fence, getting ejected at 66hspd to save an extra frame.
Green Star 1 - 5.91 IGT - 1:17:01.66 real-time split (49.25s segment)
This is an improvement of 0.25s over my 2020 TAS (https://youtu.be/H2jkvp5i-1w).
The main improvement is neutral walljump strain, where going neutral on the joystick after walljumping can bypass the restricted movement walljumps usually have.
This was particularly useful for the second walljump, where I held Luigi back enough to avoid ledgegrabbing, and then forward to maintain speed when landing.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] Super Mario Galaxy 2: Yoshi Star Green Star 3 in 1:01.10](https://i.ytimg.com/vi/QJkLo8j6kSE/mqdefault.jpg "[TAS] Super Mario Galaxy 2: Yoshi Star Green Star 3 in 1:01.10
Luigi, Yoshi and P2 go soaring through Yoshi Star Galaxy to collect a Green Star in 1 minute 1.10 seconds [IGT, 3666 frames].
The spring bonk strat was found by Qr1pt1k, and is 0.3s faster than just long jumping up and onto the Magikoopa.
For the infinite flutter, I tried emulating and optimising Jeromes instant triple bounce off the fruit (https://youtu.be/FduJb1he_Qs?t=26), but it was frustrating trying to angle Yoshi correctly with the flutter.
I eventually managed to get it, but 8 frames of lag are lost getting grabbed by the Smeech, so theres improvement to be made there. For some reason, I actually needed to get grabbed by the Smeech to bounce off the fruit in that way. Trying the same angle and P2 freezing the Smeech ended up with no bounce at all.
After the flutter, I jumped out of the ground pound and spun up to the sign, and from here was able to long jump with a slight rightwards strain to barely land on the next ledge.
Backflip spin was faster than long jump wallkick spin by 5 frames, and by landing on the edge of the spring, I was able to backflip off it which would have saved time if finishing this planet without Yoshi was faster.
Because that route is not faster (see below), and because the Yoshi egg isnt active until just as I land on it, the spring backflip isnt faster nor is it slower than backwards spring jumping.
It is also because of the eggs late spawn that P2 couldnt break it to save time. You can only go as fast as Yoshi eggs respawn.
For the ending route, I first tried going without Yoshi through the launch star (which is a faster way to the launch star) and climbing up to a Yoshi egg near the first moving platform. The route continued with a couple of Z flutters (infinite flutters) out of the first flower swing to the second (the moving platform doesnt move as early as the Yoshi route), landing on the second moving platform and jumping to the next flower swing to do some swings up to the star.
This ended with a 1:01.53.
I then tested taking Yoshi through the launch star, doing a sideflip with flutters up to the first flower swing, which in retrospect had suboptimal execution. The elevator was in a better spot to get a momentum-boosted sideflip up to the flower swing, and then I crouched to get lower in the air to bounce off the Octogoomba.
This - plus a quick spin at peak bounce height - helped to grab the next flower swing much earlier than the no Yoshi route had, resulting in 0.43s of timesave overall. Thats just comparing a much more optimised no Yoshi against a slightly sloppy Yoshi route, so it can be a bigger timesave.
Taking 4 swings to the star is 0.3s faster than any alternative routes, such as taking 2 swings and fluttering up to the last swing.
With further optimisation, I expect this to reach sub 1 minute.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Shiverburn Green Star 2 in 19.98s](https://i.ytimg.com/vi/QdUV1xkWiuo/mqdefault.jpg "[TAS] Super Mario Galaxy 2: Shiverburn Green Star 2 in 19.98s
Luigi cloud jumps to another planet to collect a Green Star in 19.98 seconds [IGT, 1199 frames].
This isnt optimised as well as it could be but sub 20 seemed a good milestone for testing against.
This is an improvement of 31 frames over sonicpackers 2014 TAS: https://youtu.be/tH5n3u8wjCo
The majority of the timesave comes from single jumping into the cloud and turning, rather than long jumping and ground pounding. This gets to the platform 20 frames earlier.
The remaining timesave was in the cloud jumping, particularly the first cloud jump. I formed the cloud a bit earlier, and also over lava so I could spin cancel with the lava being the target ground instead of the cloud. This allowed me keep more speed for some reason, so I could continue running when landing and could long jump earlier too.
A potential improvement would be to optimise the first and second cloud jumps to skip needing a third. Skipping this could be possible, considering I had to set up the third jump very soon after the second so the earliest long jump wouldnt go too high and miss the star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Puzzle Plank Green Star 1 in 3.30s](https://i.ytimg.com/vi/QnROcaphJX0/mqdefault.jpg "[TAS] Super Mario Galaxy 2: Puzzle Plank Green Star 1 in 3.30s
Luigi unleashes his true jumping ability to collect a Green Star in 3.30 seconds [IGT, 198 frames].
Diamentos found yet another mega triple jump, transforming my original triple jump strat (https://youtu.be/ZSKT4ldZKJw) into something extra special, beating that TAS by 5 frames: https://youtu.be/wIN_DU4zrZw
[fun fact: my triple jump strat was originally found for minimum spins challenge]
The specifics on how to optimise mega triple jumps are still a little unclear (so I am certain it could be improved) but this is how this one worked out:
The first jump is angled slightly to the left of a straight line to the star, and is strained left as well. It is a short hop of two frames holding A that makes a tight and fast swing around the edge.
Before landing, I held neutral to conserve forward momentum into the double jump. This jump was made to be higher so it could get over the next edge, and I continued to strain slightly left while holding A [need exact frame count of holding A, it was roughly 20 frames].
Once letting go of A, I started to strain right. This left and right strain seems to have some effect on the resulting triple jump as, when taking a straighter line, I couldnt get anywhere near what I did here. Possibly to do with how it interacts with gravity, adjusting tilt, or just small adjustments of forward speed that couldnt be done with differing forward inputs? Or even simply counteracting the strain right before the triple jump explained below, so that it ends up being a straight jump to the star.
Getting closer to the edge, I held back (with strain to the right as well for a tighter adjustment of backwards speed) so that Luigi would land on the other side but with all the tilt stored from earlier. This causes the triple jump to go significantly higher than normal, with the first few frames having 50+ y velocity.
I held neutral again before landing to again conserve more of that momentum., and then forward as soon as I pressed A to go towards the star.
Despite not having faster yspd than continuing the jump, spinning underneath the star saved 1 frame due to the spins extended hitbox (it might also reset Luigis tilt so the hitbox goes forward and slightly upward, changing from almost lying down to standing up).
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
