Submission #9161: Gibran & KM's N64 Super Smash Bros. in 05:12.32

Nintendo 64
baseline
(Submitted: Super Smash Bros. (USA).n64 USA)
BizHawk 2.9.1
18739
60
0
PowerOn
Submitted by Gibran on 7/5/2024 3:01 AM
Submission Comments
Super Smash Bros. for the Nintendo 64 (also known as Smash 64) is the first installment in the Super Smash Bros. series, developed by HAL Laboratory, directed by Masahiro Sakurai, published by Nintendo and released in 1999. It sold 5.5 million units worldwide, the 5th best selling game for the system.

About the run

This is a speedrun of the 1P Game mode that improves on the submission of Haru by 12.26 seconds. Most of the improvements comes from using American version where the AI is significantly more aggressive. We suffer 1 extra frame of hitlag on every direct hit compared to Haru's run but in exchange we can SDI further gaining up time in certain stage.
For RTA timing, this would be a 4:47.30 where the time start when pressing start and ends on the final hit of Master hand rather than after getting your final score.
In this run (as it is also in the previous one) Kirby was picked by how fast he can KO opponents. He has an incredible arsenal of movements like forward throw, neutral b (inhale) and the spike on the up-b falling down. The up-b spike speed is around 170 units per frame and that's regardless of percentage of the opponent, just imagine it!
Here's some key improvements that were made:
  • Every bonus obtained at the end of the stage last 10 frames.
  • By losing a stock on the first possible stage, we remove the No Miss bonus for every stage, saving 110 frames (almost 2 seconds).
  • Taking damage whenever possible to remove the No Damage and Full Power bonuses.
  • Removing the timer at the start saves 60 frames (1 second) per stage, saving 11 seconds for this run.
  • RNG manipulation to get the CPU cooperate or spawn in favorable location.

Version differences

A lot of version differences were misunderstood back when the original TAS was submitted. The walking / jumping / falling speed was unchanged for Kirby between both version.
Here's the relevant universal mechanics changes:
  • The distance traveled by Smash DI is 40% longer, allowing the defender to get out of combos much easier and making multi-hitting attacks less reliable.
  • All attacks have 1 extra frame of hitlag.
Here's the relevant changes for Kirby:
  • Kirby is about 3% smaller (0.94 → 0.91) This gives his attacks less range but also makes him harder to hit (granted he has not performed the bigger Kirby glitch).
  • Nerf Up aerial deals less damage on the clean and mid hits (12% (clean)/9% (mid) → 10%/8%).
  • Nerf Up aerial has a smaller hitbox (150u → 105u).
Kirby being ever slightly smaller could cost some time in a very specific scenario, but it's almost always a non issue. This could also be mitigated by doing the Bigger Kirby Glitch although most stages ends so quickly, so we pretty much never have the time to set it up.
The nerf on the up-air reach does impact us as we could potentially hit sooner or from further away, but this also allows us to trade more easily in certain scenarios and its damage change actually makes it so it's 1 frame of hitlag less, thus mitigating the extra frame added to every move in this version.

1P Game Stages

We'll be indicating the time for the first input in a stage to the GAME SET screen showing up and we'll be using the in-game timer of the game to convert the frames to seconds.
Stage 1 - VS Link
259 frames (4.31 seconds)
We push Link to the right and we throw him off to trade stock as fast as we can remove the No Miss bonus. Another small optimization came by doing a weak forward smash to remove the smash-less bonus. This strategy ended up barely faster than using final cutter (up-b) to trade stock.
Stage 2 - VS Yoshi Team (vs. 18)
819 frames (13.65 seconds)
Kirby's up-air or final cutter works as the best tool to deal efficiently with Team matches, especially at higher difficulty. Since we're purely going for speed, we up-air every group of 3 Yoshis until we're done. A lot of manipulation with our position / timing was done in order to make the following group of Yoshis spawn in favorable locations.
It is worth mentioning that it is necessary to not KO the Yoshi's in their spawn order as it would get another bonus Yoshi's rainbow losing 10 frames. This is an unlikely event, but something to keep in mind.
Stage 3 - VS Fox McCloud
280 frames (4.67 seconds)
Unfortunately for us, Fox will always laser if we are horizontally aligned with him, so we just took the damage, which will remove No Damage / Full Power bonuses. Nonetheless, a better scenario would have been for Fox to run straight towards us. Some manipulation is done to get Fox in a favorable location for him to think that doing an up-b towards the edge of the ship is a good idea. We intercept him with a final cutter to make him reach the bottom blast zone earlier.
We could have KO'd Fox 2 frames earlier, but those were sacrificed to get DK as an ally instead of Link for the 4th stage, which loaded 1 frame sooner. Link has the most polygons to render in the game and it affects the versus cutscenes by a few frames.
Bonus Stage 1 - Break the Targets!
107 frames (1.79 seconds)
For in-game time this would be a 1.75 since the 1st and last frame doesn't increase the timer. This is a 6 frames improvement over what Haru did (0.1 second).
Instead of a stick jump + up-b to our doom, we land as soon as possible on the pillar, do a dash and taunt cancel to not fall off, walk 1 frame into down-b to do what is called rock slide/rock cancel. By walking and doing the down-b, you do the move grounded which comes out way faster than if the move was done airborne.
It's unfortunate that we skip those iconic bonus stage, but we are on a quest to defeat master hand and sadly we don't have time for it.
Stage 4 - VS Mario Bros
195 frames (3.25 seconds)
Similar to Fox, at very hard difficulty, Luigi will always throw a fireball on the first frame because he sees your ally (DK in this case) directly in front of him. With some clever movement, we can barely dodge the fireball while hitting Luigi with an up-air to send him to the right and fall on Mario with the end of the same aerial. We now have them in a good spot for a double KO using final cutter! A very clean looking fight!
We tried to trade hit with Mario to save 20 frames on the bonuses, but where we catch him he only cares about recovering, so he won't throw aerial moves.
Stage 5 - VS Pikachu
143 frames (2.39 seconds)
We couldn't replicate what Haru did, potentially because of the version differences, so instead we dash towards pika and push him just enough to grab him above the void, so when we throw him, we both lose our stock.
Another occasion where starting with more stocks than Haru, allowed us to lead to more different routes.
Stage 6 - VS Giant DK
172 frames (2.87 seconds)
We use Giant DK's aggressivity to our advantage and we can just run to the right, swallow him up (neutral left + b) and fall both to our doom. Lucky for us, we have more stocks than him so we'll proceed to the next stage instead.
Bonus Stage 2 - Board the Platforms!
258 frames (4.30 seconds)
In-game time, this would be 4.27 which is almost 0.5 improvement versus what Haru did! Haru would get to the top, short hop fastfall back air to pass between the bumpers, and up-b to the bottom blast zone. Instead we get to the top, dash and taunt cancel walk down-b (similar to what was done on the Break the Targets level) we undo the transformation right under the bumper and SDI for 4 frames to go as far down as we can. The moment Kirby is airborne again, we down-b to the bottom blast zone.
So once again, we cannot take the time to complete the stage. This was overall a fantastic time save on such a short stage found by Gibrani.
Stage 7 - VS Kirby Team (vs. 8)
744 frames (12.40 seconds)
We start by attacking the DK hat Kirby so we won't get Kirby Ranks bonus and from there we manipulate our position to make each Kirby spawn on the left side so we can intercept them as they fall. We trade with the last Kirby, so we avoid No Damage / Full Power to save another 20 frames!
A few frames were sacrificed in this fight for a more favorable RNG against Samus.
It's important to note that the RNG can only be modified during the actual match. Waiting more on the point or versus screen doesn't change the expected outcome.
Stage 8 - VS Samus Aran
259 frames (4.31 seconds)
Similar to Fox and Luigi, Samus was programmed to use neutral-b if we are aligned with her, so that's what she does as soon as the match start. We run towards her and that makes her cancel her charging animation and thanks to our previous delay manipulation on Kirby stage, Samus thinks that she can grab us leaving her open for our up-air attack.
There's very high chance of her simply stuffing you with a down-smash and it would be hard / slower to dodge due to the stage layout. After the up-air we keep dashing and trade a neutral air with her fair on which we SDI to the right to prepare a nasty down-air spike the moment that we're out of hitstun. We minimize the number of hits that we do on Samus to reduce the hitstun and we steal her double jump with a second down-air and we wait for her to hit the acid, followed by another 1 hit down-air to make Samus pass through the acid and be KO'd by the bottom blast zone.
This was extremely difficult outcome to manipulate and way faster than what was done in the previous version of this TAS.
In the neutral air / forward air trade that we got, Samus inputted 1 frame of SDI to the right which seems to be fairly rare. Often she would input an SDI towards the center of the stage instead, making the ending much slower or sometimes just not possible.
At the end, when Kirby hit the acid, we could SDI to lose a stock but it would give an extra bonus Mystic and end up losing time. Quite unfortunate.
Stage 9 - VS Metal Mario
128 frames (2.13 seconds)
This is the fastest non bonus stage and it's similar to Link and Pika except Metal Mario comes to our grab instead of us having to go for it.
Two interesting observations:
  1. For some reason, we didn't lose our stock, maybe because of the heavier weight of Metal Mario, we're not entirely sure...
  2. We were facing the stage and the throw still didn't land on the stage!
Turning doesn't change your x position, so once you're in position, you can carefully turn around and it'll still work. Maybe useful to know for versus players. This state can be achieved in several ways, running towards the edge and grabbing before you're airborne, doing a dash attack before you fall off or simply rolling towards the edge. A property of this "further grounded" position is that if you jump straight you will not land back where you were. This is further than the state called Teeter where the character is looking down barely hanging on the ledge.
Another small observation that may be noticeable, the grab throw was not done in the first possible frame, the reason is because we require a good combination of polygons for the next bonus stage, so 3 frames were sacrificed.
Bonus Stage 3 - Race to the Finish!
1036 frames (17.27 seconds)
This stage is the primary reason why we used this version. The increased aggressivity of the CPU and the extra SDI allowed us to save an impressive 5 seconds versus what was previously done!
A lot of planning and investigation went into what combination of CPU yield the best results and here's what we found.
  • First polygon should be Link as he has the strongest dash attack for a solid 14%. We then keep going and roll on purpose in the explosive barrel and even SDI in the 2nd one to take even more damage. We end our SDI in order to hit the wall next to the bomb chute so we can reduce the hitlag.
  • The second polygon should be Yoshi which should forward air followed by down smash or forward smash depending on your character. In this case down smash is just enough to get over the ledge to fall to the second level. We then use the bumpers for a quick boost roll and we're down to the final section.
  • The third polygon is not as crucial as the other two since we're no longer building damage. We just want to be sent as far as possible towards the end goal. In this run we got Luigi and he back air us towards the end where we can tech roll on the floor and finish with a dash attack to save 1 extra frame.
Ideally the last polygon would have been Captain Falcon, he could have used his up-air to send us in a fully horizontal momentum at higher speed.
See this optimal run as an example below
This would have save an extra 0.5 seconds if we could get the correct seed. We tried looking for around a 1000+ seeds by delaying the throw at Metal Mario and changing the fight, there was only a handful of seeds with the beginning combo of Link, Yoshi. Maybe in the future if there's better refinement on how the RNG works for this game and we know which seeds to aim for, we could make it happen.
All the polygon investigation of what combination would be faster was discovered by Gibrani alone.
Stage 10 - VS Fighting Polygon Team (vs. 30)
1565 frames (26.09 seconds)
By far the hardest stage to TAS. The CPU are notoriously more aggressive and very hard to deal with on this version of the game compared to the Japanese version. We start by getting rid of the first 3 polygons in order to set up easier triple KO down the line. Each group of 3 polygons is essentially a cycle so you need to deal with 10 cycles. There's several things that can go wrong on this stage. Each polygons have different weight so even hitting them at the same time might not KO them at the same time which will affect the next cycle timing.
There's very heavy manipulation to get the cpu to spawn in favorable location where you can hit them before they attack you. A lot of it was playing around and figuring out at which frame, what height or position we need to be and sometimes it was not always possible or advantageous to make them all appear on top of each other so it's something you have to deal with.
Lighter character like Jigglypuff or Pikachu tends to get star KO'd if you hit them from too high, but if you wait more, they will throw a move to prevent you connecting Kirby's up-air. This is very hard to deal with while also manipulating spawn for other CPU.
In one instance, KM traded an up-air for a luigi down-air which cancelled my first up-air setting up for a really fast kill on the other 2 polygons that were falling a little later. We ended up the fight with a triple red shell KO and Gibrani made them all die on one side, which is quite precise, because it has to be at a specific spot for this to happen and if the shell is thrown too early, one of the polys will get star KO'd. Even so, the result looks pretty nice!
The fight was maybe about 1 to 2 seconds slower than Haru's original fight, but that was expected since the polygon are now level 7 instead of 5. Although we did save 50 frames (almost 1 second) by having only 2 bonuses instead of 7 at the end of the fight.
Stage 11 - VS Master Hand
704 frames (11.73 seconds)
We did our best to maximize the damage per seconds while manipulating which action would Master Hand take. The gun is notable in the speedrunning community to waste 5-8 seconds and being the worst attack when going for a fast completion of this stage.
We use a mix of aerial when the boss will move, forward-air or back air with a neutral-air as we land. The key move that drains most of the health is the notable Kirby up tilt which can be spammed every 0.4 seconds and that's with the hitlag, ridiculously busted move!
It does look very cool after we tech the first attack of Master Hand and we have so much momentum to the left that we are sliding on the floor during our up tilt animation.
The time of this TAS ends on the last input of this run after the final high score is calculated which is 19+ seconds after the last hit was dealt. That's a very long KO animation for the final boss.

Final comments

This is 1m 51.16 second of pure pink ball destruction with 3m 22 seconds of transitions and points calculation, basically more transitions than actual battles.
Despite kirby proved to be the best choice for this speedrun, other characters that could stand up to him were also considered:
  • Kirby is the 2nd strongest character in the game only behind Pikachu, but he lacks moves to deal efficiently with team matches, and also he would not be as good as kirby for 1 vs 1 matches
  • Another consideration was Fox, he is very versatile, with a pretty decent moves, in addition to his iconic shine (down-b) which is quite good on multi team. He has a faster btp, basically around 3 seconds faster than kirby there, and runs really fast for race to the finish. But even so, he could struggle too much on some 1 vs 1 matches.
  • Captain Falcon could be good too, as his attacks are strong, and he is fast for the bonus stages, but he would certainly be a bit slower for 1 vs 1 compared to the kills that kirby does.
  • Lastly DK was considered since his up-b is a trivial way to deal with multi team stages and you can cargo suicide on 1 vs 1, but it's overall slower. Besides his board the platforms stage is also 12 seconds long, so that would be almost 8 seconds deficit just there and he could not catch-up on race to the finish either.
No doubt kirby god

Credits

Thanks for Haru for getting a very solid run for us to challenge. This was a fun run to attempt but next time we'll be back with some massive improvements to Break the Targets and Board the Platforms run. We cannot give a timeline but if you thought this was great, wait until we destroy almost every stage with brand new strats :D
Special thanks to my co-author KM for minor improvements on several stages (Yoshi, Fox, Mario, Metal Mario, RTTF) and doing almost entirely the polygon fight. And for writing this entire submission

References

Check the Smash 64 RTA speedrun records as well as a spreadsheet which has every record in the game (mainly Break the Targets, Board the Platforms and Highscore run) if you want to know more about single player mode

Darkman425: Claiming for judging.
Darkman425: Replaced input file with a 52 frame improvement by the authors.
The change of game version at first glance would seem to be the slower choice. The increase of a frame hitstop on each hit and the nerfs on some of Kirby's abilities in the US version would make it initially seem slower. However, the fact that there are no regional changes in Kirby's movements thanks to better understanding of regional differences, the more aggressive opponent AI in the US version leading to faster strats in most cases, the use of intentional deaths to cut down time across the run by removing the No Miss bonus, and general optimization improvements throughout makes incredibly impressive. It also helps that there's a better understanding of the regional differences that make this more understandable. In the Smash Bros. announcer voice: Nice work!
Accepting to Standard as an improvement to [2842] N64 Super Smash Bros. by DennisBalow in 05:25.43.

EZGames69: Processing...
Darkman425: Replaced input file with one that sets the rerecord count as unknown.
Last Edited by Darkman425 on 7/24/2024 4:28 AM
Page History Latest diff List referrers