IcySpeedruns
[TAS] Aztec - Secret Agent - 1:26.12 (Human Theory)
updated
In this TAS, I use every playable character in the Home-Run Contest, and instead of going for high scores, this is focused on fun, cool, and interesting gameplay!
I spent some time looking into some strategies for this and found a few flow chart methods that work really well. From the ledge, most characters won't hit you and you can safely roll in most of the time. From here, you can grab the opposite ledge or use Down B. Rolling in can cause a few characters to fall off if they use an aerial, while Down B is a good disruptor and does some damage. Down B can also KO Kirby and Jigglypuff.
If you isolate an opponent, you can do a fairly safe BThrow if you're snappy with it. Fair from the ledge into regrab can stage spike DK, and Up B can act as another disruptor. Finally, ledge attack can clip a few characters for easy KOs if you get a reverse hit, but it's quite difficult to do safely.
This was a pretty good attempt overall and I'm happy with the result!
For the most part, this minigame is just mashing A as fast as you can. Steering is an interesting mechanic however, as you don't lose any speed when steering along flat sections, but you can lose some on the turns, both by steering in the wrong or right direction. Because of this, steering should be minimized, although there are only small sections of each turn that can cost time. Some of the turns also don't need any steering as there is no risk of falling off. I used the second team to demonstrate this by excessively steering the entire time.
Braking can also be used instead of steering, and allows you to use less of it, but this is overall slower. Braking is not necessary at all.
Unless I missed an important mechanic or nuance of this minigame, I believe this time is maxed out, beating the non-TAS WR by 4 frames.
While this minigame appears to be pretty simple, optimizing it is very difficult. Many of the turns are actually fastest when slowing down and grinding the wall a bit, as opposed to going at full speed and wider around. Compared to the current non-TAS WR of 56.53, this run gains 3 frames at the start with mashing A faster, and 3 frames throughout the course.
From my testing, it seems the blue team can be 1 frame faster, being closer to the wall, despite having to take a sharper hairpin turn.
This is almost certainly not maxed out, but pretty good overall.
This is a quick little TAS demonstrating that The Frog Flap can be won while hitting all of the red platforms that speed up the game, and also avoiding all the blue platforms. Nothing too interesting comes out of the increased speed, while on the other hand, too many blue platforms softlocks the microgame.
This TAS goes through the mini-golf mode for Luigi's Garden using fast terrain in as few strokes as possible. I also opted to go for these in a playaround style.
Hole 5 is the only one that I believe is impossible to get a hole-in-one on. The front side creates too much distance with all of the angles, and the backside hits the wall too directly to get around the corner. The area where the hole and pin are is also sloped on both sides. The only way I think this could be possible would be a weird collision with the geometry to send the ball flying over there, but even if such a hit can be done, the likelihood of having the ball land in the hole is also low. It's also worth noting that this hole and all others is possible in one stroke on normal and slow terrains as the hole is much closer.
Note that there are graphical errors due to emulation limitations.
Special thanks to the Pokemon TCG (GBC) disassembly project as a resource for information and research.
Final decimal is 8.40s. This is pretty tough to optimize with the tight movements, lag, and high speeds, but I'm pretty confident that 0:07 isn't possible.
The dinosaur vs the monkey for a quick 15.
With Turbo Mode and not really having to deal with guards, this level becomes pretty much pure strafing and lightning speed. The final time on this is 27.45, but I don't think 0:26 is possible even with really polished movement.
This run shows off the game through the wide variety of abilities and mechanics in fun and entertaining ways! Very happy with how this project turned out!
This is an improvement to my 1.88 from a few years ago. Optimization for this stage challenge is to simply have 2 golden Chu Chus enter your rockets as fast as possible, but even just one is very rare. As a method for manipulating RNG, I made a stage in the stage builder that has Chu Chus constantly running into rockets. Interestingly, because RNG doesn't advance when not called, the Chu Chus that would have come out in the stage builder could instead come out of the rockets in the stage challenge if I quit beforehand. Essentially, I would wait until I had two golden Chu Chus come out in close succession, and then hope they would also come out of the right side rockets in the stage challenge.
After about half a million frames worth of RNG calls through this method, this was the fastest that I found. A bit more time could hypothetically come off, but I couldn't find such RNG to make that possible.
In this quad TAS, I compare all 4 of the meter timing games with the audio only coming from Log Chop. The first 3 use identical inputs and you can see that they remain completely synched together. However, Shingle Smasher has two differences compared to the others. First, it has lag frames which is why over time it falls further behind, and completes 75 rounds slower than the others. Second, the meter starts up higher than the other games, which slightly adjusts the timing and cycles, but is otherwise not very different.
For the gameplay itself, the top bar starts off lasting for 4 frames, and then goes down to 1-2 frames as the game speeds up. There initially are also 3 cycles to time pressing A, but at faster speeds, there's only enough time for 2 cycles. And then once even faster, there are 3 cycles again for a brief period of time. Pressing A also stops the game on the next frame, so you generally want to time it slightly before the meter reaches red for hitting the range most consistently.
Overall, the results are mostly just confirming what most people would expect, although it's interesting that Shingle Smasher is different.
Hot Dog Hog is fairly easy, even at high speeds, and the WR for it is maxed out at 999. Initially, you can bite every 45 frames, but this frequency increases to a maximum of 11 frames as the game gets faster. It's also possible to each the hot dog on each difficulty in only 2 bites.
This TAS mostly functions as a playaround with clearing each round in fun ways.
Nunof Yerbizness informed me of a softlock in this microgame on the hard difficulty by hitting multiple blue platforms in a row while avoid red ones. With some timing and calculations, I was able to make this TAS by making a very tight jump to skip two red platforms. One excess blue plays the game much slower than normal, while two excess blues decrease the speed and gravity so much that you get stuck at the top.
Once in this state, you can neither win nor lose, and your only option is to quit out. Very cool!
Parking Prowess is simple and straight-forward, with just properly timing the movement. Each difficulty has different variants, and turning is quite sensitive, so precision is required at the faster speeds.
There aren't really any fancy tricks for this microgame, but as the speeds get very high, timing gets incredibly precise. For the earlier rounds, it's possible to simply tap and hold left at the correct time, but for faster rounds, it's required to fly down through the middle.
Hare Scare has a few interesting mechanics that encourage playing reactively instead of mashing. When you squash a hare, the next time you can attack is 7 frames, but if you strike a hare that's already squished or nothing at all, you have to wait 11 frames. Hitting nothing also generally delays the hares, so to put simply, hitting only hares is optimal.
However, the delay for each hare to come up is random, and is often slow enough that it makes the game impossible. Sometimes at high speeds, it can be bad enough that you never even see one start to come up. Pausing before rounds is heavily used for RNG manipulation, and I often had to do many RNG checks to have the hares come up fast enough. Reaching 75 with no losses is proof that the score can go to 999. For real play though, even getting to 50 would be extremely improbable, not to even mention how tight the timing for execution would be.
Overall, the best way to play is to hope for each hare to come up and never hit anything else. A great score requires both high skill and luck.
Hurry Hurdles is a simple microgame and has no variance. The timing does get fairly difficult as the frame windows for each hurdle start at 14 and quickly goes down to 3 frames. However, the frequency at which you can jump also increases, all the way to every 7 frames.
This microgame is very simple and is just about timing A presses. There is no randomness or variance at all. At the initial speed, the frame window for each heart is about 11 frames, and this slowly decreases down to 3 frame windows. You also only have one chance to lick the blue heart on medium difficulty and the gold heart on hard difficulty. You can lick once every 16 frames.
As the game gets faster, the speed at which you can lick means that on hard difficulty you must immediately lick the blue heart, then the gold heart, and then the pink heart. If you do these out of order, there is no time to complete the round as you can't simply lick them all in a row.
The score can go to 999, but for real attempts, this game gets extremely difficult with needing tons of precise timings and awkward visual cues.
This microgame has quite a bit of jank to it, which explains why the world record score is seemingly low. Every other frame, your jump is much weaker than normal, and this amount increases as the speed increases. Despite there being no RNG calls, this functionally makes just simply jumping luck-based, and at faster speeds, the weaker jump can't reach the bird. Pressing A on the wrong parity of frames is enough to make you lose.
It's also possible to do mid-air jumps and float past the edges of platforms, although gravity reduces how much height you can gain. At high speeds and awkward angles, it's also possible to clip through the bird.
It's quite clear this microgame can hypothetically go to 999, but because of the poor design, real scores will be very low.
This is another microgame that becomes impossible at higher speeds, but interestingly, it's the easy difficulty that becomes impossible due to the long starting delay. That said, the hard difficulty also requires the luck of having one of the bad guys appearing in the middle.
Movements and firing cannot be buffered and must actually be timed. Firing can occur on the first frame after adjusting aim. The game also quickly gets to the point where you essentially must guess and hope for good luck.
I'm not 100% certain that 48 is the absolute maximum, but I couldn't find a way to complete the game at the highest speeds. You can even hear the sound of a successful shot, but still not counting due to the speed.
While this game may seem impossible to complete at higher speeds, it's just barely doable, but extremely difficult even as a TAS, and certainly not at all consistent for real play.
The main obstacle is the ship can move faster than you can, it's easy to get cornered, and the bullet speeds get as fast as reaching you in 3 frames. The only way to avoid shots is to move in such a way that the ship aims at the edge of your hitbox as you move in the opposite direction.
Interestingly, depending on the spacing of the bullets, it's possible for them to go too fast and phase right through you, as one frame it's below you and the next frame it's above you, never making contact.
While the game gets to be very fast, it's always possible to complete. Attacks can't be buffered and must always be tapped to come out. To connect, an attack must be out for at least 2 frames, and switching attacks takes 6 frames. While it's possible to beat the enemies in different order, it's not possible to complete doing this.
Pro Curling has some interesting mechanics to it. You can simply hold A to both buffer and automatically sweep, making execution very easy, with each sweep taking 12 frames. Many rounds are also trivial to complete by simply holding A to buffer a specific amount of sweeps. Some rounds will require performing the last sweep a bit later as the later you sweep, the less distance the stone travels.
What makes 38 the maximum score is that as the game gets faster, there's not enough time to both move the stone far enough and to make it stop in time. The round after 36 is only possible since it can be done with just one sweep, but the two rounds before and after require two sweeps and don't have enough time to stop in the ring.
Guy Scraper is pretty simple and is just a matter of aiming correctly, but after fast enough speeds, the game becomes impossible since there is not enough time to actually land. Because of this, 33 is the highest score possible.
Once the speed of the game gets too high, it becomes impossible to stop the train, even with holding A the entire time, making 42 the maximum possible score.
For real attempts, I recommend the press and hold strategy, and just learning the correct timings since the game has no variance or luck. The following are the frame windows for the later rounds. After round...
29 - 5 frames
30 - Hold
31 - 7 frames
32 - 5 frames
33 - Hold
34 - 4 frames
35 - 2 frames
36 - Hold
37 - Hold
38 - 1 frame
After round 39, the game becomes impossible.
Unlike most of the microgames, Butterfly Stroke becomes impossible after round 33, giving a maximum possible score of 37. This is due to the initial entry not increasing in speed with the rest of the game. All inputs in this TAS are done on the first frame possible.
Many years ago, there were two different TASes that used very different strategies and both got 0:19, so I set out to beat both of them with this TAS. The final decimal is 18.73, and I don't think 0:17 is possible.
Using 2.4 and the correct cutscene, a guard is lured to open both back doors, and then to also pull a Grenade to be used for two of the cameras. Full speed is also charged like normal for the start, but the timing of it is such that full speed is only reached after turning around to avoid unnecessarily running in the wrong direction. The first pause doesn't take place until up to the glass doors so that full speed is maintained for much longer. From there, the rest of the objectives are finished and a lot of nice guard boosts are received.
This was quite a bit more difficult to make than it may look. The Grenade throw has to carefully load the back side so it doesn't fall into the void, and also land in a very precise spot to blow up both cameras. The third camera is also shot through the backside and into the lens since the timing of the cutscene results in it completely facing away when getting to it. The rest of the challenge is getting nice guard luck.
Really happy with the result and to improve over the two very ancient TASes from well over a decade ago!
After picking away at every detail on this level, I managed to get a 14.95, leaving very little leeway! 2.X makes the start much faster, the box boost gives a slightly more direct line, the left route is a bit faster than the right route, and I also got a guard boost to eek out the last few frames. Really happy with this one!
Falcon vs Fox for 99 stocks.
With some precise shooting, tight movement, and an alternate out of bounds, this TAS goes faster than the current Agent WR, and only 2 seconds slower than my Agent TAS. For the out of bounds, the drop spot is almost directly under the the wall, but because of movement being clunky and loading zone issues, it's extremely difficult to jump down into the correct area, and took some effort to reach. Even getting onto the wall is barely possible, and it is definitely TAS only. Interestingly though, it is slightly faster than the outside out of bounds, but the door is locked on Agent, making it only viable for this difficulty.
Really happy with making this one and getting the sub minute!
With some clean movement and aggressive warps, this TAS saves 2 seconds over the current WR at this time. One other subtle trick is during the fall at the end, I use a double crouch to be able to throw the ECM Mine earlier as it otherwise gets stuck on the backside of the ceiling, which helps save a bit of extra time.