SM64 ABC History Part 6 Additional Notes

Note 1 (2:00): It also calls RNG to determine a short waiting period between swings, but it's irrelevant in this case.

Note 2 (2:45): The force shown here is not an accurate force acting on a swinging pendulum, but it accurately represents the effective force used by the game's physics. The pendulum experiences a fixed angular acceleration of 13 on a slow swing and 42 on a fast swing, which is different than what would be observed in reality.

Note 3 (4:04): Because there are only two possible speeds (13 and 42), and because the same speed twice in a row returns the pendulum to exactly the same point, there is no choice other than to alternate between slow and fast until the middle point crosses angle 0. So, there is no way to make that any faster - it simply happens to take 291 swings.

Note 4 (4:58): The function has a couple of failsafe mechanisms that prevent any small closed loop from existing, but also incidentally shorten the loop a little bit due to an overzealous failsafe. On paper, the RNG function would have a loop of length 65534 and a loop of length 2, but in practice, the long loop is cut a bit short. More information: https://www.youtube.com/watch?v=MiuLeTE2MeQ

Note 5 (7:49): STROOP was initially created by danebou and came into existence on June 13, 2016, but pannenkoek took over the project for most of 2017, 2018 and onwards.

Note 6 (12:18): Some of you may have noticed that since no hat in hand is available and because instant release drops the bob-omb, there should be no way to bring a bob-omb directly to a preset HOLP in TTC. This run used a trick found in 2021, which will be talked about later.

Note 7 (12:47): The strategy was invented by ds273 and executed for the first time by BillyWAR.

Note 8 (18:42): While the gradient of the slope is important, the friction coefficient is actually the biggest factor. The floor triangle needs to be of an unusually low-friction type, and/or the slope especially steep, for this trick to work. The reason it works on the stalactite in JRB is because when Mario is half-submerged in water, he won't slide off from steep surfaces, so he can punch.

Note 9 (20:09): The specific operation that causes the error on Wii VC was converting from a double (64-bit) float to a single (32-bit) float. SM64 typically doesn't use doubles, but a small number of calculations do use them - for example, the oscillation of the BitFS platforms, and the computation of normal vectors of triangles. The glitch is unique to Super Mario 64 on the Wii VC. It doesn't happen on Wii U VC, or for other games on Wii VC.

Note 10 (25:52): Also, it's 230 forward speed. Forward speed is much more difficult to maneuver with than negative speed, because forward speed is hard capped on a lot of actions for which negative speed is only soft capped, including diving and running/walking. Therefore, Mario can only turn using text redirection.

Note 11 (28:43): The sign is reversed here because for the multiplier to be applied, the stick needs to be held back by some amount. If the stick is held back, the stick angle in relation to Mario's facing angle will always be between 90º and 180º, so the cosine of that angle will always be a negative coefficient. Also, "The multiplier goes down by 1 for every 10,000 speed" is assuming a stick angle of 180º. In reality, it goes down by one multiple of the cosine of the stick angle for every 10,000 speed.

Note 12 (29:54): His speed doesn't get capped to -16 because he bonks after a partial movement of just one quarterstep. While the speed doesn't immediately get capped, it does get reversed to -2339. This is not a problem for twirling, but the 10k glitch requires positive horizontal speed, which is yet another reason why a bonk could not have been used to land on the shell.

Note 13 (31:00): It's actually sort of a big deal. None of the common cloning glitches can be used: at 0 stars, there's no wing cap for hat in hand, and the clone's destination renders transport cloning and instant release useless. Pannen and Tyler had to resort to Pause-buffered hitstun (PBH). Just wait a second to see what that is: it's simpler and more natural to introduce that glitch on Bowser instead.

Note 14 (33:17): Mario needs 45k speed to get out of the hyperspeed punching spot directly into a freefall, which is why the 10k glitch had to be used after this point to reduce Mario's speed to a workable amount. This is why he has to slide, fall, slide again, and then twirl.

Note 15 (35:36): Technically, he has 3 frames, but in reality, he can't use them up very well.
	- Frame 1: Mario enters freefall, which cancels the remaining quartersteps of movement, leaving him only one quarterstep of movement to work with.
	- Frame 2: Mario is in freefall with 0 Y speed. Because of quarterstep spacing issues, he uses up only three out of four quartersteps of movement.
	- Frame 3: Mario now yas -4 Y speed, so he can only move by one quarterstep before he's too low to reach the pillar.
In total, he has 6 quartersteps of movement available to him, or the equivalent of 1 ½ frames, and he travels for 5 of them.

Note 16 (36:56): If this were out of bounds, Mario would lose his horizontal speed here, but thankfully, ceilings only kill vertical speed and conserve horizontal speed.

Note 17 (39:26): Mario's max acceleration is 1.15, but drag takes away 1 speed per frame if his speed is above 32. Because the stick needs to be held at a slight angle, the actual acceleration will be lower than 0.15. The sharper the angle has to be, the lower the acceleration, which limits the top speed.

Note 18 (39:37): This top speed is a factor of the floor steepness, friction coefficient, and the angle the stick needs to be held at.