Legend of Hero Tonma

This is an improvement of 13.64 seconds over the previous publication, using a variety of optimization tricks, improved boss fights, as well as a new strategy to cut down on the time spent waiting for the screen to scroll at the start of boss arenas.

Stage 1

When Tommy lands from a jump, depending on the arc of his y-position, he can either land directly on the ground or stay hovering a pixel above it. This is useful for positioning the next jump you take, to help fall off of ledges faster or grab items earlier, or even just jump a pixel higher than normal. It's useful throughout the run. Here, it lets us take more optimized jumps for the keys. Additionally, you stop moving horizontally forward for a frame every time you land from being in midair. We avoid this an extra time here compared to the old run by bouncing off of an enemy to combine two jumps into one.

The boss is roughly the same, but we use a new trick to reduce idle time spent before the actual fight. When a boss fight starts, the camera scrolls to a pre-determined position, moving at a relatively slow pace. Tommy's regular movement brings the camera forward much faster, so it's always best to land as close as possible to where it'll eventually end up. This is another trick used throughout.

Stage 2

Like most jumps in video games (and real life), Tommy's jump starts out fast and decelerates over time. Additionally, it takes several frames for him to land from a jump, even if he's inside of the floor at the peak of that jump. This means that if you have to hop up multiple ledges or enemies in a row, it's generally optimal to pick ones that are in the middle of his possible jump arc rather than near the beginning or end, so that he spends as much time with a high y-velocity as possible without needing to stall a ton. Bouncing on enemies is also quicker than jumping, since it doesn't require completing the landing animation. This is all done very effectively in the old run, but we make some minor adjustments. This cliff-scaling praxis will also come in handy for stage 4.

Stage 3

The amount of times landing from a jump was reduced compared to the old run with some clever bouncing. We also skip one powerup, since our secondary weapon is already at its max level by this point in the run.

In addition to the time saved with the scrolling trick itself, its use here also causes us to end up much closer to the boss at the start of the fight, which turns out to be the largest single timesaver in the movie. Double win!

Stage 4

As said, good use of jump spacing is very helpful here to cut off frames. We have to fire off our weapon at specific times to keep the enemies in a good position for bouncing, but it's also necessary to do so to avoid a bunch of lag frames, as this is normally a very laggy stage. Some of the statues we need to bounce on will also fail to load if the game feels as though it's processing too much at once.

The prior improvement for this game saved a cycle on the boss here, and we've cut off yet another. The right pillar man ends up having 8 health remaining when he fades out the first time, and each set of three projectiles that we fire deals four damage total, which means that for an imagined one-cycle, we'd have to hit him with an additional two full sets while also being able to one-shot the left pillar man. As it is, we kill the left one by the very skin of our teeth, so that would probably be a miracle.

Stage 5

Tommy normally moves at a steady pace of two units per frame, whether over land or in the air, but this is reduced to ~1.67 units per frame if he walks off of an edge, hits his head on a ceiling, or moves out of a jump from a standstill, operating in a pattern of 2-2-1-2-2-1. This is sometimes useful for single-pixel positioning, but not for most regular movement, so it's usually good to avoid it. The very start of the stage can't be improved much, since we're dependent on waiting for the shield enemy to come out of his door for our time-saving bounce, but making strategic use of normal jumps instead of slower hovering was very helpful during the later parts. You may also notice that we waste a jump onto some falling blocks near the halfway mark for what seems like little reason, but this is actually necessary to avoid lag that would pop up if we took the lower path.

The boss of this stage throws rocks in various arcs, and this is one of the only important things in the entire game to depend on some form of RNG. This game is generally sync-friendly, but changing earlier parts of the run will inevitably lead to issues here. As it is, the current fight is 3 frames slower than the best one we had at a different point during development, but getting that superior fight would mean stalling for much more than 3 frames before the start of the stage. There also doesn't seem to be any reliable way to affect this randomness after the stage starts. A better understanding of the systems at play here would probably be good grounds for future improvements.

Stage 6

Not much to say here, other than that we use a better jumping pattern near the end. This boss is very annoying - even matching the old TAS was difficult. Trying to figure out the best route to take between damaging the left and right set of four heads is like a less fun version of bowling. We go for the split.

Stage 7

Demon down