Mega Man 5 Theory TAS in 35:12 notes

Mega Man 5 Theory TAS in 35:12 by BaronHaynes
http://www.twitch.tv/baronhaynes/c/5017981

Goal:
Create a playthrough representing the realistic human limit for Mega Man 5 with the current route (collecting Beat).

Parameters:
1. Virtually all real-time gameplay, done without frame advance. The exceptions are manipulating RNG and setting up easier savestates, which doesn't affect the final time.
2. All feasible strategies a human player can do, even if it's not currently done. This one is subjective and I think I had mixed success with it.
3. Above average RNG, including the extra arrows from lucky weapon drops in Gyro and Charge Man.

How did I do:
It's a decent start, but not as consistent as it could be. There are movement errors, some slow menus, strats I missed or didn't understand, and somewhat varying standards for what I accepted in one room over another. I attribute this to putting it together quickly. That was part of the idea, though -- I knew I was going to make mistakes and overlook things, so rather than spend weeks getting it perfect, I made something a little rougher as a starting point. Other runners watching the playback caught things that never would've occurred to me no matter how long I spent on it. I will probably improve on this, but I think it works as a beta; I think it gets most things right.

======================
Individual Level notes
======================

Star:
1825 - Getting this hop over the turret without stopping is incredibly precise. I included it because I've done it once, but it's kind of a fluke occurence.
2900 - Similar story here, although I've done this one more times in runs.
4500 - I'm pretty sure getting a fast knockdown from the torpedo is faster, but I couldn't get it and didn't spend a lot of time trying.

Gravity:
9482 - Movement error.
9800 - Getting the damage boost/gravity flip here involves incredible timing, but with practice I think it's worth going for in runs.
11766 - You can do a very small hop under these spikes and use Star to kill the enemy, it's slightly faster than a charged buster but you run an extra risk of losing time to a drop.

Gyro:
Not much to say. The extra drop supposedly saves around 40 frames, but I haven't done comparisons myself.

Crystal:
23800 - This is the fastest way I've found of getting through the first room, although you can do an extra partial slide at the end.
24380 - I've tested this room several times and this method seems to be faster than firing the arrow from the left wall. I've had 7 ammo with this method more than once, although doing an extra arrow jump on this ladder may not be a good option, since realistically you won't have time to see your ammo count until after you fire it. You can use two arrows on the next screen instead and save about as much time.

Napalm:
34280 - The alcove drop is about 10-15 frames and not quite as bad as it looks. You can decide whether it's worth trying or not.

Stone:
Getting a good pattern from this guy isn't any easier in TAS than in real time :(

Charge:
This chicken room is dedicated to Glitchman, with an additional shoutout to Kzez.
50000 - The first two chicken kills are nearly frame-perfect without stopping, and I savestated heavily to get them. My route here is a bit slower than other MM5 runners, so I wasn't sure how to realistically approach this.
50819 - I believe this is the ideal rat to get a weapon drop from, so you can arrow from the A to the ladder (thanks bjw)
50953 - Slide-jumping to get this letter is a 5-frame trick that's also dependent on the room's bouncing animation. You can conceivably do it in runs, if your chicken room is consistent (and if the bouncing starts when the room first loads, which I'm not sure about either).

Wave:
57250 - Jumping over the spine is faster than using an arrow, the arrow introduces more lag frames than you save from sliding on it.
58285 - Two arrows are faster than 3, even without a perfect jump at the screen transition. It's about 8-10 frames with a good jump.

Dark 1:
69744 - It's not worth hopping after landing on the other side of this turret. There's not enough room to do an efficient jump, so you lose time over just walking off the edge.

Dark 2:
73800 - Double gap jump is pretty silly, but it would save 15-20 frames if someone got really good at it.
74500 - I'm not sure if my path through this room is the most efficient one, but I think it's pretty close. This is another case where it wasn't what I normally do, so I had to improvise a little.

Dark 3:
82263 - The no-hesitation slide-jump off the 1-block pillar here might be stretching the limits of "realistic play". A little.
85821 - You shouldn't pause to switch to Gyro early as I do here, but as late as possible without getting hit by the enemy. The arrow moves faster than sliding.

Dark 4:
The tower is pretty good, but it's possible to have a better boss pattern.

Wily 1:
99778 - This is about as close as you can cut this chicken room, and again stretches "realistic" a bit. But it's possible.

Wily 2:
103967 - I don't expect anyone to try this slide cancel. It's one of those things I spent enough time playing with that I started to believe "well, someone might add it eventually."
104817 - Does anyone do this slide + jump without stopping? I wasn't sure, but I kept it in anyway.

Wily 3:
Wave - Yeah, this one is really unrealistic. Hitting him without taking damage is a 1-frame trick. But avoiding most of the damage in this fight is possible.

Wily 4:
122908 - D-boosting past this walker is 10 frames faster than arrowing over him. That surprised me.
123400 - Sliding directly into him is faster than taking a backwards d-boost from the laser, unless you do the latter perfectly as seen here.
Capsule - getting 2.5 off of a free hit start is very hard and luck-dependent, but it can set you up for a 6-cycle buster finish. I actually didn't manipulate his final spawn, it just happened. Whether it's worth attempting or sticking with a sure thing is up to the runner, but cutting a cycle off this fight would be pretty sick.