Times in bold indicate an in-game timer improvement, with the number of time units improved represented by the number of asterices next to them.
Levels marked with a # are entirely unchanged from the prior TAS.
General Improvments
Shortly after the previous run was submitted, a subtle trick which can be used to reach top speed 5 frames faster from a standstill was discovered -- on the ground, Sonic accelerates at a rate of 12 per frame (this is increased to 24 while in air), but he brakes at a rate of 128 per frame. So pressing the "wrong" direction for a frame, and then immediately reversing it causes sonic to have a speed of 128, instead of the 24 (or 36, given a jump to increase acceleration) that holding that direction for two frames would achieve. This was used to shave 3-4 frames off of most levels.
Some further improvements were achieved by better optimization of rolling at the start of slopes, and downslope jumping for the speed boost.
The improvements were performed by Upthorn, but levels, and level segments, which could not be improved, were left as they were in JXQ's recently published TAS.
Specific improvements
Green Hill 2 - New strategy of zipping from the wall to the goal discovered by Upthorn and optimized by JXQ
Marble 1 - A freak accident while attempting to gain back frames lost due to a bat desynchronization lead to the discovery of a four second shortcut.
Labyrinth Zone 3 was modified very very slightly to recover from a lag variation. You probably won't notice the difference.
Starlight 3 - Better optimization of slope roll and slope jumping allows for some sick speeds here.
Final Zone - A method similar to what is used here to avoid being crushed allows the player to fall through the floor and die.
Thanks to
Adelikat, Fabian, Moozooh and Xebra for their tireless words of encouragement over the long and hard five days that went into making this run.
i noticed the sound desyncs from the avi every so often (most likely a codec issue) it every so often gets ahead of the video... um and the quality in the knuckles run is noticeably better then the sonic one.
Sonic accelerates at a rate of 12 per frame, but he brakes at a rate of 128 per frame. So pressing the "wrong" direction for a frame, and then immediately reversing it causes sonic to have a speed of 128, instead of 24
This is almost useful information, but unfortunately there is a problem with your units. The first part should read something like 'Sonic accelerates at a rate of 12 "x" per "x" per frame.' The "x"s need to be filled in, otherwise this statement really isn't informative. Likewise, when referring to speed, it would be more helpful to say 'a speed of 128 "x" per "x".' Again, with the "x"s replaced by real units.
I'm curious to know a little more about the physics of the game, and accurate units would be helpful to that end. Thanks!
The first part should read something like 'Sonic accelerates at a rate of 12 "x" per "x" per frame.'
The number is the amount by which the RAM address that holds the velocity value changes. RAM addresses do not contain information about what units they represent.
Also, since you know from context (indeed, it is explicitly stated multiple times) that the information pertains to a speed value, the units do not matter, whether it x was inches or miles, the ratio, and therefore the time savings, remains the same.
Also, your example template of an acceleration unit is incorrect and incomplete, a correct example would read 'Sonic accelerates at a rate of 12 "x" per "y" per ("y"/"z"),' where x is a unit of distance and y and z are units of time.
Now I'll satisfy your curiousity by explaining that the stored velocity value represents subpixels per frame. There are 256 subpixels per pixel, but a subpixel is not 1/256th of a pixel. A change in position by 12 (or even 255) subpixels does not actually constitute motion from the perspective of the game engine, unless it overflows the subpixel value, increasing the pixel position.
However, in the future, I would appreciate it if you expressed your curiousity by asking a question, rather than incorrectly telling me that I had made a mistake, and that my information was useless. I had deliberately chosen not to attach units to the numbers because (1) RAM addresses don't contain information about what units the number represents, and (2) Explaining subpixels midsentence would be distracting, and, conversely, make it more difficult to understand the important point, which is that the speed jumps from 0 to 128 in two frames instead of five to eleven. (I will, however, now edit the submission text as I forgot to point out that Sonic's acceleration is doubled while in air).
How fleeting are all human passions compared with the massive continuity of ducks.
Now I'll satisfy your curiousity by explaining that the stored velocity value represents subpixels per frame. There are 256 subpixels per pixel, but a subpixel is not 1/256th of a pixel. A change in position by 12 (or even 255) subpixels does not actually constitute motion from the perspective of the game engine, unless it overflows the subpixel value, increasing the pixel position.
However, in the future, I would appreciate it if you expressed your curiousity by asking a question, rather than incorrectly telling me that I had made a mistake, and that my information was useless. I had deliberately chosen not to attach units to the numbers because (1) RAM addresses don't contain information about what units the number represents, and (2) Explaining subpixels midsentence would be distracting, and, conversely, make it more difficult to understand the important point, which is that the speed jumps from 0 to 128 in two frames instead of five to eleven. (I will, however, now edit the submission text as I forgot to point out that Sonic's acceleration is doubled while in air).
That was actually pretty interesting information, I'm glad tmnt9999 asked.
If I ever finish the Super Mario Kart run, I'm going to claim the units of speed are in 7E10EA-2u's :)
Edit: Oh, and great job Upthorn :D
Wow! Nice finds. Those are no small cuts, and I daresay the movie files didn't even look fully optimized. Where do you think more cuts can be found?
Indeed. These definitely warrant a new TAS
Preliminary testing has brought MZ2 down to 0:44, and SBZ3 down to 0:10
When I improve the TAS, I'll check for any other places these techniques might be usable.
How fleeting are all human passions compared with the massive continuity of ducks.
carretero, you're a damn machine :)
Isn't it about time these glitches were combined to improve the run? I'm guessing a sub 15 minute run is now possible?
