I guess the definition of optimization for a run like this would be that you'd want to do things that would be as efficient as possible given a different goal. (For instance, in the NetHack run dwangoAC and I are making, we're up against a huge framerule, but we try to maintain entertainment by accomplishing each individual task as fast as possible even though we could dawdle and still make the frame rule.) EDIT: And a concrete definition: an "infinite loop" TAS could be measured for optimality by how many frames it runs for before it loops for the first time, smaller is better. ("Continues making progress through the game" is harder to define, though.) Still, this seems like a genuinely interesting category due to the planning involved and the knowledge about the game required; you have to have a sufficiently good handle of the game that you know your input will keep working forever.

By entering the warp zone to world 1, doesn't this run breaks its own "always make progress" rule?

By entering the warp zone to world 1, doesn't this run breaks its own "always make progress" rule?

Yeah, but the score(not sure for the "mario life point") still progress, despite it reset at some point.

I’ve been watching it for a while, and I don’t really see a change in enemy pattern…

Hammer Bros have different patterns, so sometimes Adelikat kills one on the bottom and sometimes he doesn't.

I also noticed bullet bills changing pattern, now btw, what the hell is Crown+blue brick lives?

I'll watch it later when I have some spare time.

Is there a version of Tetris where the highest speed remains playable, and the RNG can be wrapped in a reasonable timeframe? Seems like that would be the purest example of continuous progress. It wouldn't be as entertaining as this loop, though.

Could someone upload an encode? It would be much appreciated.

Could someone upload an encode? It would be much appreciated.

Who needs encodes when there's a live stream?

Well, I'm impressed. It's also interesting that the submission module supports infinity as a time.

It's also interesting that the submission module supports infinity as a time.

I assume Adelikat coded that on the fly the best part of this TAS is its stream’s chat, lol

I wonder how many iterations it would take for score/lives/position to sync up such that the exact memory state of the game was identical to a previous one. Once you can show this, you know it will go on forever.

Each loop gains 25150 points and 14 coins. x * 14 = y * 100 (x is number of loops; y is arbitrary) The least common denominator shows that 14/100 = 7/50 so we know that at the end of the 50th loop, coins will be reset to 0 a total of 7 times. Doing the same thing for score: x * 25150 = y * 10000000 Again, the least common denominator for 25150/10000000 is 503/200000. So every 200000 loops the score will have looped 503 times. Fortunately 50 divides into 200000 easily. So at the end of 200000 loops the coins will be 0 and the score will be 0 and as soon as that loop ends (starting with the first input frame of 1-1) we can consider that a "complete run". Each run is about 5800 frames @ 60fps (best I can tell). If I have the numbers right, that's a minimum of 223 and three-quarters days. If anyone knows the frame count and rate better than I feel free to recalculate. RNG is another matter. Each loop is slightly different from the last thanks to Hammer Bros and Bullet Bills. I assume that's all the randomness that occurs here. Otherwise it would require seeing how many times the RNG changes in a single iteration and recalculating. Lives is still going up (according to my calculations it's on loop ~66900 at time of this writing) and I don't know it's upper limit or if there is one. Again, according to calculations he should have at least 9369 lives. If lives is stored as a byte (which it should) then there's apparently no issue with it increasing and looping. We won't know for sure until it gets closer to 32767 at which point a signed word would become negative.

It's STILL going on? I never realized infinity was this long.

Each loop gains 25150 points and 14 coins.

Unfortunately no. Sometimes an iteration will kill a hammer bros for 1000 points, and other times it will not due to the hammer bros randomness Not sure how you would correctly calculate things based on this EDIT: and I can confirm that lives is stored as a signed 8 bit number.

will check back on may 12th to see if still running

Challenge accepted!

Unfortunately no. Sometimes an iteration will kill a hammer bros for 1000 points, and other times it will not due to the hammer bros randomness Not sure how you would correctly calculate things based on this

Oh yeah, I forgot. I just saw it happen, too. That does throw things off. I guess then it would depend on finding the pattern of hammer bros killings (that sounds fun :) ) and factoring that in. Somehow I doubt it's as easy as "one every X iterations" and is probably more like third iteration, then fourteenth iteration, then twenty-first, then thirty-seventh, etc. It looks like the only randomness involved are in 3-1 (the two hammer bros and the bullet bill). I might take the time to count the RNG steps and see if I can find a pattern to it. Sounds like a bit of a wasted effort, though.

Awesome work, Adelikat!

Could it be possible to create a screen saver out of the run?!

By entering the warp zone to world 1, doesn't this run breaks its own "always make progress" rule?

"Progress" in this case would just be defined as moving forward and actively trying to complete the levels, in strict contrast to standing still in order to run up the timer. The levels are being played throughout the video and Mario reaches the end of each one, so it counts as progressing each time.

Could it be possible to create a screen saver out of the run?!

THIS. Edit: Damn, now I'm stuck with this song forever. ><'

♪♫This is the song that doesn't end♫♪

This is probably the most unique April Fools joke I've seen here. Great job, adelikat!

It looks like the only randomness involved are in 3-1 (the two hammer bros and the bullet bill). I might take the time to count the RNG steps and see if I can find a pattern to it. Sounds like a bit of a wasted effort, though.

There's actually a bit more randomness than that, because the pirana plant timing is very very slighty different. I had to adjust my jumps to avoid this problem as well.