I'd think a run using Luigi would have a better chance of getting published, since his physics are different in this version. As it is, this was enjoyable, but not different enough from the published SNES run. Voting No.
I forgot about luigi! I'd like to see a luigi run.
I didn't like it. It was like the SNES version but less glitchy (no smashola, no passing through ramps) and less screen real estate. I do like the GBA version's redition of the castle and bowser boss music though :)
This approach only makes sense for a game that isn't hardware-dependant, with a well defined semantical model. Starcraft 2, for example, can store replays as events because it can guarantee determinism. A game as simple as megaman is wrought with oscillating speeds, emulator-dependent amounts and positions of lag and a tiny chance of executing arbitrary logic (which is again emulator dependent and not part of any well defined model).
As you've said, bots could be used to do short sections of the game for you - these have already been made and utilized. Why put the cart before the horse? Decide you need a bot that does X, then code everything that the bot needs.
Admittedly, I'd love to see the 'higher level movie' for an RPG like final fantasy. Instead of measuring durations in time, you'd timestamp things like 'at RNG #FE, attack CHAOS with BIGGS'. In games where RNG manipulation is easy (advancing it as fast as possible is always trivial) it would be a compact way of describing the run.
I thought the sonic run focused on in-game time over real time, as evidenced by ignoring the huge amount of time tallying up a time bonus takes. Why can't we ignore this too?
Has anyone considered using artificial neural networks? It's an interesting method to create A.I.
A neural net needs 1 or more inputs, processes the date and generates 1 or more outputs. The advantage of that would be, one won't have to program the A.I. itself (only the neural net), but rather, it 'learns' what it has to do. The more neurons available, the more it can learn.
However, my attempts at creating such artificial neural nets weren't very successful, and they might not at all be suited for playing or TASing games, but I think they still are worth mentioning.
Neural networks do not make a complete bot, but they can be used to discover the ideal way to tackle an isolated situation with a finite number of variables. For example: http://robowiki.net/wiki/Neural_Targeting
One last thing: have you considered delaying each room until the *last* sprite exits the room? This might make it easier to follow, and easier to spot the amount of improvements going on.
Ah ok. Game runs now, thank you so much! Can you control the speed at all? The game runs really slow, which I assume is normal since TAS is never at full speed, but it's like less than 50% speed.
That's normal for this emulator. It's emulating the processor dos games run on and not very optimized.
The creator of this game actually suggested doing this on the TAS thread I made in the GMC. This could be the first TAS that a game developer has actually endorsed. :P
Ultimortal TASed his own game, Hero Core, to make a video reallyjoelsdad difficulty, and later commented on the video nitsuja made using hourglass.
I think in a few isolated cases programmers on games have been sought out for help with understanding how a game works to improve a TAS (am I crazy?)
I already managed to solve these problems, but they were interesting to solve, so I'll present them here:
1) I have a string of n bulbs. No two adjacent bulbs can be on. How many possible strings are there?
2) I have a string of n bulbs. Bulbs can be off, red or green. If a bulb is red, the bulbs adjacent can't be red. If a bulb is green, the bulbs adjacent can't be green. How many possible strings are there?
What can you create? http://www.bemmu.com/music/index.html
Me and some people in an IRC came up with examples:
100*sin(Math.log(10*t)^Math.sqrt(10*(t%1024))<<8)
100*sin(Math.log(10*(t%9))^Math.sqrt(10*(t%1024))<<8)
100*sin(Math.log(10*(t%9))^Math.sqrt(10*(t%1024))<<(t/1024))
100*sin(Math.sqrt(t)^Math.sqrt(10*(t%1024))<<(t/1024)) <-- this is a good beat
100*sin(Math.log(100*(t%1024))<<(t/1024))+50*cos(Math.sqrt(100*(t%1024))<<(t/1024))
(128*sin(Math.sqrt(10*(t%1024))<<(t/1024)))/cos(t>>9)
100*sin(Math.sqrt(100*(t%1024))<<((t/1024)^(t/768 )))
sin((t%8192)>4096?tan(Math.sqrt(t%4096)):Math.sqrt(10*(t%1024))<<(t/1024))*100
(128*sin(Math.sqrt(10*(t%1024))<<(t/(1024*tan(t/1500>>7|8)))))/cos(t>>9)
-----
50*sin(Math.sqrt(100*(t%1024))<<(t/768))+((t%8192)>4096?50*sin(Math.sqrt(100*((t)%1024)<<(t/1024))):0)
50*sin(Math.sqrt(100*(t%1024))<<(t/768))+((t%8192)>4096?50*sin(Math.sqrt(100*(t%1024)<<(t/2048))):0)
just the new part by itself: 50*sin(Math.sqrt(100*(t%1024)<<(t/2048)))
or with a backing beat: 50*sin(Math.sqrt(100*(t%1024))+(t/768))+((t%8192)>4096?50*sin(Math.sqrt(100*(t%1024)<<(t/2048))):0)
----
100*sin(Math.sqrt(10*(t%1024))<<((t/512)%1024))
t*(t>>8&t>>12)
t * ((t>>12|t>>8)&63&t>>4)
