Gradius? I know boss fights would be very tough, but it might be fun to see if you can escape enemy attacks, grab power ups, while spamming the shoot button.
Alden, those movies are awesome! I didn't think anyone would consider, let alone attempt some of my ideas!!
Thank you so much for kick- starting this thread and allowing an idea to take shape. Now this pipe dream is a tiny bit closer to becoming a reality :).
Yeah, that's a nice extension of my ideas. Something of this nature would essentially be a big "case" statement involving checking memory values based on what appears on the screen; and hopefully not involve any recursive backtracking, which was the whole reason I proposed this idea.
Basically these heuristics would output button presses that are hopefully constructive and work toward a goal of beating the game.
When playing Mario Bros there is a few simple assumptions that can be made about it.
1. First the left+right stand jump should alway be used at the beggining of each new area in order to get the fastest inital speed.
2. After that B should always be held, with exception to circumstances where it would be theoretically quicker to go through a wall.
3. When entering a new area Mario always heads right.
4. Up button is never required.
5. A general guided path is required, e.g Mario needs to enter a pipe or reach the goal on the far right.
Of course to get something which could remotely TAS SMB efficently would takes 100's of rules and millions of re-records, but the rules would prevent the Bot from testing Billions or even trillions of pointless input.
I don't know if you even read my post, or were just musing, but the whole point of my idea was to eliminate the idea of re-records, and just complete the game.
I simply was speculating about how to do this, and using an "on the fly" approach with rules seemed to be the best way. Speed is not my primary goal. With better rules, I'm sure speed could be improved, but that's another story.
Yeah, I can see the rules become too "level-specific" and the rule list being pretty long, but just seeing a level being completed is enough to interest me. For example, it seems (and I am probably very optimistic) that world 1-1 of SMB is very forgiving, and it would be very satisifying to me to see it beaten with a small subset of rules required to beat the whole game. How forgiving are the rest of levels? No clue, probably not very.
I guess the main argument is that if the rules become too numerous and too "level-specific" you'll essentially be approaching the "hardcoded button-press" approach which I vouched against.
I also expect this to be a "trial and error" approach in programming, meaning that you would run your heuristic program, and see where it fails, and then add a rule to make sure it doesn't fail (and hope this rule is generic enough to apply to quite a few situations).
Once again I'm just brainstorming.
I was just brainstorming, and came up with a novel idea. If this has been discussed before, an admin can move/lock this topic.
Would it be possible to beat a game/level (let's take SMB for this example) by using heuristical/rule based methods? Note: I'm not asking to complete the game in the fastest time possible, but rather just beat the game. Rather than "hardcoding" button presses, it would be nice to see if you could just follow "rules of thumb" and see how far Mario can get.
For example, here are a few rules that you could potentially use to beat level 1-1 of SMB (and maybe more). This list is by no means exhaustive, or even correct.
Default: Hold the right (thanks NrgSpoon) key (and run in most cases)
1: If there's a pit, jump over it
2: If there's a goomba/turtle on the screen, jump on it or avoid it
3: If there's an obstacle in the way, and no enemy is present, jump over it
4: If there's a pit and then an enemy...
5: If there's an enemy and then a pit...
...
and so forth
Of course, implementing these rules is definitely "easier said than done." You would require the use of memory address to keep track of Mario's speed, position, and figure out a way to locate pits and enemies (and their various positions).
Actually, it would be VERY interesting to see how well a very rudimentary set of rules could perform, provided that each rule could be implemented.
Computationally, this should be very feasible, as there should be no depth/breadth first searching, as code should be generated on the fly (using the rules), and you wouldn't be able to "look ahead" by using a save-state.
I might be reaching too far, but ideally, the best set of rules should be able to clear an arbitrary arrangement of obstacles, provided that the scenario can be beaten.
If you don't understand what I am trying to convey, please forgive me. It is late, and I am tired. Also, I am not volunteering to search SMB for memory addresses, and write the code to do this. I am merely throwing out an idea that could be interesting/fun/cool to other members of this site. It sure seemed interesting to me.
After watching the Link's Awakening run, I was eager for more 2D zelda action. This may indeed quench my thirst.
Looking forward to seeing your progress, SwordlessLink. I wonder if there are any exploitable pit-glitches to be found in this game. Good luck.
I had a thought for the any%. When fighting a boss, you often reach a point where the boss can't be hurt, but you can be, right?
So, during one of these points, take damage with timing such that Baby Mario has 0 stars left when the boss dies. That way, the score screen goes faster. The entertainment comes not from the extra speed, but from knowing attention was paid to such a fine detail.
The main problem with this is not the crying, it's the fact that you lose time when you retrieve baby mario. The screen freezes completely for about half a second. It's like jumping on yoshi in SMW.
It also looks stupid.
Um... In Misty Star World..Where the heck did that fish come from???
I second this reaction.
If I had a "WTF" stamper, I would apply it to this movie.
I had a lot of fun watching Misty Star World and Pipe World 4 particularly, .
Xkeeper wrote:
There was a movie of SMW a while ago demonstrating that weird property of P-switches. Namely, if you play with them just right, a fish will pop out...
This isn't a DWTLC-specific bug, in short.
DK64_MASTER wrote:
jaysmad wrote:
Um... In Misty Star World..Where the heck did that fish come from???
Just so I don't have to repeat myself again :P.
I had actually found that glitch a few months before than what I posted. Mrz had mentioned something about being able to step on a pswitch, and then take it down a pipe, and keep it intact. So I was able to do this with yoshi. Also to my surprise, a fish magically came out. Now, I wonder if they will actually come "alive" if this is done near water?
a few problems:
1) Good luck implementing the AI to detect such a sequence.
2) If you ever hang around the IRC channel, you'll find that messages such as these are often ignored, scorned, and are unecessary. We don't need constant reminders to tell us that we should be bug hunting when we are just want to idle, or are discussing other matters.
3) Often there are only a few active members in the irc channel. The chances that a message involving tas strategy that would concern said members are quite slim.
4) If there IS TAS talk on the irc channel, chances are that the current discovery is already being discussed.
5) Who checks IRC history? A static message posted in a forum is MUCH MUCH more likely to be noticed rather than a fleeting message by NVA.
6) The idea is stupid (read points 1-5), and will clutter up the IRC channel.
7) Also cocks.
The idea about NVA announcing comments on IRC is pretty bad, IMO. I definitely think that the comments should be addressed to specific TASers, and not every runner is on IRC.
Also, remember that there are quite a few members on the irc channel who don't TAS at all. The new topic indicator by NVA is sufficient enough, and anything else would be overkill and, not to mention, extremely annoying.
I'm finding this fairly easy to do, one problem: The emulators desync away from each other. So when I watch the movies again, messed-up things happen. How can I fix this? If the only solution is hitting frame advance very slowly, I think I'll just give up on this run...
Giving it a shot, here I come! My completion time goal is 12:00. I figure SMB will be about 6 minutes, SMB2j maybe 9, SMB2 about 10 (because boss-fighting won't be too optimal).
So what do you plan to do with the 6 minutes of down time in smb? Expect us to stare at a blank title screen? Maybe you'd want to complete some more levels instead.
I've seen a lot of the little tricks that have been presented in this run. Each of them mindblowing in their own right, but when you string so much greatness into such an optimized package, I can't help but smile, and gaze in awe at the amazingness of this run.
A huge yes vote, and YAY for yoshi fishes!!
And for those who don't understand the file-name, try reading the submission text :P.
Very nice job guys. And...I thought the SM64 wars were over. Who knows what Rikku/you guys/someone else will come up with for the next submission?
A yes vote, and give it a star please.
A month after last April Fool's day, it's still confused as to what the rules are as to what makes a publishable April Fool's day run. What are the formal rules for these joke submissions?