This is a new Rockman TAS, mostly an update to the previous version though.
Changes:
Cutman battle improved by 4 frames. Thanks to Vatchern!
Wily1-Wily4 (excluding Wily4 refights) improved by about 16 frames.
One more refill in Wily2. Costs 20 frames. Wins a couple of seconds in Wily4.
Bombman refight glitched, death. Resulting lots of gain. Other refights skipped totally. Thanks to Arne the Adjective and FinalFighter! (Thanks to Arne again for reminding me about FinalFighter's refined technique to this.)
New attribute: Uses death to save time.
90% of the last battle was played by BisqBot. I created a new algorithm core in the bot that simulated literally what a human would do in such situation; only saving lots of work. (Previous versions of BisqBot have always used random input combinations.)
AngerFist's contribution to this movie comes from his part
in the previous version. He played the Gutsman fight and
the Bombman fight (the first), and he prevailed at luck
manipulation in Cutman stage, and helped optimize
a couple of rooms.
Ps: The rerecord count is arbitrary. At some point, it was going to say 6268371 -- somehow the value was accidentally copied from BisqBot's progress file. So I edited it manually.
Ah, indeed. BisqBot did play some parts of this movie. Specifically, the 14-frame improvement in Wily1 is thanks to BisqBot, and the Bombman refight escape technique was perfected by BisqBot. And of course, he manipulated the refills in Wily2 and beat CWU-01P. And, played 90% of Wily4 battle.
As for Bombman refight escape... I disassembled Bombman's AI entirely. That way, I knew exactly how to make it throw a bomb and how to make it jump a certain height of jump.
Edit: Submission file replaced! (15:52 --> 15:49)
Edit 2: Submission file replaced again! (15:49 --> 15:48)
Edit 3: Round numbers are nice. 56900 frames is the length now, making it 1289 frames (21.5 seconds) faster than the published movie.
Ps: This is a bigame movie. It completes on both Rockman (J) and Mega Man (U).
An improvement this great to a game that is already pushed so far is just incredible. I do hope you get to the bottom of what you felt was a loss of time. Yes vote and great work the both of yous.
Wow, that boss skipping trick is pretty funky.
The highlight for me was the very short and very weird Cutman battle though, I think. Something from the next few frames after selecting 'G' would make an excellent screenshot.
BisqBot...I'm sorry but I don't like the idea of that tool. 6 Million...you know how long it would take a human to do that? Isn't a human suppose to be the one playing the game, not the program? Granted it does help with complicated stuff, but what's the fun in that? No one but that bot can play the game for randomness now, no one will get close. It's tool ASSISTED, not tool RELIED video...just my opinion.
The point of a TAS is to produce results as close to possible as what a robot built to play games optimally would make. Wouldn't you say it makes perfect sense to have a program TAS for you as much as possible?
What, you don't find horrendously glitched runs fun to watch? I certainly do.
Same, thats why I love the lttp runs, plus the game is still fun to play.
DonamerDragon wrote:
It's tool ASSISTED, not tool RELIED video...just my opinion.
It is tool assisted, they didn't use the bot the whole way through the movie, so I don't see how it was tool relied. In that sense actually, wouldn't using the frame advance and the emulator itself be relying on tools, not being assisted?
*Waits to be corrected and burned*
Voting yes because of the new changes :P
The point of a TAS is to produce results as close to possible as what a robot built to play games optimally would make. Wouldn't you say it makes perfect sense to have a program TAS for you as much as possible?
No, like I said we are doing tool assisted videos. We always assure everyone that we're not cheating, I think having a program doing for you is just as bad as using a game genie. You put in the code and you get a desired result. TASing is suppose to be fun, and done by a person who is playing the game....
It's tool ASSISTED, not tool RELIED video...just my opinion.
It is tool assisted, they didn't use the bot the whole way through the movie, so I don't see how it was tool relied. In that sense actually, wouldn't using the frame advance and the emulator itself be relying on tools, not being assisted?
Maybe so, but a human player is still playing. A human player is still figuring out randomness, if the human player is still playing the game, not a program. I don't want to start a huge dramafest, it's just how I feel about the nature of TASing...I just think a human player should be doing it entirely, even if he is using frame advance (in my opinion, like a frame number ruler).
Hmm, why is it that the Select-tapping to beat the second Wily form sounds faster at the end? Can you make it that speed the whole way through? Is that why you lost time?
Unpausing can be done by both Start and Select. Starts brings up the weapon menu. However, when a shot is on the screen, the weapon menu cannot be displayed, and thus, the game continues immediately. With Select, there's the transformation delay. The penalty? With Start, the onscreen objects get a frame of actual life (hence shots can move past enemies so that they won't hit anymore). With Select, they stick in place. (However, Wily's carriage defies this logic because its movement is programmed in NMI, not object code.)
> I think no one.
Wrong.
I just made it a bit unpredictable.
FreshFeeling wrote:
The highlight for me was the very short and very weird Cutman battle though, I think. Something from the next few frames after selecting 'G' would make an excellent screenshot.
Here are some shots Dehacked took. Can you guess the technical reason behind these?
DonamerDragon wrote:
We always assure everyone that we're not cheating, I think having a program doing for you is just as bad as using a game genie.
We're still only going as far as the game lets you.
And like GeminiSaint said, someone needs to program the bot.
For instance, in the Bombman refight, what BisqBot did was only: after I luckmanipulated Bombman to first do a tall jump and maneuvered Rockman to take damage from Bombman while throwing a bomb himself, it pressed Start randomly for ~200 frames and repeated that ad infinitum searching for the fastest way for scrolling up to happen. (Pressing Start more often yields more lag, hence slower, but pressing it less often may involve Rockman not levitating high enough to scroll to the next screen. And there's Bombman's actions (luck manipulation) to consider; his throwing a bomb is essential for the trick.)
I could do that manually, but it would be an utter, unmotivating waste of my time. That's what tools are for.
if the human player is still playing the game, not a program. ... I just think a human player should be doing it entirely, even if he is using frame advance.
The thing is, the program is just doing what the player is telling it to, it's not like it's playing the game on its own like a person would. A human could likely duplicate those feats using less rerecords (and more time), but why bother if it's something so simple to describe that you can automate it?
Saying that is like saying you're not really playing the game if you use autofire or other macros because that puts in input for you, or that hex-editing your movie isn't really playing because you're not actually recording those frames. Those could be true, nobody claimed to be playing the game and it's probably not correct to call it "playing", but since when has that been a problem here?
DonamerDragon wrote:
Isn't a human suppose to be the one playing the game, not the program?
No, that's never been the point. If a program could beat the game it would, but it just happens that humans tend to be much better than most programs at finding fast ways to beat games.
I don't really see what's bad about figuring out what the best solution ought to be, even if it isn't humanly feasible to find it, then figuring out a way for that result to be found another way.
Be it programming a robot to troll for the solution or what have you.
I've never TAS'd Mega Man, but I would bet that if you had to manipulate three big items drops within the spam of about 40 frames (like in v7 I think), each having a 2% (again I could be wrong) chance of appearing, the fun part of TASing would take a backseat to your ever-increasing rage.
I see building a bot like hiring a big mobster for your TAS. "Oh you won't give me what I want? Well, I'd like you to get acquainted with this bot I have here. I think you'll get a long real well. I'll be back in eight to ten hours to see if you've changed your mind."
<Swordless> Go hug a tree, you vegetarian (I bet you really are one)
Keeping in mind that I don't know anything about low-level Rockman mechanics, the two Cutmans somehow remind me of the way that, in RM2/MM2, Rockman can stand next to Wily 3's boss while detonating a crash bomb to damage it. Are they similar, um, phenomena?
Keeping in mind that I don't know anything about low-level Rockman mechanics, the two Cutmans somehow remind me of the way that, in RM2/MM2, Rockman can stand next to Wily 3's boss while detonating a crash bomb to damage it. Are they similar, um, phenomena?
I don't know about the Rockman trio in Iceman battle, but the Cutman duplicate wasn't of that kind of origin at all.
The previous run already had a lot of Bisqbot luck manipulation, didn't it?
At any rate, I'd vote yes to this run, just as I would have the previous run. I was going to say that I hoped there weren't too many improvements to this run, because I didn't really enjoy watching the same run over and over again with very slight improvements, but then I actually watched the improvements and they were ridiculously cool. So scratch that, keep on improving until you've got it down to 13:37 or something.
I am not sure if this is considered a full run with all the debate going on about concept demos and stuff.
Since when has a "full run" been a requirement for a speedrun (tool-assisted or not)? Regular speedruns skip *a lot*, usually by abusing some design mistakes in the game, if not even bugs (just look at QdQwav, and especially the HL2 speedrun), yet they are still speedruns.
I don't see any problem with a run skipping the route that the game designers intended being considered a normal and regular TAS.
The vast majority of videos published here complete the game by running through the intended route, simply because there are no glitches to exploit. It's refreshing to see at least *some* runs where things can be skipped in a cool way. It would be a bit boring if there would be no such runs.
BisqBot...I'm sorry but I don't like the idea of that tool. 6 Million...you know how long it would take a human to do that?
If you want to watch games played by humans, go to SDA. There are plenty of those there.
The whole point in TAS runs is to try to find the most optimal way of completing the game using *any* available tools. What those tools are is irrelevant: The goal is to find the optimal list of keypresses to complete the game in the minimum number of frames.
We always assure everyone that we're not cheating, I think having a program doing for you is just as bad as using a game genie. You put in the code and you get a desired result. TASing is suppose to be fun, and done by a person who is playing the game....
What is the difference between using savestates+frame advance, and coding a program which searches for a short optimal path? How is the latter more "cheating" than the former?
It isn't. Both are tools, both tools have been created by someone, and both produce the exact same legit result: A series of timed keypresses. I don't understand why using a program would be any less "acceptable" than using savestates.
We have never claimed that the videos have been "played by a human", because they aren't. They are actually played by an emulator, which is provided with a file containing precalculated keypresses. How that file is generated, however, is another story. Why should there be any limitation on how that file is generated?
No, like I said we are doing tool assisted videos. We always assure everyone that we're not cheating, I think having a program doing for you is just as bad as using a game genie. You put in the code and you get a desired result. TASing is suppose to be fun, and done by a person who is playing the game....
Gamegenie codes modify the game mechanic in some way or other.
BisqBot does not. Anything that BisqBot does could be duplicated by a sequence of input that a human created.
No sequence of input, however, will be able to change the values contained in a game's Read Only Memory, nor will it be able to lock any RAM values in place, preventing them from changing when the game code writes a new value to them. Game genie and pro action replay do this. BisqBot does not.
How fleeting are all human passions compared with the massive continuity of ducks.
It kind of feels disapointing watching this after recently seeing the S3+K run, since it stole the glitched out movie crown. I understand why most of the zipping was happening ,but how come when MM walks normally into a certain walls he imeadiatley zips through?
Still a yes vote though.
Here are some shots Dehacked took. Can you guess the technical reason behind these?
