Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
If something happens at random and needs to be reacted to before it happens (e.g. by shooting a projectile), there's no way (barring perfect luck or rerecording) that you can react to it appropriately every time.
A TAS can get perfect luck and can rerecord, but this sort of AI can't.
Wow. Any fan of low%s should watch this. It isn't technically a low%, but the "2 pauses" restriction makes it play out very similarly to how low% looks in other games.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Actually, a good Chess playaround would be to manipulate the AI of a weak engine into following the moves of an actual famous game between grandmasters (while entering the moves of the other grandmaster).
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
JPC-RR does support subframe input (e.g. #3080: ais523's DOS NetHack "fastest death" in 00:01.15 which is another submission that's shorter than this one; and in the more serious NetHack TAS that dwangoAC and I are working on, we use subframe input to do luck manipulation faster). In order to avoid Warp's concerns, I played using JPC-RR's default clock speed.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Zowayix wrote:
- It looks like a major initiation point of the glitch is when the game figures out who to send out at the start of a battle. The submission text says that the game first checks the team, and if that fails, then checks the data of the Pokemon last sent out. But what's the point of the second check entirely? In normal gameplay the first check should always succeed.
I can't answer your other questions, as I'm not an expert on the game internals, but this one can be answered just based on normal programming mechanics. The second "check" isn't a check exactly, it's a loop fallthrough. What's almost certainly happening is that the code does something like this:
for each team member:
if that team member is not fainted:
set that team member as the currently sent out team member
break out of the loop
This would be a perfectly sensible and idiomatic way to code a check like this. You might notice that there's no code whatsoever dealing with the case where no Pokémon is found; because that situation is something that should never happen, then if you're writing for limited hardware (like the typical games console), you don't bother putting in any code to handle that case; it'd take up space on the cart and you'd hope it'd never have a purpose.
So what happens if all team members are fainted? The loop runs and doesn't do anything, so the "currently sent out team member" variable still has the same value it previously had, because nothing changed it. So you start with that team member sent out.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Personman wrote:
I more or less understand why this happened - we have some loose model of gameplay as by default only including interaction between the game and the player, and not the player and outside objects. For the bot, (some of) RAM is included in our perception of the bot's self, so it doesn't seem to violate the model, but for the human it does.
Sorry for the random tangent!
Right, the difference here is that the RAM is part of the bot's "brain", but not the human's. The parallel breaks down because the equivalence here is "computer memorizes information" vs. "human tells computer to memorize information", which makes it clearer why the situation isn't quite symmetrical.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I guess the way to look at it is: getting a cloud is a glitch that would, in the absence of ACE, crash the game. You're using a minimal amount of ACE in order to prevent the glitch crashing the game, which is a pretty minor use of ACE, but there's still ACE there.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Given that it's a handheld, this has to fall under the language preference rule. Which basically says that this movie can be obsoleted on [U] or on [E] (just like a movie on [U] could be obsoleted on [U] or on [E]), but the difference in time between the localizations (if any) doesn't count as part of the movie time for comparison. ([J] would be just as fine; I don't know whether there's a [J] version or not.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Yep, this is actually the normal way to ID scrolls (and to a lesser extent other items). The difference is that humans need to keep notes, bots can just memorize the prices.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I implemented stable rerecording in a game I was working on in order to make debugging easier, and to make it possible to recover a game if something went wrong.
I don't have a full-on suite of TAS tools for it yet though.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
As far as I understand it, the TASvideos rule is that you need to use clean SRAM unless you have a good reason not to. Accessing a category that would otherwise be unavailable is normally considered a sufficiently good reason. I think I'd prefer to see the no-SRAM version, though.
Also, I'd be very much against any use of the not-available-in-game cheat codes. (The ones that are available in the game, I wouldn't be against using them after they're discovered in-game, although that would be a different category. However, I suspect that's probably slower?)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
jlun2 wrote:
Regarding whether or not it can be done in real time as a criteria, there's things like Pokemon RNG abuse. It's used often to get perfect stats and/or shinies a lot faster than soft reset over and over. I think that situation still makes Pokemon Emerald "heavy luck manip", since the frame precision to get everything perfect is unfeasible.
As someone who does realtime manipulation of Pokémon games a lot (for fun, even), I'd say the point at which it crosses into heavy luck manipulation is the point at which you need a separate calculator program to work out precisely what sequence of apparently arbitrary actions will give you the result you're looking for. (The fact that it then takes me 2 to 3 hours trying to enter that sequence frame-perfectly simply by pressing buttons by hand is of course irrelevant when TASing.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I think it'd be funnier if people cross the finish line and then realise they haven't won. (Sadly, the way I think Mario Kart's lap counter works, it'd be obvious when you didn't go onto lap 2 after lap 1 had "finished".)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I don't personally think NetHack would be a good fit for a GDQ. However, if it somehow happens anyway, I'll do my best to make it entertaining.
And yes, I haven't been working on the run. I'm incredibly busy with non-TASVideos-related things at the moment.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
Huh. I started doing that casually a while back (still have the save file), but by the time I beat Blaine, I decided that I'd proved it was possible and then got stuck thinking up ways to entertain my audience.
My route was quite different, though (also horribly inefficient, especially outside TAS conditions).
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I like Masterjun's suggested screenshot too, as it'll instantly tell someone who knows the game what category is being run, and it looks pretty silly even in the abstract. (Not to mention that Viridian being in that position is basically unsalveagable for a human player, so it shows off TAS skill as well.)
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
The behaviour JMC47 reports is exactly what we'd expect with a CAV disc. The behaviour Dolphin's using is not CAV.
This opens up a path to fixing the loading times: change Dolphin to use CAV.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
This is actually good news; it looks like you've identified what may be responsible for the inaccuracies (and why they were inconsistent). Perhaps we can get accurate loading times after all.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
QuizmasterBos wrote:
TakaM wrote:
The gameplay wasn't the worst part of Other M, it's enjoyable to play imo and would make for an entertaining TAS.
Could you skip cutscenes, or was that only in newgame+?
As far as I can remember, you cannot skip cutscenes the first time playing. I don't even know if they can be skipped in a NG+ since I never played the game twice.
Although I've never played it even once, if I remember correctly, you can skip cutscenes in a second playthrough but they count against your in-game time even if you skip them. You should probably get someone who knows the game to confirm that, though.
Editor, Experienced Forum User, Published Author, Player
(44)
Joined: 7/11/2010
Posts: 1022
I know that back in the original discussion, I was advocating having, rather than tiers, tags: "this run is entertaining", "this run is a speed record" (with many games being classified into both), but the admins disagreed with me.
The plan of having a "speed records tier" and a "side goals tier" is similar, but different. (I assume that side-goals runs would be accepted only if entertaining, and speed records would be accepted if they were verified as a record, regardless of entertainment value.)
This would also fix the poll on each of the submission threads to be something that would be used as intended. It's uncontroversial whether a run is aiming for any%/100% or not (sometimes it's controversial whether it succeeds, but that's not important here). Thus, for side-goals runs, a "do you think this should be accepted?" question would be equivalent to the current "do you find this entertaining?", and for any%/100% runs, it doesn't matter whether the run is entertaining, and so the "do you think that this should be accepted?" question is as relevant as any other.
Thus, I'm going to throw my support behind the "speed record" versus "side goal" distinction; it achieves most of what I want, and also simplifies things noticeably for the judges and submission viewers.