Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
From what I'm gathering, basically the RTA community's NSC category definition only forgoes save data corruption via resetting during the saving process, while the movie class guidelines describe both that and save data corruption via underflow as a single category "Corrupts Save Data." So this movie is a bit ambiguous in that it could fit both forgoes save data corruption and corrupts save data.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Yes, it sets the same flags as old save corruption TASes would as it uses the same hall of fame warp, it just doesn't use the act of hard reset save corruption as a means to obtain ACE.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
To clarify,
Upon checking Gifvex's NSC run is completely playable on continue after "THE END." this TAS can continue but crashes quickly, and the same for the Yellow NSC RTA route.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Thanks for your input Doomsday,
I could follow your logic in a vacuum, but as it sits now Gifvex's Pokemon Blue NSC TAS is the more recent and relevant comparison for HOF completion criteria and this TAS satisfies the same objectives. I believe that Blue NSC also results in an unplayable save file if the game is allowed to save after the credits, but I haven't tested this personally. An NSC branch following those criteria would also diverge from the criteria for RTA runs as current RTA runs of Yellow NSC result in unplayable save files. Personally since my interest in TASing grew out of the RTA scene I tend to work under their rulesets where applicable.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Sadly the game crashes upon attempting to continue. Historically the "no save corruption" TASes have only avoided the specific save corruption abuse of hard resetting during a save, so they frequently result in unplayable save files upon completion.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Hi everyone.
There's been some previous discussion that hinted at Red/Blue and Yellow no longer obsoleting each other when the TASes are sufficiently unique and entertaining. I'm submitting this Yellow NSC TAS with the intention of it being a unique branch from Gifvex's Blue NSC TAS. I think this is a good idea for Red/Blue and Yellow in general due to unique elements in glitchless (instant text in Red) and better reflecting the state of the RTA community, in addition to different trainers and the technical differences with the introduction of color in Pokemon Yellow which cause it to be inherently longer than Red/Blue.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I've just officially submitted a Yellow NSC TAS which will more directly bring up the problem Gifvex was inquiring about. I believe that the glitches used are sufficiently unique and entertaining to warrant an independent branch.
http://tasvideos.org/6119S.html
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
It's unlikely to work differently in GBHawk as it was also tuned for cycle accuracy in Pokemon Gen 1. Lior's SameBoy core is more accurate than either core at passing test ROMs in general and we might want to enable it for gb/gbc/gbc in gba in the future instead of just SGB. However at the moment SameBoy doesn't even have cycle accurate RNG for Pokemon, so some more test ROMs are needed to cover enough corner cases to make SameBoy the clear winner.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
It would need to be done with gbc mode and gba as gbc mode set to true to match the game boy player. But the current Gambatte core has some sync issues outside of the Pokemon Gen 1-2 so we're nit sure how well it will work even with that configuration.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I've cancelled this submission in light of Feos and Nach's reasonable discussion and finding a better way to highlight the verification effort that was made here, and also given the lack of discussion since Feos's post.
For the moment efforts are being refocused from Glitchless to a Pokemon Yellow NSC TAS preparing for the Pokemon Speedrunning Marathon August 10-12 where a prototype will debut, then another showing at GDQx in October!
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
We've run into a potential use case for the Sameboy core getting updated to .11 in Bizhawk. I've run through some test TASes of Mickey's Dangerous Chase on both GBHawk and Gambatte, and neither in Bizhawk 2.3 properly emulate the game's sprite lag. As soon as the first enemy shows up, the emulators lag more than on the real console. Sameboy may already be accurate enough for this game, and it would be interesting to see if TASing Sameboy .11 could allow for a proper console verification.
We believe that there already at some point was a Sameboy .8 core in Bizhawk, and the API has not changed much between the releases, so it may not be a complex update.
I may be able to assist with updating the emulator myself at some point but right now I'm tied up with Pokemon Yellow NSC TASing.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I also wanted to mention, I think it's important to look at the inverse of the problem presented here. If a new movie were to be submitted that is faster but relies on inaccurate emulation and does not play back on console, it would violate existing rules about taking advantage of emulation inaccuracy. I feel like in order to keep consistency with that rule, some standard needs to be set for accuracy related improvements as well.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Feos, thank you for your input here. It's helping me understand the problems faced with the current rules and proposed rule changes.
To the problem of perfect accuracy with the existence of the potential for non-deterministic behavior: Pokemon's RNG is as precise as Battletoads, it updates every frame and the update calculation relies on the rDIV register which is basically a CPU cycle counter. This is why such nuanced CPU behavior needed to be solved to sync the new movie. In effect, we had to find "perfect accuracy" as is relevant to Pokemon Yellow to achieve full sync.
In fact the gameboy does have some non-deterministic behavior observed in gekkio's test ROMs related to audio, but it wasn't relevant in this case. In many other cases with degrading hardware or inherently non-deterministic hardware design like the SNESes we ran into at SGDQ, this level of consistency is impossible. I don't think you have to throw out the general concept of "perfect sync" for a movie because specific other consoles or games have limitations including non-deterministic behavior that make it impossible.
Moving past that, I have another idea: what if after a submission is accepted for publication, the submission's authors become maintainers of the submission page and can update status messages for things like the level of console sync, status of sync with the latest BizHawk versions, and movie files to sync with newer versions? Effectively making the submission page a kind of wiki for the branch like Dolphin uses to track how games are running across versions. Then when an updated movie file on the page sees a meaningful improvement in optimality or entertainment, it could be considered for submission.
As mentioned elsewhere, my primary goal here was to document information related to the status of console verification of this TAS, which I think that solution could account for.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I didn't flesh out that last idea much. It probably doesn't work very well.
You mentioned not wanting to require console verification to accept something, does that cause a problem with a rule like "a TAS that has been independently console verified can be accepted as a resync improvement when the existing movie cannot be console verified"
I like the cleanliness of saying the last one didn't sync, now this syncs.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
This TAS is actually closer to 75 frames slower than the previous due to a variety of factors involved with rewriting 180k frames and the frames lost with the accuracy improvement. Your specific case would be slower because gen 1 misses get "Sandslash's Attack missed" as text whereas Kineses fail gets "But it failed," and finding a 1 in 4 enemy stat modification miss is likelier to find a fast setup than a gen 1 miss.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I'm not sure which of these would work well for drawing a hard line, but here are some ideas:
Allow resyncs when a movie is to replace another movie from the same author(s). This sounds quite reasonable in any event; what's to prevent an author from retracting their old runs before submitting the new one to circumvent the resync designation? Unless you can't retract a run from the site.
Allow resyncs when console verification is independently vetted, but not necessarily by a judge. Yes judges may not have access to verification hardware and indeed are unlikely to, but on the flipside there are often other individuals with their own equipment to replicate the acheivement. Indeed this would be a considerably more thorough vetting than any RTA world record could ever be vetted, even if not from an official judge. In this case Extrems has done an independent console encode, and at least one other individual in the TAS community has the equipment.
Allow resyncs when a known improvement is made to a core as opposed to resyncs across cores. This could tie resyncs to improvements in hardware knowledge better.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I believe Thunderaxe asked at some point in this thread about Brock Skip. Brock Skip is only possible in Red/Blue, not Yellow, and in RB it's slower than just beating Brock right away with Squirtle to get Nido experience for Horn Attack. So it will likely never be relevant for Glitchless pokemon TASing
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
Oh, I inquired about this problem in Discord and several people said the opposite. It is just a resync. A gameplay improvement is possible over the original movie but I don't plan on writing it for a few months.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
There are VBA GB runs? Yikes!
I definitely agree that timing should be done by cycle count where possible. From what I understand the definition of what a "frame" is breaks down during LCD disables, this is why console verification requires cycle count precision for inputs and incidentally could cause confusion about run length between emulators that implement LCD disable differently.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
HappyLee wrote:
None of where I slowed down looks like an oversight or mistake. Perhaps you're not paying attention to the details. The truth is, this run is as optimized as an SMB TAS can get. Did you also feel that, Scepheo?
TiKevin83 wrote:
I'll just make a note that I heartily agree with the idea that more people would be satisfied with the final entertainment focused submission if it were accompanied by an input file that has the optimized individual level frames. The final product of this TAS is awesome but without the other file it's much harder to improve on, even if such improvements are hard to impossible anyway.
So you're saying that we need to make two version of this movie just to make some people satisfied, or for it easier to improve on?
Certainly you shouldn't -need- to make two versions, but in my opinion the 2nd movie file as described would be easier to improve upon and therefore lead to broader appreciation of the final product by making it more accessible.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
I'll just make a note that I heartily agree with the idea that more people would be satisfied with the final entertainment focused submission if it were accompanied by an input file that has the optimized individual level frames. The final product of this TAS is awesome but without the other file it's much harder to improve on, even if such improvements are hard to impossible anyway.
Ambassador, Experienced Forum User, Moderator, Published Author, Site Developer, Player
(120)
Joined: 3/17/2018
Posts: 348
Location: Holland, MI
A revised version of this TAS has been console verified and will be submitted upon an official release of BizHawk. The following pastebin explains the verification process.
https://pastebin.com/DXiDT9ZT
