I finished an attempt at the European version a while back, but I canceled the submission because the community wanted to see the 60-Hz original version. This made sense because most of the text in the game is skipped immediately, removing the benefit of using English.
The techniques I used still apply to the Japanese version, of course. In fact, I converted my video to the Japanese version, but was discouraged that my run was slower overall relative to a run published on Nicovideo. I was exhausted from creating my video and wasn't able to contact the creator of the faster TAS for more information, so I sort of just sat on my speedrun and gave up.
It's worth noting, though, that while the Nicovideo run is faster than mine overall, it's quite a bit slower in certain sections of the game. This gives me hope that elements of my speedrun might actually be accepted on this site... given somebody who's sufficiently dedicated to finding the remaining optimizations.
It's been a while since I dusted off my TAS folder. I'll dig into it, find the latest version of my TAS on the Japanese game, and share it here within a week. It's my hope that having an existing run will be a huge help to those working on the game.
Also, here is an older topic on the same game. There may be some helpful resources in there.
Good luck!
You should ask the TASer to upload it somewhere for you, since it's just an input file; an avi wouldn't help at all how to pull of a frame-specific trick.
FractalFusion wrote:
Nope, video uploads are not input files. If the uploader didn't put a link to the input file in the description, then there's absolutely no hope to get it so don't even bother trying.
If for some reason you are really lucky, you can find a blog or some other contact outlet that the uploader uses, contact him there, and hope he complies with your request. Nicovideo is not a contact outlet.
Alas, this is what I feared. Thank you for your replies.
This presents sort of an awkward situation. I have a completely independent run of a game and the run happens to be a few hundred frames off the record. The thing is, I actually do some levels significantly faster than the recordholder (on the order of thousands of frames)--it's just that he gradually makes up the time with frame-precise optimizations in nearly every other level (ten frames here, ten frames there). While I'm confident I can get my movie to beat the existing record overall by making a few obvious tweaks, I know for sure that I will have a slight positive split on the majority of the levels when compared to his.
The obvious option in most cases would be to closely inspect what the recordholder did in the input file to make up the time. The issue is that I can't actually access this file (as discussed in this thread), and trying to reverse-engineer every frame of a flash-embedded movie seems a ridiculous proposition.
So, the existing record is 37:22. I know I can get my movie below 37:20 or so just by making obvious improvements. To be sure, this breaks the existing record. What's frustrating is I know that the recordholder has about 20 seconds' worth of indiscernible optimizations sprinkled throughout his run, meaning that a theoretical movie which combines his best levels with my best levels would push the final time all the way under 37:00.
This raises a question: the criteria for a speedrun's acceptance to this website is that it "breaks all existing records." It is relatively trivial for me to break the overall record for this run, but I know that many of my individual levels are actually run slower. Would the judges still accept my run knowing that I was slightly suboptimal in many levels, provided I achieve a better time overall? If not, then I'm stuck, because there's no way for me to dissect the existing record. Surely I could spend ages trying to replicate every last pixel-perfect stunt the leader did by watching and re-watching a streaming video ad nauseam, but I frankly wouldn't have the patience to do this and my months of effort on this run would be wasted.
Any thoughts?
Apologies if this has already been answered--I looked around the forum a bit but couldn't find anything.
Is it possible to download the actual emulator movie file from Nicovideo? For example, I'm curious how the author of this movie pulled off a particular stunt, but I don't know if it's possible to actually download and look at the SMV file.
Any help is appreciated. Thanks.
After some (pretty exhausting, tedious) investigation, it seems that waiting on the title screen until address 7E005E is 26 (decimal) grants good outcomes for many of the "randomized" levels:
Catacomb logic puzzle: The random variable is changed to 205. The guy who knows Nick is in the far right cave, meaning that a quick sweep from left to right is all that is necessary, forgoing the annoying upper-left room entirely. Value goes from 205 to 207 upon getting a "true" response from the correct guy.
Anconda fight: I didn't realize this fight was randomized until just recently, but it is. Thankfully, both attacks are drops from the left, which is very time-efficient. Value goes from 207 to 209 for the first attack, then 209 to 211 for the second attack.
Pirate ship "vase" room: Key is in the first jar. This saves lots of time, especially compared to the case where the key appears in the fifth vase, like that in the movie I canceled. Value increases from 211 to 212 upon breaking this vase.
"Karma" level: Women are found in odd-numbered boxes relative to the order you open them; i.e., the first box opened is a woman, the second is a monster, the third is a woman. It may be possible to manipulate this so that the first two boxes opened have women inside, but this would be an extremely tedious process considering the amount of randomization used earlier in the game that affects this outcome... I don't even want to think about it. Anyway, value increases by 1 for each box hit, and increases by 1 again to 216 at the end of the level.
The Tower 3: I didn't know this level was randomized until just now. It's another pretty significant detail in the game that goes overlooked. Anyway, each mini-stage seems relatively flat, which is good for Whirlo's dash. I haven't actually TASed this level yet because I'm saving it for when I convert everything from the European to the Japanese version of the game.
With these changes, the movie goes considerably faster. Even though I still have plenty of work to do, I'm expecting the Japanese version of the run to take about 37:05--well below the 37:22 run posted on Nicovideo which apparently does not manipulate luck. I'll get back to work.
Jigwally wrote:
Nicklegends, after reading the differences you listed between the Japanese and European ROMs, the Japanese ROM seems better suited for a run. It's also the best version to use because the European version uses PAL timing and there is no US version.
Jigwally wrote:
I second (J) version. The language difference doesn't matter much since most of the English text gets skipped too quickly to read anyway.
I wouldn't have minded starting out with either version; I was just trying to follow the movie rules where the only applicable line is "English text is preferred. Thus the use of Japanese text in the movie would be a negative aspect towards a switch to the J version." The page doesn't say anything about the use of a European version, nor does it mention anything about timing. I wish the page were a bit clearer when it came to situations like this.
So, I looked into the game's randomization a bit more, and it seems that the seed is determined by how long one waits on the title screen before starting a new game. This means that there should be an optimal number of frames to wait so that the outcomes in the Catacomb logic puzzle, the "vase" pirate ship level, and the "karma" Tower level are all the most convenient possibilities.
Specifically, check out memory location 7E005E (apparently 1 byte). When booting up the game, it increases by 1 per frame until one selects "New Game." Right at the start of the Catacomb logic puzzle, it gets permuted a few times and becomes some other pseudorandom number. It gets messed with a bit during the Anconda fight but stays the same until the vase level, where each broken vase increments the counter by 1.
I'm looking into this more as I type and should have some better results soon.
Ladies and gentlemen, I've decided to cancel this submission because the Nicovideo movie alerted me to a few spots in my run that can use better optimization. Specifically, I determined that it is very possible to skip by the two boulders in Toadstool Cave 1, and I expect that other maneuvers executed in the Japanese ROM are also possible in the European version. I also need to do more research about the random variables in this game that affect the vases in Pirate Ship 4 and the boxes in The Tower 2.
As is, the Nicovideo version beats my movie by the equivalent of about 400 frames. But now that some improvements have been pointed out, I should have absolutely no problem overcoming the difference and then some. By redoing Toadstool Cave 1, I save 223 frames right off the bat.
Thank you for all your comments. I'll be back with a better run soon.
Adjusted for NTSC vs PAL timing differences, but not version text differences:
that rejected run would be equivalent to a 51m 59s run on this ROM version...
The Nicovideo run would be equivalent to a 44m 50s run .
And this run would be equivalent to a 37m 29s run on an NTSC rom.
The two versions differ by more than just their text--there are a number of level differences (the Japanese version has a few more enemies, for example). With this said, I will compare my run with the Nicovideo run and see if any improvements can be made. At first glance, I noticed that the player in the Nicovideo run is able to
progress directly to the platform after bouncing off a particular enemy in Mt. Alsandra 4 (I talk about backing up to get a better angle in my own video).
skip by not one but both problematic boulders in the Toadstool Cave levels. I tried to make this happen for ages in my run without any luck.
uncover the key in a vase in Pirate Ship 4 that wasn't the last one to be broken.
reveal the women in The Tower 2 in a different order.
seemingly bounce higher when performing downward stabs in general.
Some of these suggest that I may have missed opportunities to manipulate luck; others suggest that there may be subtle changes in the (wonky) physics engine between versions. I will keep my run submitted until I can say definitively that the time savings in the Nicovideo movie are not limited to the Japanese version of the ROM.
Is this yours? The user name is different and you said that you would use the Japanese version of the game, so I guess that submission is from a different person.
OH COME ONNNNNNNNN
I'm sorry, I didn't know you were also working on this. Hopefully we can learn some tricks from each other.
Interesting that very shortly after the creation of a thread about this game, a complete and clearly-carefully-researched TAS is submitted. Had you just happened to be working on this, or were you holding it in secret, waiting for other people to discover this game? :)
Holy cow! I had absolutely no idea that thread existed. When I started working on my run a few months ago, I couldn't find any discussion of it on these forums. That was actually one of the reasons I went through with it--I liked the appeal of sailing into (relatively) uncharted waters. The fact that a topic was started mere hours before I submitted this run is a complete, crazy coincidence. But I'm glad that people are talking about the game, at least. It doesn't get enough love.
The movie rules page makes it clear that movies are to be recorded from power-on. How is this accomplished in Snes9x 1.51? I see the option to record from reset, but I'm not sure if this is considered the same as turning off and on the power. If it is the same, the movie rules page is misleading; if it is different, please let me know how to record from a power-on.
Thank you in advance.