Posts for PoochyEXE


PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
OtakuTAS wrote:
But I'm guessing if you delayed a frame now and then you may get a better result.
As far as I could tell, the RNG only picks up entropy in menus. I delayed pressing A to pick Chikorita by a couple frames to get a starting board that was more conducive to getting 2 simultaneous chains going quickly, but after that, the panels that spawn seemed to remain the same no matter what I did differently.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Yup. In general the pauses are to manipulate either the column the next bomb spawns in or the position and/or color of the next diamond that spawns, because the RNG for both of those are influenced by the lower 4 bits of a frame counter.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Yeah, the SNES version doesn't have a Mode B in Round Game, and the Round Game cutscenes just show Toad making his way through the overworld map towards Wario's castle. VS COM became the mode with actual cutscenes with dialogue, which are also different cutscenes from the NES version. I picked the SNES version since I still think it has overall better entertainment value, mainly due to better QoL with things like being able to pick up diamonds, turn around and kick objects on the same frame using L/R, etc. which allow for a wider variety of setups and chains. I also like the SNES version of Birdo's theme way better -- I'm pretty sure hearing the NES version 99 times would've gotten old very fast.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I haven't actually paid much attention to player points in all my time here, but I was curious as to why that number on my profile was changing when I hadn't submitted or posted anything recently, so I searched the forums and saw there had been quite a bit of old discussion on it. I have no idea if anyone still cares about player points, but I have a simple suggestion in case: Update the computation algorithm so that for each non-obsoleted movie, its Player Point value is the greater of (A) its value under the current calculation system OR (B) the highest value any of the movies in its obsoletion chain would be worth if they weren't obsolete. The idea being that if movie X should only obsolete movie Y if X is strictly better than Y, so it should also be worth more points. And this would remove the perverse incentive to avoid obsoleting your own movie lest your points drop due to fewer people voting on the new movie.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Technically there aren't 255 rounds. There's a fixed rotation of 32 extra rounds that start at 100 and loop until 255, i.e. rounds 100 through 131 are identical to rounds 132 through 163, rounds 164 through 191, and so on. Then round 255 loops by itself, and the only indication that R255 is the end is that the starting layout is exactly the same and the round number doesn't increment. So one could make a 255-round video but the latter half or so would be pretty boring to watch, and there isn't really a sense of completion at the end. Or one could also make a video up to the point where the extra rounds loop, but again there wouldn't really be a sense of completion at the end. Edit: On the other hand, rounds 101 through 105 have patterns that spell out W-A-R-I-O with the monsters across the 5 levels, which would be cool to incorporate into a TAS somehow. VS COM would be a good third branch though.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Cool to see my work finally complete and published! Thanks for the prompt judging and processing! I noticed the publication page says it obsoletes the previous run, even though this was meant to be a new branch (and Samsara accepted it that way). Was the decision changed, or is this an error? Also, if you'll excuse the nitpick, the description mentions stuff from the NES version that aren't in Round Game in the SNES version, namely minibosses and cutscenes of Wario trying to discourage Toad. The SNES version doesn't have minibosses, and the cutscenes are different -- in Round Game they just show Toad moving through the overworld map, and the cutscenes with actual dialogue were moved to the SNES-exclusive VS COM mode, with different dialogue.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Nope, I'm finally done. I had found a couple other optimizations, but a few minutes ago I just finished testing out the last of them, and they actually end up losing frames, since I ended up having to spend more time on luck manipulation on subsequent rounds.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I'd like to delay this again since I re-watched the encode and spotted a couple more optimizations. I should have another new version ready in a day or two. Sorry if this is getting annoying, I keep thinking I have everything optimized only to come back several days later and notice another little imperfection. Edit: Done! Saved another 320 frames. https://tasvideos.org/UserFiles/Info/637811121948338188 None of the changes are quite as interesting as the R59 improvement from last time (R59 is one of my favorites too), just a bunch of subtle smoother movements and setups here and there.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Sorry, please hold off on judging for a little bit, I spotted a couple optimizations that I think I can add in while still being able to resync the RNG state afterwards. I'll probably have a new version in a day or two. Edit: Done! https://tasvideos.org/UserFiles/Info/637806096666496015 -- Please replace the movie file with this new one. I've included a changelog in the upload description. (Note that this still includes the fanfare at the end.)
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Thanks for the temp encode! I've uploaded a modified version with the special fanfare: https://tasvideos.org/UserFiles/Info/637798894512418864 (It literally only inserts 291 blank frames before the final Right input to proceed to the ending cutscene.) I do agree that it makes for more entertaining viewing, so judges, please go ahead and replace the movie file when you get a chance.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I noticed the TASVideosChannel on YouTube seems to be uploading higher quality re-encodes of older publications recently. Just out of curiosity, what specifically has changed in the encoding methods? I can't quite put my finger on it since I can't find the old encodes for a side-by-side comparison. But kudos for the dedication to high-quality encodes!
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Wow. Just wow. That was absolutely brilliant! This belongs in Stars for sure, and I smell a TAS of the Year award or two as well. A yes vote isn't enough to do this justice, I need an option for my vote to be the entire "I'll have what she's having" scene from When Harry Met Sally.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Definite yes vote! Also, just out of curiosity, what was the reason for intentionally hitting the walls on Route 101? It looks like there's some oddity with the physics engine being exploited here, but I can't quite tell what it is.
Post subject: "Morimoto About SMB Movie" page translation corrections
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I just stumbled across this page by chance while browsing the wiki, and while I realize it's an old page that probably gets viewed once in a blue moon, I feel I can improve significantly on the translation and clear things up in a couple spots. Unfortunately the page doesn't seem to be editable, so I'll post it here in case any staff member wants to update the page:
2ちゃんのスレにはマリオと一緒に、緋蜂ノーミスムビと、 アリカのテトリスデスモードクリアのアドが貼ってありました。
An advertisement of Tetris Death Mode Clear from Arika was pasted to the 2chan thread about Mario together with Hihati's no-miss movie.
I'd translate this as: "In the 2ch thread, together with Mario, a Hibachi no-miss movie and Arika's Tetris Death Mode Clear advertisement were posted." 緋蜂 is Hibachi, the true final boss of DoDonPachi DaiOuJou, a famously difficult shmup. Even getting to Hibachi is quite the accomplishment in itself, much less a no-miss against it. "No-miss" is actually Japanese shmup lingo for "never getting hit", I think it comes from a corruption of an abbreviation for "no-mistake" but I'm not sure. And Tetris Death Mode is referring to T.A. Death Mode in Arika's Tetris: The Grand Master 2 PLUS. At the time, it was widely considered to be by a country mile the hardest Tetris game ever commercially produced (although it would later be outdone by Tetris: The Grand Master 3 in 2005, which holds that distinction to this day).
オレはデスモードクリアできるんですわ。 っていうか結構初期のほうのマスターGmとかなんですね。
I can clear death mode. Saying like this, it's the master Gm of the way of the splendid beginning and such. (?)
I'm not entirely sure if I've gotten this right, but my interpretation is: "It's not like I said I can clear Death Mode, or that I was able to get Master Gm very quickly or something." Note that the "I" in "I can clear Death Mode" is "オレ", which (a) the author doesn't use anywhere else to refer to himself, and (b) generally comes off as overflowing with self-confidence. "Master Gm" refers to the Grandmaster rank in TGM2's Master Mode, also an incredibly tough feat to accomplish, probably even harder than clearing Death Mode.
んで彼は大往生の2周目もクリアしてる凄腕ゲーマーで、 実は緋蜂のノーミスは彼の依頼を受けて製作したのです。
... He is also second clearing resourceful gamer in Peaceful Death, and the no-miss of Hihati received and, to tell the truth, produced his request.
I'd translate this as: "And also the resourceful gamer who can clear the 2nd Loop of DaiOuJou, he actually recorded the Hibachi no-miss in response to a request." 大往生 is "DaiOuJou", short for DoDonPachi DaiOuJou. The subtitle literally translates to "peaceful death". 2周目 is official term for the 2nd Loop of DaiOuJou -- normally the game ends after the regular final boss, but if you've done extremely well up to that point, you go into the 2nd Loop, carrying over your remaining lives, score, etc. and starting over from the first stage, but with enemies' firepower drastically increased. Make it to the end of the 2nd Loop and you fight Hibachi after the final boss.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Cancelled in favor of the Expert run.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
OK, I've cancelled the Easy submission in favor of this one.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
In fact, a 20 combo was actually the very first thing I tried to do when attempting an Expert run. This was the closest I managed to get after many attempts. The problem with executing a 20 combo is that it requires the sides of the playfield to be filled almost exactly to capacity. More precisely, I can leave up to 3 cells empty total across both sides, and those cells must be in the upper corners and connected to the main well. And leaving more than 1 empty cell would significantly limit the choices I have for the last couple hits in the combo. There's absolutely no wiggle room whatsoever. (Case in point: In this run, I had to dump a Z piece in the top left corner before starting my combo. With a 20-combo setup, there wouldn't have been any room in the corner for that piece.) There's one tiny exception if I can manipulate my way into getting 1-2 white stick pentominoes (and no other attacks just before them) just as I'm about to start my combo, in which case I can leave 1-2 additional cells empty in columns 3 and/or 7 as long as they're not in the same row nor in the bottom 2 rows. I still suspect it's just barely within the realm of possibility, but I'm pretty stumped as to how to luck-manipulate my way into it.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Based on the feedback, I've submitted the Expert movie as well. I'm aiming to get one or the other to get accepted but not both; if and when one gets accepted, I plan to cancel the other. I'll leave it to the judges to decide which one is better. Link is now in the submission text.
Antronach wrote:
Well, you were able to keep it entertaining and make it a time improvement over the current submission. I think the expert run is preferable.
I think you misread my previous comment. This submission is faster than the Expert run. Still, people pointed out to me that runs on harder difficulties can obsolete faster runs on easier difficulties (and have before), so I figured the Expert run is worth submitting too.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Considering the playfield is exactly 20 blocks tall, I think a 20-hit combo would be a very tall order (no pun intended). It's theoretically possible, but it'd require an insane amount of luck manipulation. I have managed to accomplish a 17-hit combo in 2 other attempts at further optimizing my previous movie, but both left behind terrible board states which made both the subsequent All Clears and KO'ing the AI take significantly longer. Quite a pity too, since a 17-combo is worth a x3000 multiplier, and 3 All Clears plus a good time bonus would be worth 335K base bonus, enough to hit a billion. Edit: I went back and tried to optimize one of my previous failed runs, and it turns out I was able to pull off a run with a 17 combo and 3 All Clears on Expert. It's still slower than this run by 123 frames and 5 seconds of in-game time, but it makes for a decent proof of concept. I've uploaded it as a WIP in case anyone's interested. Thanks for the inspiration, Qlex!
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Hm, I had previously only been thinking of trying to get 5 2x2's to get a second All Clear immediately after a first one, although that idea was also sunk since the first All Clear would counter some stuff. I hadn't thought of the possibility of getting more than 2 5x5's, but now that you mention it, it just might be possible in Expert mode. And it'd be awesome if it were actually executed -- I can imagine it already, "All Clear! Pentris + All Clear! Pentris + All Clear! Pentris + All Clear!" There is one additional difficulty in getting 5x5's though: As I mentioned in the comments, I don't have much choice but to go for the big combo right away. This sends a bunch of pentominoes to the AI, which, if they're volleyed back and forth, would result in the 5x5's going to the AI and not me. Meanwhile, waiting for the AI to send pentominoes to me isn't much of an option either, since there's a delay between clearing the line and countering, plus you can't counter the next piece to spawn, and pentominoes would make the combo significantly harder. It might be possible to combo, let that first set of garbage get countered, use up the 2x2's, build up a counterattack while waiting for pentominoes, then volley to get the 5x5's. I might give that a shot when I get some more free time.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Yep, this is intended for Moons, and possibly Notable Improvements as well. I intended for this run to be both significantly faster and significantly more entertaining than the previous run; in this run I use a wider variety of strategies to get All Clears and use a new technique to synchro hard drop + twist and save frames (which requires frame-perfect precision). As for higher difficulties, the strategy there would probably be to time my attacks to either counter or turn the AI's attack into a counter, since countered pieces are just squares and not crazy blocks. It'd be a longer, slower-paced run, and probably less entertaining, but I think a one-stage run on Expert is possible.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
Thank you very much for the encode, Spikestuff! And thanks everyone for all your comments!
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I can't get it to sync on 0.9.10, which is a pity since I'd really like to see how you optimized it so well. In any case, I'll take your word for it since you're a judge, and cancel this submission since it's already been obsoleted.
Post subject: SRAM saves don't work, causing movie desync
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I'm in the middle of making a TAS on Desmume 0.9.10 (x64), and I've been periodically playing back the whole thing on fast-forward to make sure it hasn't desynced. Today, I noticed that it's suddenly desyncing at the first time the game auto-saves in the movie, and the save clearly didn't update. The movie proceeds to desync since the failed write to SRAM appears to introduce additional lag frames. This didn't happen before; earlier today I played it back and it would get through that section just fine. Additionally, if I load up the ROM and play manually, the game saves just fine. But if I load the movie, saving doesn't work, even if I stop the movie before the first auto-save and play manually. I've tried deleting the SRAM file, but the problem still persists. Is there some way to fix this? Did I accidentally flip some setting in Desmume? Edit: I just opened up the movie file in Notepad++ and inserted an extra frame at each auto-save and now the movie syncs again. Saves still don't work.
PoochyEXE
He/Him
Experienced Forum User, Published Author, Player (223)
Joined: 12/5/2010
Posts: 29
I imagine the optimal path for eliminating the AI is to make large combos over and over, pausing occasionally to counter whatever the AI sends until it turns into 5x5s and then using those to clear a Pentris. So somewhat different, but not completely different. I think it could be moderately interesting to watch.