Posts for PearlASE


Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
ior's MarisaA LNN has a replay here: https://maribelhearn.com/replays/lnn/ior/th11_udirMA.rpy Scoring 750 Million points higher than this movie with the same restrictions. (2,843,113,180)
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
This is a very, very unoptimized run for a category like this, I'd go through every unoptimal segment in this movie, but the post would be about 3000 words, so here's a snippet from stage 1 and 2. Since this is a TAS, every nonspell and spell card should be timed down fully to 0 for extra graze, as the spell card bonuses do not matter when compared to how graze works. Graze in Touhou 11 affects the Point Item Value Multiplier, for every 100g, you get an extra 0.01x point item value. A rough estimate of the item count for a NoMiss NoBomb play of SA is going to be about 2,150. It could be a bit higher than this depending on 4midboss kill technique, but let's just assume that there are 2,150 items in the game, shall we? Stage 1 : Already we're off to a very unoptimized play; the WR by ior gets 2,960 Graze after Kisume (stage 1 midboss), while scratchtas gets only 652. With some rough calculations, we can estimate that the score loss from this is a massive 80 million points, already. (2110 * 200,000 = 422m | 2110 * 200,000 * 1.2[graze multiplier change] = 506,400,000) Entering Yamame, Ior has 4,070 Graze, and scratch has 1,328. A difference of 2,700. Mind you, while the two categories are different (One is allowed deaths and bombs, the other is not) The two runs routes are identical to this point, so this lost graze is purely the result of very poor optimization. On Yamame, Scratch also just doesnt time down anything. Opting for spell bonuses, which is so completely not worth it it's not even funny. Spell 1's bonus is 3mil, while the score you would get from timing this spell down fully is approximately this : 2,020 * 200,000 * 1 = 404,000,000 | 2,020 * 200,000 * 1.06 = 428,240,000, so the score LOST from going for spell bonus is 25 Million points. Non2 is also killed asap, you can just estimate the score loss if you want. But once Yamame final begins is the final time we can compare with ior. Ior will bomb the 2nd wave of the final spell, and their graze difference is 4,800. Stage 1 alone shows a lack of care for maximum scoring, but let's just continue through the rest of the game, comparisons to Ior from this point on will no longer be around, as the route will differ. But I'll do my best to estimate a score loss at the end (if I bother to look at everything) Stage 2 : The grazing on the yinyangs before the midboss are very poor, losing chunks of points as you usually will want to let them fire everything they can before killing them in a TAS. Let's be generous and say 350 graze is lost here. (350) The yellow/green yinyangs are also killed instead of stalled for graze. (850) Parsee non1 is instakilled (1,850) Same for spell 1 (2,850) And non2 (3,500) Spell 2 doesnt even do the scoring strat. (5,000+4,800) This feels more like a speed run than a "scoring run" due to how much score is just lost from a lack of actual scoring. The rest of the run also features millions of points lost at every attack in the game, I can't really be bothered to look at the rest of it from just skimming stage 3. Notable exception is Stage 5 boss, Orin's 2nd nonspell where he misses out on like 50 point items aka 40 million points at the very unoptimized graze count at this point in the game (and when we add the 10,000 graze lost in stage 1 and 2, this score loss jumps to 50 million, let alone the rest from stage 3,4 and 5.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
Signing up as individual but keep in mind my brain is not wired for time attack TASing as of yet.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
moozooh wrote:
Lol, hi Pearl. :D Sick run, loved it! I thought for a bit for some potential avenues for improvement, but it doesn't look like any of them compare to simply grazing more (or, to put it more precisely, minimizing the time spent not grazing). So if any improvement comes along, it'd probably come up with more efficient grazing routes that minimize idle time between bullets. Also, am I to assume you'll be also doing RF2 and/or Jet at some point? I'd love to see either, but particularly Jet, of course.
Thank you for the feedback moozooh! Glad to see you here. I do think this is possible to optimize a bit, but the graze routing is pretty tight when combined with the stage layouts for all of the secrets; so any improvement i think should be minor. I plan to do Jet eventually as it retains the same graze chaining mechanic at 1 (2 has 2,220 per frame flat with fairy, BORING)
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
Patashu wrote:
Finally got around to watching this; the graze chaining is hilarious, yes vote. There's some times where you're doing nothing and there's a bullet on screen you could have grazed, or times when there's plenty of bullets to graze and maybe you could have woven in a kill or two you didn't do; I don't know if these are 'there's a reason why this is optimal' moments or 'the points gained from this fall below the threshold of optimization effort I'm applying this time through' moments. In either case I don't mind, since the overall strategy is justified and held up. (And you do indeed say in the description you haven't done all possible micro-optimizations.)
There are a few instances of this, especially if the bullets are fast; the points that would be gained is under 1,000 and realistically would only be about 200-300 per instance. Any planes that I didn't kill that I *could* have killed are only worth 1k In total, im pretty sure all those little instances are worth 30k
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
WarHippy wrote:
I really like these kinds of runs, but I’m curious why you let so many enemies escape
In almost all cases, it's not worth killing them. Most enemies have a value of about 1,000 or so points, which is one frame of grazing. Shooting at an enemy could end up killing an enemy that's about to fire at me, resulting in a loss of points.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
excite wrote:
I'm trying to tasing the PC engine port of klax. If you don't mind, could you tell me the techniques and glitches you used in this movie? I thought maybe I could use a PC engine port. (The text was translated by Google Translate, so it may be difficult to understand or unnatural.)
For the most part, I play the game as Vanilla as possible, as the goal is to score as high as possible; I focus more on specific tile patterns for higher scores. The only glitch I use is at Wave 100, to place tiles out of bounds, which has no purpose in a score or speed oriented TAS. "H Pattern" - Good for diagonals. Score : 640,000 X = Tiles that are useless O = Initial klax A = Primary scoring tiles. B = Other scoring W = Wild Tile specific O A A A A _ A X B B A A B B O A A A A O B B A X B B "Basic" - All around pattern used. Score : 810,000 O A A _ A A A A A A A A O O O O A A O A A A A O A A "???" - Something that happened on wave 70. Score : 855,000 O W A _ A A W W W W A A O O O O A A X A A W A O W A
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
PearlASE wrote:
If I were to play for full speed, I wouldnt be able to manipulate the pieces I like, especially late game. The pieces are impossible to catch at max rank at high speeds, in which you see me not speeding up at all unless I use it to intentionally finish a tile wave.
Are there instances where the pieces could be sped up without compromising manipulation and ability to catch? Or places where manipulation is yielding a desired result and the rest could be sped up?
PearlASE wrote:
Also, I wish it gave notifications on the website for replies on this threads
It sends an email if you "ring the bell" on a thread.
Basically, the early bits of each wave have the possibility to be sped up more than shown here (but i often checked speedup for rng seeding, used it when needed) but after the first 35 seconds or so of each wave, it becomes unmanageable. The first few stages could have been sped up.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
Sometimes you make the pieces slide down quickly, and sometimes you wait for them to arrive at normal speed. Why wait? Also, how exactly do you manipulate luck?
The piece speed differences are used in luck manipulation. Luck is manipulated by player 1's inputs (and inserting a coin) this shuffles the rng seed that causes the tiles to spawn as certain colors. Youll see me wiggle around or stand still to force the colors I aim for. If I were to play for full speed, I wouldnt be able to manipulate the pieces I like, especially late game. The pieces are impossible to catch at max rank at high speeds, in which you see me not speeding up at all unless I use it to intentionally finish a tile wave. Also, I wish it gave notifications on the website for replies on this threads. If you have any other questions, Feos, just let me know
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
Redownloaded the game, the ltm file and mame 220 on my new VM. The playback still syncs (atleast until frame 250,000 when i stopped) with everything set to default on mame (the .mame folder being removed) If yours is desyncing early on, then it's an NVRAM issue.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
Did you configure mame according to https://tasvideos.org/EmulatorResources/MAME? I can't get this run to sync at all (mame 0.220, libTAS 1.4.4, klax ROM from 0.220 set).
On my older VM, I ensured that the configuration matched the link above, let me try to re-download the movie file and get it working.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
PearlASE wrote:
Time to enter. The 999 timer only applies to the first boot
Okay you won't have to provide nvram, you'll only need to provide a movie that generates it, as a verification movie. We already do this for DOS games you have to install for several minutes. And we'll attach the verification movie to the main publication. Just don't skip any actual game.
Sounds good, cant wait to make a movie file that sits on the loading screen, itll be pretty funny. So now all that's left is the autofire thing, how difficult is it to remake it in lua with a separate keybind as I need to also use no-auto at times too for specific setups
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
PearlASE wrote:
Raiden Fighters, on it's Japanese romset (and many romsets) require you to wait through a 999 tick timer on the first boot, a pre-packaged nvram file would skip this timer.
Is any kind of pre-gameplay required? Or is it just the time it takes to enter gameplay?
Time to enter. The 999 timer only applies to the first boot
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
How long does it take in real time?
It takes about 10-15 minutes in real time. Also thank you for the misc keys thing, I appreciate that. If you can get the lua thing working with 1/1 then hit me up on the Discord cause my mind is very fresh for this game, would love to make another sample stage 1
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
Should be.
This would be amazing.
feos wrote:
We just need a way to recreate it on our end. For example through a separate movie. What is nvram needed for?
Raiden Fighters, on it's Japanese romset (and many romsets) require you to wait through a 999 tick timer on the first boot, a pre-packaged nvram file would skip this timer.
feos wrote:
You can reassign libTAS keys however you want.
Not the libTAS binds, but the keys used in MAME. Using control for fire and alt for bomb is very awkward, especially alt for bomb causes issues with window priority.
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
feos wrote:
PearlASE wrote:
Hi, I want to know the feasibility of being able to submit an input file (movie) that utilizes the autofire plugin of MAME for libTAS. The consequence of using it is the file would not playback on full default settings, and 2 or 3 small configuration files must be downloaded before being able to sync. However, getting these files in their right place is as easy as downloading them and placing them in their correct directory. It'd work similarly to an NVRAM file. The reason I ask is because manually tapping the button even frame by frame is frankly horribly inefficient to me, its not the accuracy that matters in shmups but rather the movement and positioning of my shots, so with what I've mentioned above, would this be permitted if the process is as easy as I've made it out to be?
I'm thinking of trying to reproduce the autofire functionality in libtas through its lua. Which patterns do you need?
1 on 1 off ideally, Is it possible to make it user configurable? Also- on the topic of controls and NVRAM; is packing an NVRAM file with a movie file allowed as Raiden Fighters may require it. Also can I somehow use a non-default control binding for mame?
Post subject: libTAS MAME Autofire
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
Hi, I want to know the feasibility of being able to submit an input file (movie) that utilizes the autofire plugin of MAME for libTAS. The consequence of using it is the file would not playback on full default settings, and 2 or 3 small configuration files must be downloaded before being able to sync. However, getting these files in their right place is as easy as downloading them and placing them in their correct directory. It'd work similarly to an NVRAM file. The reason I ask is because manually tapping the button even frame by frame is frankly horribly inefficient to me, its not the accuracy that matters in shmups but rather the movement and positioning of my shots, so with what I've mentioned above, would this be permitted if the process is as easy as I've made it out to be?
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
Thank you!
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
pachunka wrote:
I'm a big fan of this game (I grew up on its C64 port), so I'm biased, but this was amazing to watch. The end of wave #78 with the huge pile of drops would be a great frame. What on earth was going on in wave 100 with the out-of-bounds drop spaces? Is that something that can only happen on that wave, or was it something fun you were saving until the end? Knowing how brutal the tiles can be, the luck manipulation seems impressive. Sidenote: Seeing that drops and vertical-klaxes were necessary, even in a TAS, made me feel better about my struggles with this game as a human-player.
You can place tiles out of bounds as seen on 100 on any wave, but they are permanently onscreen until the run is over, hence why I opted to make that build on wave 100. Also, because wave 100 is permanently on minimum rank which gave me time to build the scoring pattern as well as the funny. The RNG manipulation was pretty extreme at times, I only have 6 inputs to use to manip (directions, button 1, coin) and sometimes there just is not enough frames to get the color I'd want!
Twisted Eye wrote:
i dunno that final score screen kinda suspicious
:)
Experienced Forum User, Published Author, Player (204)
Joined: 12/5/2022
Posts: 20
Location: Canada
FractalFusion wrote:
Nice 2048. I wonder what the fastest completion (fastest time) TAS looks like, being that the gameplay is pretty unusual for a Touhou game.
A time attack TAS for this game would boil down to a rather boring game loop to be honest, use bomb to get score which gives lives and then suicide for more bombs to finish the stages faster It's also a lot of flashing which I tried to avoid. Glad you got it to sync, fsvgm. I wasn't sure if the version I had was the latest from the master but the executables name mentioned Rev 22.