Posts for berrimeow


berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
Wow, there's so much more optimization that can be made with this game than what I tried checking. I wasn't sure if anyone knew anything more about Dual Shot than me at the time, so I really just tried giving it the best shot with the knowledge I had, which was frankly pretty limited. I'd really love to see this game get cut down to sub-hour if a time like that could be possible.
CtrlAltDestroy wrote:
I'm replying to this thread because I once tried TASing this game, and gave up when I realized this game should be brute-forced for maximum efficiency and entertainment, similar to the old Lunar Pool movies. Peggle has massive potential to be a really entertaining TAS if someone wants to write a brute forcer. Every shot should be at least as impressive as a Zen Ball shot. As it stands, this movie is indistinguishable from a casual run that uses savestates.
When I was in the process of making this TAS that felt like the standard to me, frequently I felt pretty stumped on levels though and I knew they probably weren't the best shots that could be made. I didn't use any external tools, simply just testing combinations myself over and over again until for each shot I got what seemed was as optimal as it could be, but with a brute forcer there's definitely way more that could be done, which would totally make it faster and more entertaining as you said.
CtrlAltDestroy wrote:
Did you know that you can subpixel-adjust your shot by tapping the D-pad left and right? Here's something you did about 5 minutes into the run: In this same position, here's a shot I found after about 2 minutes of attempts (though admittedly I wasn't holding the fast-forward button when it happened):
I really wish I knew you could subpixel-adjust shots beforehand, that probably would've helped out with so many levels. I figured at the start that the d-pad control scheme would just get you the same angles, about 1 pixel per left or right press on the d-pad, but having access to even more possible angles per shot is pretty huge and probably way more efficient for getting good combo strings.
Really_Tall wrote:
First of all, it's nice to see a full TAS of this game! I spent some time working on Lua scripting for it a few years ago on DeSmuME but I only actually TASed 1-1. Here's a TAS of 1-1 that aims for style rather than speed, and here's the result of my research on RNG. I also have a TAS of 1-1 that aims for speed, but I haven't published that yet. Do note that it's somewhat unfair to compare TASes of 1-1 to a full game run, because they can use perfect RNG by changing the system clock before startup, and it is infeasible to have perfect RNG in later stages of a full run. I noticed that the submission text doesn't mention RNG manipulation (for orange peg positions in the next level) at all, which is a red flag. Along with changing the system clock to get perfect RNG for 1-1, you can manipulate RNG for the next level quite quickly by waiting on the fever screen, because all of those particles rapidly advance RNG. Even if no scripting was used to find good shot angles and/or RNG seeds, I'd expect to see at least some testing of different RNG advancements at the end of each level, to attempt to cluster orange pegs together, and to avoid them appearing in the worst possible positions, for example.
Being able to manipulate luck at your will honestly wasn't something I was sure existed for this game. The fact that you can change it on the fever screen is pretty great, I figured myself that the only way you'd be able to advance RNG would be by just stalling for longer periods of time during/in between levels but your idea makes a lot of sense thinking about it. It also felt to me like no matter what you'd still have to deal with semi-random luck with where the orange pegs show up, making luck manip mostly arbitrary, but I clearly see now that's not the case. Also, that TAS of 1-1 and Lua scripting you made look absolutely incredible, it really looks like you can get way crazier shots than what I thought was doable. I had no idea that someone tried reverse engineering Dual Shot like this already, looks like it'd be super useful for lining up the exact shots you want.
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
ThunderAxe31 wrote:
Thank you so much for your patience! I can't claim the submission right now because I already have my hands full with other ones. I'll claim it in the next days, unless there is another judge who wants to pick it. In the meantime I'll leave a quick review: I've confirmed sync with the provided verification movie. Also the execution looks really optimized. Great job!
Yay, the sync thing is really great to know, I was still a bit worried it wouldn't sync properly. Take as much time as you need for verifying
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
ThunderAxe31 wrote:
Hello there. As WarHippy mentioned, a confirmation movie doesn't need to be a TAS, it just needs to produce the same SaveRAM file you used for your submission. You don't seem to have the confirmation movie, so I'm willing to produce one myself. I'm familiar with this game and I wouldn't mind playing it again. Just tell me everything that you did in-game before recording this submission, so that I can replicate it in a new recording.
Sorry for the late response but thank you so much for the verification movie you made, that's gonna be super convenient to hold onto To be honest I hadn't gotten my own SRAM in order to make the 12:19 All Puzzles TAS, I just used an already completed save I got from the page below and converted it from .duc to .SaveRAM in order to save time (there's no way to be able to use cheats). I'll definitely make sure to create my own verification movies for TASes I create later on if they need them. https://gamefaqs.gamespot.com/ds/605448-plants-vs-zombies/saves Looking forward to working on All Puzzles again in the future, hopefully I'm able to improve it somewhat as well
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
The video below is a pretty basic demo of what luck manip could be used for in longer categories like Any% The number of times/specific frames that you do stuff like grabbing the shovel and collecting sun at affects the game's RNG, meaning near perfect manip for the plants you get from conveyor belts, the rows zombies spawn in, etc. is probably possible, which would allow way riskier but faster strategies. The main thing that makes speedrunning this game possible to begin with is that every time you kill between 35% and 50% (luck-based as well) of the current wave's HP, the next wave spawns immediately, meaning way less time you have to spend to clear levels. Link to video
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
Finished a huge improvement to quest mode 100% today bringing it down to sub 20, was able to optimize a bunch with the main things being level routing and gemswap luck manip. Hope everyone enjoys Link to video
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
Dimon12321 wrote:
I never knew PvZ on DS had puzzle levels. Looks enough entertaining for me to vote Yes. Unfortunately, you need to provide a movie file which achieves the conditions you use to begin the game.
Oh, that's kind of rough then. Unlocking puzzle mode in PvZ to begin with requires beating adventure mode first which is more than 3 hours long, making a TAS of Any% might be something I'll be up to do someday but for now it's just way too much work.
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
GamesFan2000 wrote:
This was a treat to watch! Inhuman Bejeweled gameplay is fascinating. And, to clarify, TASVideos always times from power on to last input, so the time should be a few seconds faster than what is specified. Yes vote, and I'm hoping for an "All Badges" TAS of this game in the future (or whatever those achievements are called).
Ohhh thanks for the info, I retimed the original TAS project myself and it comes out to 8:41.026. Doing other stuff with Bejeweled 3 eventually would be pretty neat too, I kinda wanna cover a lot of the PopCap games on DS tbh (or at least however many have interesting TAS possibilities)
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
Dimon12321 wrote:
I didn't expect anyone to contribute to beating this game. What a great improvement! 2,5 minutes ahead of my LOTAD. I suspected invalid matches would affect the random, but I didn't dig much into it because my brief tests showed no effect.
Thank you :) and yeah, pretty unusual game to make a TAS of lmao. Found it interesting to create nonetheless
berrimeow
She/They
Experienced Forum User, Published Author, Player (86)
Joined: 7/16/2023
Posts: 9
Location: California
Dimon12321 wrote:
Bejeweled 3 is not a really good game choice for TASing. For 2 reasons: 1) Entertainment. As we can see in this submission from 2012, it gets boring quickly. Yeah, this is one of the games which are entertaining when you play it, and not so entertaining when you watch a TAS. 2) Optimization. When you make a combo, a set of jewels, which falls from the top, is selected at random, depending on the time moment at which the combo was done. That means, each quest should be heavily bruteforced high and low in order to achieve your goals in the fastest time. I'd say, optimization is the main restriction here. Bejeweled 3 has not a big variety of goals. It has it's own set of achievements, but I don't think any of them, besides Levelord, worth its submission, as they are aimed at an extremely long play. So, I would emphasize 2 reasonable goals: Levelord (short) and Quest Mode (long), which we'll focus on below. Special Gems glossary Flame Gem - collect 4 gems in a line. Blows 8 cells around it. All gems above the explosion jump up, but it's still the fastest interruption among special gems. Star Gem - collect 5 gems forming an 90-degree angle: these are 4 variations of "L" and the "+" variation. Clears the whole row and the whole column which are traversed by the gem. Hypercube - collect 5 gems in a line. Clears all gems of the kind a hypercube was swapped with. Note: combining 2 hypercubes clears all the gems on the field. Supernova - collect 6 gems in a line. Can be made only the game drops multiple gems from the top after you've made a combo. Provides the multiplied effect of a Flame Gem and a Start Gem, i.e. clears 3 rows and the 3 columns which are traversed by the gem. Quest Mode Quest mode consists of 40 stages, introducing 11 game modes. Alchemy Vertical combos cover columns. Horizontal ones cover rows. 8 matches on each line in 1 direction is enough to cover the whole field. Things get a little complicated when you need to cover the field twice, but I think the strategy is still regular combos. Special gems are not that game changers here. Avalanche This mode is slow-pacing. Not much can be done here. Combos should be done during a fall of a gem line, to make them fall at the full speed. Making combos in other moments will pause the timer which drops a new line of gems. It's possible to swap a gem with an empty space if there is no question mark above it. Blowing a Star Gem at the end will save some time comparing to making 3-4 combos after the last needed line of gems. Balance It's about manipulating the frequency of red and blue gems. Don't let more than 3 columns be unsortable because, the more combos you do, the more gems fall down and red/blue ones are seen more often. Special gems are not that required. Butterflies Every time you make 1-2 combos, a random gem at the bottom turns into a butterfly and each combo makes it jump one space above, until it reaches the top. It's about the frequency of combos to spawn more butterflies. It's not that hard to make combinations with them. Gold Rush The main problem is making combos at the bottom to break the ground cells. Blowing a Supernova, especially, near the bottom, is the fastest strategy. Well, a Star Gem along the line with gold cells is good too, but who knows if this will be faster than a Supernova. Ice Storm It's about making frequent combos and blowing Flame Gems because of their short interruptions. Poker Obviously, to get a Flush hand, you gotta rely on manipulations of the random. Say there are 3 available combos of the same color. You do these combos and make the game drop gems of the same color to do 2 more combos, resulting in having a Flush (5 of a kind). Unfortunately, each your gem swap result gets recorded on one of 5 cards, so there are not much possibilities to prepare the field for the next set of cards without ruining your card combination. I don't think it's possible to collect constant Flushes (5 of a kind), but the rotation of Flushes and quads (4 of a kind) is definitely possible. Also, combining a special gem gives you +50 points to the result. Sandstorm / Buried Treasure You gotta clear specific areas to pull the items out. The strategy is the same as in Ice Storm. It's about making frequent combos and blowing Flame Gems because of their short interruptions. Stratamax Since we have a limited number of moves, it's about manipulating cascade combos and blowing Flame Gems because of their short interruptions. Time Bomb You gotta make combos with specific gems with timers. 3-4 combos trigger a new bomb to drop down from the top. Not much to say. Flame Gems may come in handy. Wall Blast Flame Gems are not so useful here because they may not cover much of the wall blocks. Supernova is the fastest strategy because it covers much space. I actually managed to make 2 hypercubes and join them to cover the whole field, but I think it's a slow strategy. Jewels Random The game drops new jewels at random for each of 8 columns which have less than 8 jewels after combining some of those. It also keeps track of total quantities of each color on the board, as well as the amount of possible combos, but it's not too restricted. In rare cases, but easy enough to face it in a regular play, the game drops multiple jewels of the same color, making it easier for you to get powerful jewels. In even more rare cases, you can get 5 (hypercube) or 6 (supernova) similar jewels in a row just dropped down for you. This is the main point of manipulation for the fastest strategy to complete some of the non-trivial goals. In all modes, except Classic, the game doesn't let the player run out of combinations by always providing a set of jewels which make at least one combo possible. For a TAS, if you get into this situation, that definitely means you gotta replay a stage or rewind continuous playthrough to get a better random. Here is my TAS which beats the Quest Mode. It's rather a low-effort run, cuz' I didn't bother beating each stage in the fastest time. A human may spend dozens of hours on each stage to get objectively fastest time, I think it's a work for a bot to mesh buttons. Link to video Here is the input file. I don't know why you'd need it, but anyway. User movie #638173207203289282
I finished making a TAS today for beating 50% of the quests in Bejeweled 3. I'm not sure how known it is but I think it's possible to manipulate the game's luck a bit farther by purposefully making invalid matches at precise moments, causing the gem colors the game is about to drop to be rerolled (this manip doesn't seem to work all the time such as sometimes during ongoing combos, but otherwise it's pretty helpful). It's also possible to manipulate the luck of the starting gem board on each quest depending on the order you select levels in, which is especially helpful for levels that you want certain beneficial starting boards on (Balance quests starting with a lot of reds/blues, Stratamax quests starting out with more 4+ gem combos, etc). Still a pretty luck-based game overall, but with this manip and fast level routing overall I managed to get this run under 9 minutes. File link: User movie #638269707232963338