Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
This was pretty samey to me. I felt like I saw everything there was to see by about the second stage, even if the ceiling bonks and pause buffering were pretty cool. But that's a criticism of this particular game, not the amount of skill and effort you put into the run.
A solid Meh from me. Could've been a Yes if this game were more varied...but it's not :/
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
So I'm not aglasscage, Shinryuu, FinalFighter or pirohiko...but I've got a WIP run for you guys. It's zipless, and plays up to the beginning of Wily 2. It's a little rough around the edges, but I think it's a good start.
Link to video
I'd love to hear what you guys think! Plus, I could use some advice. Mostly for the Quick Man fight, but also for Item 1 routing.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
Link to video
Wanted to post my art% proof of concept. The drawing system is super clunky and limited, so I think it's really fun to watch the little cursor draw out decent-looking art with cold and calculated precision. With a good artist or community submissions or something, I think this could be a fun little freerun.
I was also toying with the idea of making a narrative where Potato Brains are shit artists (played by me in real time), but it just isn't the same when you can't hear the offending artist justifying his incomprehensible strokes :/
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
Hate to be a bother, but has anyone had an issue with the scroll wheel randomly deciding to jump ahead by several thousand frames, instead of the usual 6? This happens to me seemingly at random, usually when scrolling up.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
Hmm. On 2.2.2, it looks like I can only select two rows at a time, and can only affect them with the arrow keys; I think the ability to select more would be useful. The "paint brush" thing and arrow nudging are definitely pretty useful! Thanks for the tip.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
What's the recommended way to use virtual pad with TAStudio? Currently, I can only figure out how to use it with "Record input" enabled, which is an exercise in extreme frustration as I accidentally overwrite inputs and can't undo them (clearly I need to set a bigger undo stack, but still).
It'd be nice if I could fiddle with aX and aY for a given frame via virtual pad, rather than typing out byte values.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
You're absolutely right. However, curiosity (and a strong urge to do anything but study for finals) got the better of me, so I threw this thing together:
Link to video
It's not entirely devoid of merit, but 48 squares really is a lot...you have to make 12 identical trips in the box stacking one, and the fireman one will probably put you to sleep (even if the RNG is god-tier). I wanted to make a no-damage run, but if you've ever played the fireman game, you probably know that I've come as close as anyone will ever get, at 2 lives lost.
Even if it's not submittable, at least we know what it looks like to be good at Pictionary!
Beyond this, I have one idea involving the drawing mode, but it involves actually being good at drawing...
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
Just to address everyone's comments: Mothrayas is absolutely right about the firefighter game. You could delete all those inputs and the outcome would be exactly the same...it's 100% playaround.
While I personally think the box stacking one turned out kinda neat, I don't have much to say about the other two. The only reason I chose suicide is that it's the faster option by far in all four games. I was hoping it'd be a mix, but that didn't work out.
A 100% TAS is in the works. Keep an eye on the Pictionary thread if you're curious.
Thank you for your votes and comments!
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
http://tasvideos.org/5939S.html
Finally found the time to submit this thing! My submission text is pretty long for such a short movie, but I think I made a pretty convincing case for it.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
Mothrayas wrote:
If you do this, your run no longer is eligible for the Vault.
You don't suppose doing the minigames could be considered a goal or category? "Contains speed/entertainment tradeoffs," maybe?
Don't get me wrong, I'd be thrilled to see this run published in any tier! I'm just curious when something stops being a category/goal and starts being trivial and stupid.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
c-square wrote:
It looks like you slow down...
DrD2k9 is exactly right. It looks sloppy, but it's worth it for those sweet, sweet first-framers. Even on the ones where it looks like I don't waste time, I actually do; notice how early I let go of right for SWAN. It's a 128 frame rule there, so in some cases, you could waste two whole seconds and still lose no time!
I just looked at your submission, Spikestuff. Seems oddly familiar... :p
Masterjun's comments definitely don't bode well for us, but he did say this:
It also has video game elements... However, these are skippable...
I have high hopes that people will consider it a video game if we don't skip the video game parts!
I do have plans for a deathless TAS. But oh man...have you tried doing the fireman game??? It's barely possible to complete the image, let alone save all the innocent lives!! I have a WIP right now where I barely save enough clones to reveal the image (48), but the time runs out before the guy hits the square. I'm sure there's some frame where we could make it work; then we could rebrand the run as a "minimal damage" run, or something like that.
Btw, how do submissions work here? Do I sbumit things for vault or moons, or do I just hit submit and let the judges decide where it should (or shouldn't) go?
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
I decided to put my faith in the "TWINS" sequence. Here's my TAS:
Link to video
I managed to save 6 seconds from my old movie, mostly due to this better word sequence. I also found a faster way to die in the box stacking game, as well as manipulate a faster frame for the firefighter one. The first frame dice roll thing probably didn't help much, but hey, it's still good to do it.
I might submit this.
Btw, the scene in the thumbnail was an unintentionally hilarious message that I got :D
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
I had an RNG breakthrough last night on my stream! I was trying to manipulate the congratulations message by finishing the word on a different frame, but I noticed that it weirdly seemed to affect dice RNG when I did that. That discovery you had about entering the number of teams seems to apply to any group of lag frames. Due to the text entry frame rule, this means that we can easily manipulate a first frame 6 roll, every time, while wasting zero frames across the whole run!
Here's my theory on why this works: Pictionary probably has an RNG variable that updates every frame, except lag frames. Whatever arcane RNG math gets performed takes into account the current frame, as well as the previous value. After sitting through different numbers of lag frames, even though we make a roll on the same frame, the formula is looking at a different previous value in each situation, and so the results are completely different than before.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
DrD2k9 wrote:
waiting just two frames before selecting the game mode changed the die rolls enough to save over 30 frames...I still enter the team name on the same frame...
So what you're saying is...we can manipulate better dice RNG...by wasting ZERO FRAMES!?!? That's incredible!!!
DrD2k9 wrote:
Please don't take my comments wrongly...I just feel that there are too many factors to automate everything..
You're good! And I agree. While something like frame 530 (2145 frames) is probably not worth testing, frames 524 and 529 (889 and 770 frames) are probably both worth testing, and maybe a few more. These are hundreds of frames better than the average sequence, but close enough to each other that dice RNG might fuck you.
Like I said in the comments of my lua script: "I think its results are worth considering for a TAS." Not "using without thought!" It's only meant to give us a few good starting points.
Here is the script, if you're interested.
Btw, I've just console verified the sequences we found...they're 100% accurate! Your spreadsheet is extremely helpful, too; I didn't record it, but I may now be the world record holder for Regular Mode :0 :0 :0LOL NEVER MIND
Apparently this sequence thing has been known for some time. Weirdly, all the runs on the leaderboard were done earlier this month. They seem to like to grind for the "NO, PIN..." sequence. But maybe our faster sequences might still be useful anyway.
You know, if you could play it off like you're fidgeting with your phone, checking texts and stuff, making these "blind" guesses would make a great party trick :D
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
DrD2k9 wrote:
Great....but if your script doesn't account for changing directions, what's the point...?
You're absolutely right. Additionally, I forgot that the cursor always starts on N!!!!!!!! My old code assumes that it always starts right where you want it.
Are you familiar with dynamic programming, or DP? It allows us to compute all possible solutions (in this case, which direction to go from each letter), while storing stuff along the way so we don't have to compute the same thing more than once. With a few simplifications, this problem lends itself really well to this technique.
Here is the output of my word typing DP. Out of curiosity, I left in the results of the dumb, "hold one direction" algorithm to compare them; they're the second numbers after each word. The new numbers are much smaller, but interestingly, the results barely change! The only difference is that my original "NO, PIN, THINK" sequence is no longer a relative minimum. Ouch...
I plan to post the script after I clean up a few things.
The frame rules are going to be a pain to deal with. I thought about having my code round up to the next multiple of 128, but that's not really how frame rules work, is it? However, the results of my script are so much better than the average word sequence that they'll most likely still be the best. It may be worth it to try all of them out.
DrD2k9 wrote:
Something else to factor in is the time required for die rolls.
I don't think this is going to have a big enough impact to make us choose another word sequence. In my fun run I think I had to wait like 20 frames for one roll, but otherwise, a 1 in 6 roll is extremely quick to manipulate. The way I see it, all possible runs are equally affected by die roll manipulation, and I doubt we'll be unlucky enough to see an outlier. But you never know...
Testing frame 529...
TWINS (124)
SWAN (108)
COMEDY (168)
LEFT (160)
LIP (112)
STOP (116)
26 characters
788 frames to type + 44 frames of waiting = 832
!!BEST SO FAR!!
My original sequence takes 1092 frames. Yours takes 1148 frames. None of these numbers account for the 128 frame rule. It also assumes you type each word by either holding left, or holding right; it can be optimal to stop and turn around, but my script doesn't account for that.
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
DrD2k9 wrote:
Thanks for the mini-lesson.
Happy to help!
DrD2k9 wrote:
Does it take into consideration how many frames you have to wait to initiate the other sequences...?
sum = sum + spacelessLen(word)
...
if sum < best then
emu.print("!!BEST SO FAR!!")
best = sum
elseif sum [greater than] worst then
emu.print("...WORST SO FAR...")
worst = sum
end
(I wrote [greater than] instead of a greater than symbol because BBCode bizarrely parsed it as an HTML tag)
Nope. Because of the stupid accelerating shoes, I couldn't figure out how long a given phrase would take to type without actually typing it, so I just looked at the shortest length. Although, I guess we could come up with an estimate based on holding right the whole time; then we could factor in the waiting time.
DrD2k9 wrote:
my experimental WIP of the TAS currently uses the sequence starting with "CUP."
Can't wait to see it! Is it finished?
I'm going to code up my estimate that I described to see if some longer sequences might not actually be faster, so don't submit it quite yet!
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
I went and wrote my own script after all.
I'm too lazy for a spreadsheet, but here's the output from it: https://pastebin.com/FBfEpLfU
You can ctrl+F for "BEST" and "WORST" to see the results.
Looks like we're off by 1 frame from each other somehow...but we got the same data. I ran mine out to about 600 sequences, but only read 6 words per sequence; I just had the TAS in mind.
Lua script on pastebin, if you don't feel like downloading it.
It's pretty sloppy atm. Read from range is much less useful than I thought. It converts the bytes to an ASCII string, so you have to manually convert them back and then re-encode them. I should just read the bytes one by one...
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
DrD2k9 wrote:
Thus far I haven't gotten to this sequence.
Yeah, that's fair. It's the 24th sequence. I'm curious, what's your method? I was theorizing a lua script that mashes start and reports any changes in offset 0x0448, then loads a savestate at the name entry box and tries the next frame. I may decide to code it up if it's not too similar to what you're doing.
Here's some data on the sequence I chose, which has 29 letters:
Word/Phrase: Frames to type -> frame rule hit
NO: 13 -> 128
PIN: 128 -> 128
THINK: 182 -> 256
SHIN: 140 -> 256
HOUSE: 237 -> 256
THIRD FLOOR: 286 -> 286 (end of movie, frame rule doesn't apply)
Can you tell me the 27 and 28-letter sequences? I want to see if they're faster in practice than my one.
DrD2k9 wrote:
Part of what I'm looking into with my work is total letters of input for the run.
I'd argue that that's an insufficient metric. Not all words of the same length take the same amount of time to type; just look at THINK and HOUSE in my table. But I'd also argue that, unless you're right on the edge of a frame rule, it most likely makes no difference. With that in mind, I'd say it's the second best metric to the "spend a few hours making a whole TAS" metric :p
Experienced Forum User, Published Author, Player
(248)
Joined: 3/15/2018
Posts: 229
Location: United States
"DrD2k9 wrote:
I'm curious to see the community's response if you and I both end up submitting runs using different rules/strategies. Perhaps one will be deemed vault eligible.
Hell yeah! Let's see where this goes.
While I disagree with you about the minigames, it almost doesn't matter. Because if you and I press start on the same frame, I do the minigames, and you don't...we still get the same sequence of words. Meaning your findings will be optimal for every category, and I can steal your work collaborate with you easily.
Have you found a better sequence than "NO," "PIN," "THINK"...?
Due to the 128 frame rule between entering your guess and the board appearing, the first 6 words almost don't matter. The final word (on which we end our movie) is much more important. And since my final word is "THIRD FLOOR," I think we can do much better.
I assume you've figured out the text encoding, but here is the table file anyway. Check out RAM offset 0x0448. Also check out ROM offset 0x1C84C while you're at it! I highly recommend it.
c-square wrote:
I think this has a high chance at getting published
Sweet! I'll probably hold off until DrD2k9's lua script (I assume) finds some faster word sequence. Knowingly submitting a likely suboptimal run just to see if the rules stick is probably a sin ._.
What's the TASmania project, btw? I had trouble finding results that weren't about the actual island of Tasmania...
