Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
What's happening?
My current approach is not very different from the IL (that skips the entire level): gather speed on the start with a left boost instead of a right boost (may need to do a half reverse boost).
Then, the stage is a bunch of curves: 'thankfully', the edge of these curves are straight vertical, and you can bounce off them. The strat involves falling down, doing a frame-perfect pixel-perfect position-perfect orientation-perfect on the black ledge (orientation for the second bounce) into another frame-perfect pixel-perfect position-perfect orientation-perfect bounce on the yellow ledge, sending us straight to the goal.
If this succeds, it will be one of the craziest strats in the TAS, and may save up to 40 frames. Hovewer, it's going to be extremely hard to optimise; movement will be extremely tight and unforgiving to route.
This is the last of the three extremely hard stages in World 1 (Curvature, Progenitor, Anaconda) and it's the hardest on par with Curvature.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
I've completed a pretty satisfying TAS of the game and am ready to submit. While fooling around, this little thing happened:
Link to video
I cal this 'physhic arrow' because the arrow you fire becomes really wierd. Also, on the frame the glitch happens, desmune freezes for about 1.5 seconds per frame for about 5 frames. I didn't look further than that, but there are probably interesting things going on. Will lend the savestate for anyone interested in looking around this.
Also, if you wonder what happens when you do normal actions (sword slash and boomerang):
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Muhahahahaz wrote:
-snip-
The rules on improvement are: if it's faster, it gets published 99.9% of the time. The only exeptions are emulator changes (load times) and obscure cases.
You're welcome to submit any improvement you find, even if it is less interesting/reuses 90% of input, but if you reuse input you should co-author whoever inputs you're using.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Weegeechan15 wrote:
Oooh, nice! Anaconda looks like it'll get pretty interesting at a TAS level of optimization. Keep up the good work!
The IL (individual level strat) isn't too interesting: boost, go up, fall off, done. The all banana strat for this level, however, will probably turn out to be one of the most impressive and mechanical strats in the entire run.
Here is the world record, for reference:
Link to video
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Next level up is Anaconda, aka 1-6, aka WINDO6. There was 1-5, but it's a bonus level that progresses whether you die or not; as such, you finish them by falling off as fast as possible.
I've put togheter a strat in about 5 minutes that shows what you can expect (1 second behind WR) and am currently researching a good TAS strat.
Link to video
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Link to video
Completely redone the strat (again) which saves a whooping 4 frames.
I think this is as optimised as it can be. I won't bother doing the Fladdermus bounce for 3 reasons:
1. You need a stupid angle to get it. This angle means completely redoing grinding, but with about 10 speed less than what is in this current run.
2. It's very inconsistent. Even if it gives 30 more speed for about 1.9 seconds, going faster on the rails (the grinding part) I think is overall faster.
3. It's stupidly hard to get. Trust me on this one: would be more frustating than trying to do Curvature. I don't want to spend 20 000 rerecords just to be 'damn, the angle was too wide, time to restart!' and I don't even think it can speed it up.
This is why I think doing this strat is faster than doing a Fladdermus bounce, and I can safely say I'm done with this level (both normal strat and all bananas). Thank god.
Also just discovered you can record AVIs within DesMuNe: I'm going to be doing that for all tas world records for now on, but not for the others videos (this way I can go back and watch in slow-mo what went wrong, my current speed/angle when it went wrong, and improve it and potentially do something better)
Movie file
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
In the beginning Malleoz created the heavens and the earth. The earth was formless and void, and darkness was over the surface of the deep, and the Spirit of Malleoz was moving over the surface of the waters. Then Malleoz said, "Let there be light"; and there was light.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Link to video
Beats the Fladdermus TAS by a whopping .07 seconds, the human record by .30 seconds, all while collecting all bananas.
The normal strat (aka not all bananas) I estimate to be around 51.3x .
Movie (not bothered to upload it to userfiles)
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
AKheon wrote:
Would a multi-TAS of the three Wright games (without SRAM) be vaultable? Would a trivial thing done three times simultaneously become non-trivial?
Would holding B and doing stylus inputs separately on each game (because the stylus inputs conflict almonst never) be something to watch that doens't make you want to say 'uh, what's the point?'
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Just discovered something: found the ram adress for Aiai's angle (the location he looks at) and for SOME REASON, Aiai's angle is slightly adjusted to the right each time you start a new level (before the start of the level, where you fall off).
This means doing a right boost is faster in situations where no other factors come into play. All of the levels are unchanged, but it is good to know.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
Entierely redoing Curvature (bad start). Currently at around 7500 rerecords for this level.
Link to video
New start that saves around 18 frames compared to old.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
More than a week without an update?! What has been happening?
Actually, nothing that's interesting enough to show. I've been optimising grinding:
Link to video
All of it is done manually (since this is the only time I will have to do it for such a long period of time). The main point of doing all this grinding right now is because it will be easier to find an optimal Fladdermus bounce this way (since the 1-3 all bananas strat cannot use that bounce; saves me a lot of trouble optimising it now) LOL DISREGARD THAT GOT THE ALL BANANA STRAT BY DUMB LUCK WHILE TRYING TO FIND THE TIME STRAT L M A O just kidding was unoptimised
Now, the thing with doing this type of grinding is that sure, it's optimised, but if you watch closely it is optimised very tightly: all movement is in a array of about 16 positions (all going forward) and I grind so hard that in this video, Aiai looks like he's grinding the left side of the rail, but in reality Aiai doens't actually grind off the top of the left side of the rail, but the side!!
Why is this important? When we want to get back on the right side of the rail to execute the Fladdermus bounce (because none of the grinding frames allow to get the Fladdermus bounce if we get 'back on track' without slowing down, and I've monitored them) we will have to:
1. Find the earliest frame in our grinding section we can get back on track and still get the bounce (although it won't result in a full Fladdermus bounce because the angle is too low, see the "cya nerd" video).
2. Redirect grinding to the right side of the rail without falling down inbetween the two rails, because it would slow us down, just enough that the Fladdermus bounce makes us travel enough distance (since we will have the right angle). We don't want to go too much right or it will waste one frame.
3. Find the right frame to start going upright to hit the Fladdermus bounce, hold upleft as the collision happens (2 frames before the hit) and hope it works.
All of these are not easy things. Finding the earliest frame at which I can redirect the grinding to the right side of the rail, because starting too early will waste frames and starting too late will not get the bounce: the problem is it is reliant both the frame at which we start grinding to the right side AND the distance we go right: I could start at an earlier frame (that didn't get the bounce before in the initial situation, which is the main problem: this adds a lot more situations in which I can start redirecting the grinding right), but if it's too early we'ill waste time; I could start later, at a frame in which we alderaly get the Fladdermus bounce without redirecting, but in that case redirecting us right would lose 1 to 2 frames, so now the 'sweet spot' between these two options has to be found.
Once that sweet spot is found, the grinding to the right itself is another problem: we have to keep an exact position to get the right angle for about.. 12 to 30+ frames, all while grinding in itself! So a frame that could work in theory would not work in practice because the array of inputs we have while grinding to the right of that rail is even more limited, so once I start grinding I have very little options and unless a good position is found, the directions are so limited just a pixel different in the stylus touchscreen position (230,0 or 231,0 for example) can mean the difference between I will be forced to either fall or grind more to the left which will not get the bounce.
All in all, extremely tough to find. It's a World 1 level though, so it's not surprising that it's this tough.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
GoombaHeart wrote:
I thought about starting with Super Mario Bros or something, but I'm found myself much more interested in SM64. Though I know it's rather complex, and has very high TAS (though that also goes for SMB).
Don't. The only time I would ever recommend it is if you really need to get the hang of how basic tools work, but you've hopefully gotten past that.
I myself am a 'complete' beginner at TASing and am currently tasing a very very high skill DS game (one of the hardest skill-wise DS games, if not the hardest on par with Super Mario 64 if we don't count games with RNG).
These 3d games are great because they have a relatively low skill floor, but an extreeeeeeeeeemely high cap. That is the great part of speedrunning, for example, Super Monkey Ball; you have, for most levels, a range of strats in the double digits ranging from extremely easy to it will take you 8000 rerecords on a single frame on TAS. New things are always being discovered, even 15 years after the original game.
Hovever, there are a number of things to consider:
Be good at your game. This isn't needed for all TASes, but for the high-level games it is definitely needed. I became one of the world's best players before trying to tas Touch&Roll (and co-author with the best player in the world), although that level of mastery isn't often needed.
Constantly seek improvement. Redoing an entire section or an entire level isn't fun, and you'ill do it pretty often: so it is also important to stay fresh. You can do a multiple of things in your game: create strats, better routing, trick/glitch hunting.. Also stay fresh with your game, when you're frustated or something else, go outside, take a break, go for a walk, relax.. I personally play piano to relax. It's also important to devote your entire free time to your TAS as long as it stays healthy to yourself: you are doing a work of art, and you will be recognized later in your TAS as the unsung hero, the nameless artist.
Consult communities. Unless you are TASing the most obscure-hyperro-bizarre game, there's always at least one person, one group which can help you, which has done what you've tried to do. If you are new, I recommend checking out the TASvideos IRC: lots of cool people there. Also try speedruns communities, which are generally very receptive to TASers.
Have fun! If you don't enjoy TASing, you may as well stop right there. Enjoy what you do and take pride in it: it's your own little thing you'ill build up in weeks, months, years.
Editor, Experienced Forum User, Published Author, Experienced player
(818)
Joined: 5/2/2015
Posts: 671
Location: France
TASVideoAgent wrote:
i looked in the sewer there was u
lol
Congratulations Samsara, you turned a TAS made in intent of the vault into a dance showdown by TASing like a professional. I hope you are happy about your TAS and your result.
It was really impressive to watch and I did not expect it, with all parts of the run really polished up to subtle little things, so I'm voting yes.
My current approach is not very different from the IL (that skips the entire level): gather speed on the start with a left boost instead of a right boost (may need to do a half reverse boost).
Then, the stage is a bunch of curves: 'thankfully', the edge of these curves are straight vertical, and you can bounce off them. The strat involves falling down, doing a frame-perfect pixel-perfect position-perfect orientation-perfect on the black ledge (orientation for the second bounce) into another frame-perfect pixel-perfect position-perfect orientation-perfect bounce on the yellow ledge, sending us straight to the goal.
If this succeds, it will be one of the craziest strats in the TAS, and may save up to 40 frames. Hovewer, it's going to be extremely hard to optimise; movement will be extremely tight and unforgiving to route.
This is the last of the three extremely hard stages in World 1 (Curvature, Progenitor, Anaconda) and it's the hardest on par with Curvature.
This is probably my favorite screenshot in the entire game.
I did some background exploration:
Link to video
Beats the Fladdermus TAS by a whopping .07 seconds, the human record by .30 seconds, all while collecting all bananas.
The normal strat (aka not all bananas) I estimate to be around 51.3x .
Movie (not bothered to upload it to userfiles)
The rather awkwardly-named homebrew.
This game has some potential, will try to get something out of it in the couple next hours - maybe the next day.