Yeah, that PB refill slowed you down by around 2-4 frames I think. The morphball jump through the upper block can also be done a bit faster by optimizing the subpixel position, as shown. Also my Botwoon refills were pretty bad as well, only getting 2 Supers, and even not having full energy after the fight (288 instead of the full 299). Although it's only 11 points, it could still slow you down a bit later, as it would require you to eventually get one more health refill in order to have enough energy for LN. When you reach him in your new run, you should definitely try to get full energy after the fight already.
Yeah the refills where not the best and I'll try to make them better when I reach him, though when I did him in that run I simply killed him as fast as possible (I think the fight itself is unimproveable) and his drops in no way forced me to do the fight over thus I continued. Now that I have a decently optimized run to compare to things will be easier, I really hope that I will be able to skip those pipebug refills in norfair...
As of now I'm at the diagonal room in crateria, I think I'll try to see if I can find the new strategy but I'm probably not going to keep at it for to long, after all the run will not be published and we are talking about 1 maybe 2 frames.
Also I remember moozooh saying he saved 2 frames at the torizo, where did you gain those frames? I improved by 2 frames from my old run there but it was not anything I did in that room, I improved the entry and landed 2 frames earlier but found it impossible to do anything different in the jumping part.
49'20
I was able to skip both frame delays in Ceres (one before the scientist room and one before the steam / debris room). I always knew it was possible somehow, but only just now got it to work.
Great Taco :)
I also have good news, after a couple of intense hours I've got a time of 295 frames for the diagonal crateria room :D
I enter with a perfect subpix position, keep a constant speed of 2 px/f and have tried to skip as few armpumps as possible, I also believe that the three pickups are from the optimal enemies (maybe the second or third enemy could be exchanged for the one between them, but I found no frame that allowed me to). The rest of the possible improvement for the room is to find more optimal positions to shoot the enemies thus allowing more armpumps and boring as it is AFAIK the only way to do so is by bruteforcing. Feel free to do so moozooh and others, I'm out.
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Taco wrote:
49'20
I was able to skip both frame delays in Ceres (one before the scientist room and one before the steam / debris room). I always knew it was possible somehow, but only just now got it to work.
I daresay that was nice:) Good job Taco, you certainly are a machine.
OMG 0_0!!
Taco is herewith the one and only king of Ceres! 49'20 is just beyond me. Your incredible patience, dedication, and will to keep trying Ceres over and over for ~a year is indeed something no TASer would ever be able to manage, and I'm glad for you that your hard work finally payed off. This is subpixel optimization to the absolute limit! There just can't be any more improvements left (except the extremely unlikely elevator boost). MUCH respect Taco, you are simply incredible!
I really can't wait to see you in action on Zebes, and whether you can manage the impossible to save at least one frame over my TAS demo until pre-Torizo. :-)
Also very nice diagonal room improvement, Cpadolf! I always felt there is more than just a frame over 299 possible, but didn't got around to try myself yet. AFAIK hero, moozooh, and evilchen spent around a 1/4 year on that room alone, bruteforcing every possibility there, and now you come out of nowhere blowing everything away in just one day of work, as if it would be the simpliest room to optimize. From your very first serious attempt we could see in your 0:24'04 TAS, it was already clear that you were just superior in this room, and now we have the proof once again. In fact, you were always showing exceptional skills when it comes to generally geting refills in a minimal amount of delay time, your last SMI WIP showed it pretty clearly as well. Great Job!
Joined: 11/14/2005
Posts: 1058
Location: United States
I didn't spend even close to a 1/4th of a year in that room. What are you talking about Saturn? Please refer to my old 1.51 wip to see how I clearly did not give a shit about optimizing that room.
Great job Taco!
Joined: 6/13/2006
Posts: 3300
Location: Massachussetts, USA
Saturn wrote:
I really can't wait to see you in action on Zebes, and whether you can manage the impossible to save at least one frame over my TAS demo until pre-Torizo. :-)
I love how you manage to praise Taco through the roof, yet still manage to squeeze in just a tiny little "saturn is awesome and the best" moment.
You're starting to get better, by showing WIPs, removing the "my work is perfect and I appreciate how everyone is slowly getting to my level" comments and encouragin more, but you still need to remove the urge to insert "saturn is awesome" comments completely.
Hmm... tas forum used as a therapy... someone should trademark that.
Maybe a nerdy mascot.... like Clippy... Mr. Savedframe... some user feedback ("tas-forum saved my life! I can now deescalating count to 10 three times as fast before I snap"), and Link after 5 hours of "Listen!" nonstop as commercial actor.
SCNR.
49'20
I was able to skip both frame delays in Ceres (one before the scientist room and one before the steam / debris room). I always knew it was possible somehow, but only just now got it to work.
Well I can't imagine anything other than that I found another enemy to refill from without stoping (probably the sceond one, that one took the longest time to find a good spot to shoot) because I can not imagine that they have been working with this strategy for any large amount of time without result. And I still think it is improvable by at least 1 frame, but finding better spots to shoot from just takes so long time.
hero of the day wrote:
I didn't spend even close to a 1/4th of a year in that room. What are you talking about Saturn?
I think he was talking about
moozooh wrote:
So far hero, evilchen and I have been bruteforcing it,
this added with the fact that their run has been standing still at that room for a long time.
It didn't mean we were doing it every day for all this time akin to Taco doing Ceres, it's just that no-one really had the motivation to optimize it. :) Well, except evilchen, who I know dedicated a lot of time to it and managed a 298 runthrough by himself (somewhat differently from hero's version).
I don't know which version to go with, as I don't feel too good about involuntarily "waiting" for players not directly involved with the project to achieve what I want and reap their results afterwards, so I'll leave this decision to evilchen. On a lighter note, this weekend promises some TASing time, as long as nothing unexpected happens.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
I don't know which version to go with, as I don't feel too good about involuntarily "waiting" for players not directly involved with the project to achieve what I want and reap their results afterwards
Sure I will. It doesn't leave me much hope for success, though, given how unpredictable dealing with slopes is: even if you're faster at one point, you may get fucked over two screens later exactly because you were faster. My experience with SMR is very rich on this assholery.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
True, personaly I'd say that both my successes with this room are largely based on luck. Sure this second time I was very systematical when searching for the best frames to shoot but there could just as well be so that a subpixel position of for example 4096 instead of 0 at the begining would have saved me or lost me a couple of frames. This room is stupid.
there is more than just one slope in the game.
Could it be useful to make a mathematical model and write a brute-force-program that'll output the fastest way / optimal subpixel position, or would that just be a waste of time?
The unfortunate part is that almost all slopes aren't straight lines. Based on my experience, they should actually look like a sinusoid going 1/4 a cycle each block, hence why a miniscule change in subpixel position means a drastic change of incline factor (basically, on such slopes, Samus's vertical movement is constantly alternating between going "slightly down" to "almost falling") that modifies vertical speed of the arm cannon projectile (the variation isn't that big, though, but there's no easy way to get the modifier you need). That combined with the fact that Samus moves 4.75 pixels per frame at full speed, and that arm-pumping makes slope advance somewhat nonlinear (you skip a horizontal pixel worth of incline changes or something), makes it very hard to construct a mathematical model without making it directly out of the game engine.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
that leaves the options:
- do it manually like before
- try DeHackEd's LUA-Scripting
- sacrifice a goat to the almighty Kejardon so he might bless us with disassembled knowledge ;)
it is! it is...
while iam not so good at subpixels .. this room brings both X and Y subpixels which makes it even harder for me :/
edit:
moozooh wrote:
Well, except evilchen, who I know dedicated a lot of time to it and managed a 298 runthrough by himself (somewhat differently from hero's version).
i didnt even check how much i got.. big lol! i just saw yeah all pickups and never stopped -> success
- sacrifice a goat to the almighty Kejardon so he might bless us with disassembled knowledge ;)
actually, I prefer homework solutions over goats.
Unless you mean these kinds of goats. ;)
Anyways, if you want, I'll work out a formula for arm-pumpability on slopes when I get the time. But really, I'm pretty sure it's about what you'd expect it to be - slopes are approximated by pixelized paths rather than a strict slope, and Samus can only armpump when the path is two horizontal pixels in a row. Which would depend virtually entirely on Samus's pixel position (i.e. ignoring block and sub-pixel position).
kwinse wrote:
Kejardon wrote:
Kriole wrote:
Samus is damaged by a Rinka in the opening.
That's a script action; no damage.
... it just dawned on me I know way too much about SM.
That's useful info you put up right there.
AFAIK, each sloped block is the same 16 pixels wide. If we know what pixels should be "skipped" inside each block to not lose contact with the ground, the formula that determines how Samus's position changes on them, and the amount of identically sloped blocks in a row, it should be possible to calculate the optimal path with the least amount of pumps skipped with the use of a realistically manageable script. With unrealistically small usefulness, since the slopes are either too small or have other obstacles on the way. :\
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Casual users won't notice the 1-2 saved frame(s) in that room anyway. Why spend too much time on it, Cpadolf's strat is very good.
Taco, you're godlike. :)
Saturn, stop saying Taco spent a year on Ceres.
Hero of the day, how's your TAS progress if I may ask?
Cpadolf, got my laptop back. ;)
<Deign> .dice 1d1999191023443691
<BisqBot> Deign rolls a blackhole and destroys the planet.
