Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Very nice run. Also very entertaining during the ship levels. I only wonder why you didn't use the scroll glitch to enter pipes faster like in the any% run, and especially why you didn't got 99 lives having more than enough opportunities to do so during the 8 worlds.
Anyway, it's a pretty big improvement and deserves a YES vote. Good job SMB3 TAS team. :-)
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Oh my god, I can't wait for the WIP, inichi!
Your new trick reminds me on my Secret of Evermore discovery to boost attack and defense by deequipping ammo as well. Except that I never thought it would be possible in CT. What you describe is really marvelous and absolutely unexpected once again. Bravo.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Inichi is crazy! Being busy with CT, he just randomly tries some SMW stuff and pwns every existing and long time worked out run with his discovery he found independly from Mister. From the very beginning when he showed some of his new tricks, I immediately realized that this guy has endless potential to find absolutely unexpected improvements, no matter what games he tries. It was just a matter of time until we see this coming for yet another popular game. I'm very impressed.
Looking forward for the improved version. :-)
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Although I'm not much into SPO, I enjoyed this one. Aiming for the ingame clock in every fight is the best decision one can do, because realtime depends always on the emulator you use, and thus is not a solid number to compare to.
It's a bit funny that loser Bear Hugger takes almost twice the time to defeat compared to the ultimate champ. :-P
This movie beats or matches all the previous records, and is a improvement to the previous one, so I vote YES. Well done, Phil.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
I didn't have a lua51.dll on my system at all (did a search), but gocha's v12-beta2 included it now. It didn't change anything in the functionality though, the browse button still doesn't work no matter if I click on it in when loading a smv (movie play), saving a smv (movie record), or in the "Run Lua Script" window. :-(
Unfortunately I don't understand anything in programming to analyze the source code diffs, but from what you stated, gocha, my natural guess is that it has something to do with the modification of the Makefile.mingw thing. Guess I will just use the beta20 in that case, since it isn't really that different from the final v11 version. However if this continues with the later releases, then I'll be stuck on this beta20 forever, so I really hope this "browse" button bug could be fixed someday.
Oh, and I just tested, this exact bug also happens on your v1.51 improvement. What the heck...
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
15 sec to improve from a new hero-run? Unrealistic, sorry. ;-)
But I admire your positive attitude to still believe in it. This is a important quality a TASer should have, as it can urge one to make impossible things happen someday.
I'm also glad you progressed so quickly in the last days. Did you manage to get a similar Ridley KO-grab like I did in the RBO demo?
Also a hint for Botwoon exit without HJ-Boots:
Try to not use a walljump from the breaking wall like you did in v1, but rather clear the wall with a momentum-full jump, even if it would require you to wait for the 2nd block to blow off. It would still be faster in the end than what you did before.
But very important (assuming you don't know about it yet):
Make sure to not exit Botwoons room too quickly, because if you do so, you won't trigger his death. When you return to that room, he will be alive again, which would mean that you'd have to redo everyting again from his fight on. Delay just as many frames as needed when exiting the right door after the fight, and be sure to check back the room whether he is really dead before continuing the run.
Cpadolf:
I see, bad luck. At least you got some beta run to compare to that way, which 1. eases the redoing part alot, and 2. allows you to eventually find even more improvements having more knowledge in this rooms already.
And about which run should be published, I think both would deserve it, as they both achive their unique goals, and will be of extremely high quality. Hero's one will be more appreciated here, but yours gets a 0:23 the first time ever in SM's history, which is a special thing as well.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Great work gocha! Thank you very much to contribute so many new and useful functions to this nice emulator. :-)
I was glad to see a final improvement11, but now I noticed a error in it:
For some reason, everytime I try to click on the "Browse" button to load a smv, it doesn't open the additional window where I can pick one from the list. The "Browse" button acts like there's no function programmed behind it at all, and nothing happens when I click on it. I tested it with improvement12 beta1, and it was same. Even the new Lua Script's Browse button didn't work as well. I then tried it with improvement11 beta 20, and it worked just fine. I tried to check the changes that were specially made between beta 20 and final, but you only listed the general ones, so I don't know what's the reason for it. Do you know why this happens, or what could be wrong on my side? Any help would be greatly appreciated.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Cpadolf:
Thats why you should "always" plan your runs ahead before working your ass off to optimize them just to find out that there are some inaccuracies in the ammo/health management.
Hero:
Well, 15 seconds despite a already near to the limit optimized run are still alot. This are 900 frames after all. Even with optimizing the entire route for ingame, it wouldn't be enough without a major new speed-trick. A pity that it is so close, but on the other hand you can fully concentrate to optimize the run for realtime (other than the escapes) even if it requires larger ingame sacrifices.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
hero of the day wrote:
I know this sounds crazy, but the difference between a 24 end game time and a 23 end game time is much much closer than we all could have imagined. Even with my specially designed realtime route, I think the 23 is possible. I am going to be submitting my run soon, so we will know then how much more time needs to be saved. It would be a great day when we see a real time oriented run, achieve the best in-game timer also.
This is indeed crazy. I really have a doubt this will ever be possible though, as you would basicly need to save ~30 more seconds over the currently known standard. But if the possibility indeed exists to get a 0:23 with that route, then I'm all for optimizing everything for ingame time as much as needed (without pause screen abuse). A 0:23 with the realtime route would just be the ultimate!
Btw hero, do you already have some rough estimates how close you can get to 0:23 judging by your current WIP?
Tub wrote:
just like they had to closely study your escape demo to barely match your time. right.
Keep in mind that Ceres was only about optimizing subpixels at doors. Taco had his advantages here since he (and probably everybody else) monitors the subpixel values, where I did it blindly all the time. Zebes on the other hand requires alot of strange techs to get perfect times, and without the help of my smv, it's a safe bet that most of them would be missed, which would lose them far more than the amount saved in Ceres. Even I missed a few frames there as you see, which just proves the Zebes difficulty again.
Cpadolf wrote:
Yes, I guess the reason that it was overlooked by so many is that you can't gain the distance needed to save a frame here by better speedmanagement (A rough guess is that with the best possible you would be maybe 2-3 pixles away) but you can shoot the door earlier with better speedmanagement as the needed distance to gain is obviously smaller.
Exactly. It's all about the shot, not Samus distance. And only because you have to skip a few pumps when the door opens anyway, you can regain the missing distance to finally save the full frame when entering it.
Cpadolf wrote:
Well I found out that you can actually delay the last shinespark by a good deal of time (I have not checked but I think I do it by 1-2 seconds) and still get the drops at roughly the same time (my guess is that he dies faster if he is closer to the middle of the room) , I use it to spam healthdrops from him which allows me to do an extra shinespark later in the run and in a realtime run it would mean that you are able to skip the drops from him completely, bombjumping up to the door skipping two pause screens and still have enough health for the shinespark in the big room without making huge delays to get refills earlier.
Yeah, that's exactly what I meant. You get alot of important refills without losing any time at all that way. ;-)
I think the extra shinespark will be the one before entering the grey door that leads to the Plasma area, right? Should save around a half second.
Cpadolf wrote:
I estimate my any% to land somewhere between 23:45 and 23:50 so that would mean (if it is possible) that the realtime route is much much closer to the ingame route timewise than I thought.
Same here. I seriously have some doubts the realtime route can be close enough to the ingame one to warrant a high 0:23 as of now. Let's see what hero will come up with his run to clear this mystery.
Cpadolf wrote:
Also watch Kriole/Taco go insane in the big diagonal room when searching for perfection :P I'll be really disappointed if you are slower than 290 frames in that room.
Indeed. This is the room that is made exactly for TASer like Taco, as it depends entirely on subpixel optimizations, which no other can handle as good as him.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Cpadolf wrote:
The only difference I know that can be done to the route is to remove all pause screens which is possible with good ammo management, and even easier with a thing I discovered for draygon :)
Hmmm, I wonder if the "thing" you mentioned is the same that I found when testing Draygon just before cancelling my any% run. :-)
But what about Botwoon? You would still need the Wave Beam to kill him not only faster, but also save 4 important Supers you would otherwise have to slowdown to get before the fight.
Cpadolf wrote:
I was fooling around after kraid to figure out where the drops was placed ofscreen when I to my big surprise managed to somehow get to the top level of the room. I guess it is just extreme luck (or whatever you should call it) that allowed me to find a week point in the roof or something.
Say what?
Nice. You can actually see in SMILE that the roof that cuts the room in half during the fight only exists till the end of the visible screen. It isn't even there in the right half of the room, so you can just spacejump it up there. :-)
Kriole wrote:
Wow, very nice job guys! Although you had alot of benefits from my demo, and very likely wouldn't match my Zebes times without studying it carefully, I'm highly impressed of your optimization level. As we can see now, teamwork indeed pays off sometimes. :-)
Regarding the Ceres improvement, a very clever tech you used (don't know who found it, so I'll be neutral here) at the elevator to land slightly more in the middle of it and thus cut the running distance until the trigger to gain 1 more frame. I never realized this could work out, so much respect for finding it guys! I wonder, can this be of benefit in other places, or does it only work on elevators? When jumping up the 1st stairs in Ceres Escape for example, wouldn't it be possible to land more to the right and therefore jump on the last platform with the door 1 frame earlier?
And to the main smv, you are exactly 11 realtime frames faster than my demo after OTES. For those who wonder about the improvements on Zebes, here a list with technical explanation and a few comments about them:
1 frame gained in the very first room by yet again delaying the dash button to not skip a important arm pump. I never thought it's enough to save a full frame with this tech, so I didn't even bother to try that. Thanks to the lazyness, I guess.
1 frame gained in the mainstreet room by using the spinjump instead of the normal one. This allows to shoot the bottom door a bit earlier even without the falling corner boost, since a spinjump has a slightly faster horizontal movement. A pretty obvious tech to use here, that's why I'm a bit pissed to have missed it...
2 frames gained by managing to unspin and still land on the platform right before Morph Ball. It's only possible by manipulating the falling move as shown there. Very weird but clever tech, which (I'll be straightly honest here) I probably wouldn't be able to find myself. The most useful tech in the smv IMO. Respect.
2 frames gained by using a walljump before shooting the shot blocks after getting the Missiles. Once again a pretty obvious tech, since after a walljump you have much greater jumping speed at the point of shooting the blocks, which accelerates the shot projectile in the upwards direction more than in the horizontal one, and thus allows you to shot it 2 frames earlier and still hit the top blocks that way. The most frustrating thing is that I abused this tech in other places in my previous runs, but just overlooked it here, and this sucks... :-(
1 frame saved by delaying the elevator jump on the way back to Crateria, which for some reason allows to land one frame earlier despite jumping 1 frame later. Strange tech I somehow managed to overlook as well.
1 frame saved in OTES. Once again a really stupid one, since it just required to slightly delay the key-walljump before the 3rd platform, to land on it one frame earlier. A really unnecessary overlook of me.
Oh, and for the note, I actually did a test in Ceres to aim exclusively for realtime including the escape, and I recently got a 49'06, which was even 4 realtime frames faster than Taco's and Kriole's new smv. But despite doing a realtime-oriented run as well, I decided to do a exception in the escape, and sacrificed the few realtime frames to boost the clock up as much as possible. I think it's worth to do so even in such a run, because the end result you will always see in the timer later, just looks much more optimized that way.
And regarding the HJ-Boots, it may save a few ingame frames in the end, and definitely makes the run more entertaining to watch, but is a few seconds slower in realtime, and thus not worth the take in any way if you aim for a realtime-oriented run.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Varies between ~2 days and ~2 years for me.
Depending on game, movie length, motivation, optimization level, and most important: available freetime to TAS.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
I tested both variants there, and with collecting the unavoidable PB refills you will save ~20 frames with the shinespark method, which is pretty much. Thanks to the shutter walk at the beginning of the charge that allows you to start it immediately while landing on the last platform, and still getting the refills in time. Even without the refills, I can't see the shinespark method being more than 1, at most 2, frames slower compared to the walljunmp strategy, especially with the 2 frame improvement Kriole found at the morphing part anyway.
Speaking of RBO, how is your progress Kriole? I'm really interested how the new route turns out and how well you do with the new, harder, conditions at LN.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Well, the frame by frame thing was a bit exaggerated, and it doesn't make a difference of course, I just wanted to note that this stuff was not as easy to find as it seems to be, and that there is no need to hide the fact when a TASer copies existing work. You showed to be independent of others to make top quality work in later parts of the game though, so I believe you in this case.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Cpadolf wrote:
I also would not call my pre torizo work a copy of Saturn's, I did most things by myself though naturally using his strategies which I could not find improvements to.
Cpadolf, you know exactly that this statement isn't entirely true. Many of the strategies in my new demo required deep analyzing and some weird techs that are very unlikely to be found by a casual TASer. Best example is the "in-door mockball" after the 1st Missile pack, or the completely different OMBR strategy on the way back (not to mention the revolutionary OTES). You very likely (to 99%) wouldn't get this ideas without seeing my smv, and you know that best. I realize that for most people it's just not easy to admit when they copy existing work frame by frame, but as said before, there is nothing bad about it because it's naturally for a TAS to aim for fastest time, as Mlandry already said. The only bad thing is when one can't stand to his actions when he actually did it.
Also nice vid you did creaothceann. Looks very clean. 60 fps effect on YouTube ftw. ;-)
Did you use any additional filters (maybe new deinterlacers or other stuff to smooth the picture even more) besides the "Weave" feature to merge 2 frames to 1? I wonder because your video seems to look a bit different on YouTube than mine.
Fano wrote:
when did you start the any% Saturn?
Mid August 2006, 2 days before releasing the 3rd RBO teaser. Worked on it until beginning of February 2007 where I cancelled it at Draygon.
OMG! This is the most crazy shit I have ever seen around Tetris. Especially the invisible mode where he had to remember every single blocks position combined with the inhuman falling speed of them. It's really hard to believe a human is capable to do stuff like this. Just shocking...
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Comicalflop wrote:
Now, if he done all that on his own, you might be having a huff since it wouldn't have been you that created the "currently perfect" strategies for that room.
Absolutely not true. Those who know me can tell that I always appreciate any improvements that are unknown to me, no matter of the author.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
He not only used his Ceres, he used my entire Zebes until pre Torizo as well. But I don't see what is so bad about that. If the fastest strategies are already known, why intentionally use the slower ones especially in a serious and high qualitative run that Cpadolf currently works on?
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Hey Zombie, nice to see you around here. Your Terranigma speedrun on SDA is great btw. :-)
Regarding your problem, I actually use a Super Smartjoy for more than 3 years now, and it works great without any lag at all. There is no better solution to play SNES games with a real SNES pad on your computer. I'm sure the lag problem is caused by your computer or it's system, not the adapter itself.
If you can't fix that somehow, I too strongly recommend alternatives from Logitech. Their products are known to be of high quality and have a long duration time to use, so it's definitely worth the few more bucks to pay for it.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Very nice WIP, Cpadolf. I'm glad that some of my first time shown strategies from the mixed speed demos video were of use for you.
Also, it seems you finally realized the key for speeding the run up in the end result (which is to get many important refills even on cost of a little slowdown). Good job. This version of yours now looks very similar to my any% run overall.
hero of the day wrote:
45700 - The diagonal shot in the shootable blocks was really clever! Do you know how much faster that was?
It's a nice little trick I used in my any% as well. It only requires you to manipulate the shot collision detection to make it hit right in the middle of the 3 blocks to break them all in one turn. Respect to Cpadolf to find it on his own as well. :-)
hero of the day wrote:
51850 - This was really clever too. I missed this one. How much faster was that mockball?
Thats surprising, since you basicly used the same mockball tech when entering WS as well. This one is of more use though, as it saves around 5 frames (compared to 2 for the WS entrance one). Once again good job to find it on your own, Cpadolf.
hero of the day wrote:
53700 - I would love it if we could figure out a way to descend down this room without having to unmorph.
It's impossible no matter what you do. Samus just gains too much falling speed at that point and will always hit the bottom platform if you don't use the unmorph tech.
hero of the day wrote:
54650 - Was the bomb tech faster than the unmorph tech? I never bothered to test with my 1.51 run, I just went with the bomb tech because it was different :P
That doesn't sound very convincing, hero. :-)
But yes, the bomb tech is a bit faster. I tested and compared it in my any% very carefully.
And regarding the strange turn arounds at Speed Booster and Wave Beam, as Cpadolf said, there is no other choice since you can't shoot the sphere from the ground without having wide beams like Spazer or Wave. This turn around method is the fastest possible in this conditions, I tested it very carefully when doing the any%.
The only trick I missed in my run was the upward CWJ in the green bubble room, which Kriole showed in his last RBO WIP as well. I'm not sure if it's faster than my RBO strategy with HJ-Boots (since you start walljumping from the right wall in a lower position that way), but it's definitely somewhat faster to do if not having HJ-Boots, because you have to land on the bottom platform below the wall anyway.
All in all a very nice run so far, keep it up!
Wow, that was new to me. Who would have thought that a Missile has a slightly different moving line than a nowmal shot, to still hit the bottom block directly where the normal shot fails. I didn't even try that in my run, because I have this habit to play safe and not waste any ammo when it's not absolutely required. Not a good thing as it seems... Good job, hero.
Anyway, this tricks saves exactly 13 frames, and thus makes picking the additional Missile pack a easy choice in a ingame run.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
Till Draygon. I cancelled it since then due to the missed Red Brinstar Shaft shot (aka hero shot) we could see when hero submitted his currently published any% here. That run is still frame perfect in at least 90% of the rooms I passed, but that was not good enough for me to continue, even though I could still have a chance to get a extremely close 0:23 in it (even without the Torizo skip). Now Cpadolf took it over, and he will easily get 0:23 in his 2nd version with all the new tricks that got public anyway.
Experienced Forum User, Published Author, Player
(209)
Joined: 2/18/2005
Posts: 1451
You can't compare SMW flying with a critical situation (like a Ceres door) where every single frame needs to be duplicated in order to get the needed result. "Recreate" is a fitting term as well, I guess.
Cpadolf wrote:
Between my two runs this new strategy was 8 frames faster.
Thanks. It's actually much more than it looks to be for some reason. Strange effect.
Cpadolf wrote:
Wait I just noted something in Saturns new viedo, he took 4 missile tanks before WS in his old any%... Why?
It was a route mistake. At that time I didn't thought a 3 pack 1-round on Phantoon would be possible without serious refill delays after the first shot, and I was a bit lazy to test that at first, so I just took the safe way. It's obviously not worth to take one more pack there. :-)