Very nice, easy yes vote. How does using the European version work with game branches? It's the "same game" but the physics are very different from NTSC.
Well, technically the final fight wasn't completed either (though I don't think it matters; I also don't think the ending softlock matters since the ending sequence clearly started).
Of course it was completed; it just wasn't completed in the normal way. However, the completion (without dying) of that fight is the normal trigger for the ending.
This run wasn't ended by a glitch, it was ended in the normal way of completing the final battle without dying. Other glitches done previously in the run interact negatively with the ending, but it is not a glitch-ending.
If you mean the trick in this submission, there are runs by console runners that are almost exactly the same (they get one energy tank because it's faster for a console runner to do so than to wait for the wavers to get back to the x-ray door. The out of bounds sequence is also slightly different.) The fastest I know of is just under four and a half minutes slower than this submission.
The address you should be looking at is 7ED821. I think you just need to set the second most significant bit, so 0x40 should work. My smv sets the value 0xFF.
I believe the bottom left door requires the least amount of climbing to leave the bounded area, only having to climb to the top of the screen. Any of the other doors would require climbing higher to get to the area outside of the room, and from there you would have to travel to further areas.
I'm not an expert on how the OOB area works exactly, but you seem to need to get to a certain y position before activating the x-ray to mess around with the OOB area. There is another slightly faster method for transition that has been discovered here: http://www.twitch.tv/twocat/c/4136437
It seems that any method will start from the bottom left door though for the reasons described.
It will definitely be faster than the EAC run. Here's a mockup I did based on a "human" route (theory TAS):
http://dehacked.2y.net/microstorage.php/info/1594158150/any%25%20xray%20bs%2020140425.smv
In TAS timing, this run comes out to ~15:33.30. Your run using the old X-Ray method is already a minute ahead leaving X-Ray, and you had a save that I didn't have to use.
I had a small snafu with PBs, if I had saved the PB I used in the X-Ray room (which didn't even save time :( ) I would have had one left for the last room. Oh well.
I look forward to your results.
EDIT: encode:
http://www.youtube.com/watch?v=aqW8tn2SWuQ
Yeah, I'm pretty sure you just need to have 10 when the CF would start. In either case, it's not really relevant since that run is outdated (they now use a damage boost to get stuck into the door far enough to not need a crystal flash.) Also, a TAS can simply get stuck in the door the same way it does in the gold statue room; the same type of slope is what would block ascending in a normal x-ray climb.
Well, humans have once again beaten the X-Ray Glitch category, this time playing by its rules (instead of pulling out GT code shenanigans):
http://www.twitch.tv/ivan/c/4128088
I'm pretty sure a TAS could still do this with only 6% though, it appears to be the same sort of setup as in the statues room.
EDIT: apparently the method in that video is already obsolete and a new method can cut a couple of minutes off of the time. Not that that's relevant for TASing, it's just a different method to get stuck in the door deep enough.
This is legitimately 100% botted, if you watched one completion of it (about 2 and a half hours), you watched all of it. This is because Uniracers has no RNG whatsoever, so everything is deterministic and the game hexes beautifully like no other.
It literally does every stage the exact same way 48 times.
Encoders: I'd really like to embed a Youtube encode of this in this submission text. Anyone want to help out? Preferably a full encode. Perhaps by tomorrow?
Thanks.
In 4k resolution and 60 fps. Lossless. With annotations!
That's just ridiculous! Nobody could expect that! This TAS took 2 and a half hours to make, surely the encode wouldn't take much longer!
Looks like this route might make for a faster x-ray glitch route, and less boring, too: http://www.twitch.tv/twocat/c/3988333
Also it's a 5% completion of the game.
Encoders: I'd really like to embed a Youtube encode of this in this submission text. Anyone want to help out? Preferably a full encode. Perhaps by tomorrow?
Thanks.
"This movie improves on its predecessor by 45.28 seconds, primarily by using a debug code to get 75 Kremkoins early"
http://tasvideos.org/2027M.html
That movie has a star and not only uses the above "debug code" but also uses a code at the beginning while on "Cheat Mode" to reconfigure the layout of the levels. Interestingly, that run had no controversy about acceptance. It was faster than the existing run of the given criteria, and so it was accepted.
Why would this run be any different? The reasons for using the code are the same: the run is more fluid, it has less traversal time to "get to the point".
This run is a beautiful and unique piece of work that opens the doors to a lot of possibilities. While I wasn't a fan of the pause abuse when it was announced that it would be used, I think it works out quite nicely and makes for an intense run with tricks that you'll see nowhere else.
I also think that it should obsolete the any% run, partly for entertainment reasons, and partly because using the GT code shouldn't matter. If you can use a "cheat" to obsolete a branch in one game, why not another? Nevermind you're still circumventing the intended challenge of the game by playing one frame at a time.
Yes vote.
Also in the genre of "twitch streams that don't require a broadcaster to actively be present" I have setup my stream at http://www.twitch.tv/dessyreqt. It is called "AI plays Uniracers" You can give the AI some direction by entering various commands, listed on the stream.
It is also one of the few AI's I know of that plays a game start to finish in real time (and then restarts with the next character)
Was the inability to reproduce Super Mario Bros due to the limitation of the entire system, or was it due to the hardware used to replay it on console? As in, would you be able to produce SMB on emulator?
It would be amazing if you could have written Pokemon Yellow, then started TASing that to take total control there, then do pong and snake.
Just watched the run, streamed it on Twitch a couple of times. The level of optimization is high, and of course Cpadolf lends his wonderful style to many parts of the run, keeping what would be otherwise dull parts entertaining.
Just waiting for the console verification!
Yes you can one-cycle him as Freedan. It takes 291 more frames to get him and do the one-cycle than it does to just use Will and get a two cycle.
That said, the two-cycle strat shown here wasn't immediately obvious. I was running into trouble at first because it seems you have only about 270 frames to kill both hands in the second round if you want to get his eyes on the first vulnerability window; if you miss it, you have to wait an additional 12 seconds. The timing is incredibly tight. I think I only made it by a few frames, and only because I realized you could essentially hit both hands at the same time while they are rising at the start of the second round.
I started work on this a while ago, before Vykan even posted about an improvement. For whatever reason, I put it aside for a long time. Well, I started working on it again, and since I got past the first Mystic Statue, I thought I'd post a WIP!
Here it is: http://dehacked.2y.net/microstorage.php/info/681031108/dessy-gaia.smv
A video of this file can be found here: http://www.youtube.com/watch?v=Dz7wpj7yXQw
Let me know what you think! (Nitrodon, I've heard your criticisms already.)
While the suggestion was bad, I'm happy early Ice was at least looked into. As for my other suggestions, I apologize, since I published that data, I had a second set of eyes (zoast, current 100% console WR holder) look at it and we discovered a couple of errors. They have been fixed and reality seems to skew more in favor of traditional routes (going down green and up blue.)
That said, and I'm sure you were going to do this anyway, but gauntlet after WS seems to be the best time to do it. There's a spazer wraparound that I'm sure you guys already know about that can be done from the top of the gunship room near the PB door. It's just like the plasma wraparound only the positioning is a little higher up the hill. Of course, I'm not sure it's faster than just speedballing over to the gauntlet and then doing a short charge. But I thought I'd throw it out there on the off chance you guys hadn't seen it yet.
I am having trouble understanding where any time would be saved by going down blue brinstar. To me it seems that every room would be a wash with the exception that you would need to fall down the old tourian escape shaft 2 additional times, which would add a ridiculous amount of time to the run. Also, not having the screw attack for blue brinstar is way worse than not having the screw attack for pink brinstar. I could be wrong though, so if you wouldn't mind, can you explain in greater detail where time would be saved?
As for the ice beam, I can only think of 1-2 places where it would save time. Of course as you mention, the spring ball area is the biggest. Again in this case, I cannot see how the cost of doing the ice speed gate room twice is less than the time saved having the ice beam early.
According to my spreadsheet, hitting the gauntlet, going down blue, and then going up green used 15647 ingame frames, while going hitting the gauntlet, going down green, then going up blue used 16549 ingame frames. That's a difference of 902 ingame frames, suggesting an in game savings of around 15 seconds.
"Why" is a good question to ask here and unfortunately I don't have the conclusive answer you are most likely seeking. I would assume it's to do with the way the game naturally flows, as an example you can save a couple of seconds traversing the Dachora room going left instead of going right. However, I specifically did all of this examination so that I WOULDN'T have to think about what conceptually "feels" faster. I'd just have numbers to point to (see the bottom of this post.)
I apply the same reasoning to getting the Ice beam early. However in this case, it's easier to explain. You have a lot of really short rooms with a lot of door transitions, so while you're burning a lot of realtime, you're not burning a lot of gametime. I wouldn't have suggested this route with a realtime goal in mind. For realtime, you either get ice, or you get grapple before WS, but not both. For game time, you get ice for sure, and you maybe get Grapple before WS. Only about a third of the time spent in the ice beam area actually counts toward the game clock. The places where it saves time either substantially reduce ingame time (springball being the prime example) or just allow you to move more quickly (like the red brinstar shaft)
I will also note that I don't agree with going for early power bombs in this run because you will likely be heading up the red shaft anyway, and also you'll likely be getting Crocomire's PBs before WS. You can find my comparison of early PBs vs. no early PBs as B1 and B2 in my spreadsheet.
--
Here's a zip with the spreadsheet plus all of the movies files relevant to it: https://dl.dropbox.com/u/6998359/SM100routes.zip The spreadsheet is Excel 2007 format and the movies are SNES9x 1.51.
Feel free to go through at your leisure and determine if you do or do not agree with my conclusions. An effort was made to have a consistent cutoff point for each room (the script I used cuts at the first frame of a door transition.) I also did my best to eliminate things that were unnecessary to time, for example anything before charge beam. Rooms that were applicable to other branches simply had the times duplicated. Hence a segment labelled "Norfair elevator to Ripper shaft" could happen at several points in the run, and likely in consistent ways. I tried to used common sense when determining whether I needed to record a new segment or not. Room times listed in blue in the spreadsheet are duplicates and simply point to the original cell. Room times listed in black are the originals and should be represented in a movie file somewhere in the zip.
This package is not geared toward TASing, it's geared toward console play, and thus tries to emulate a "perfect" console player (but not to the extremes that a TASer would be.) Rerecords were not spared in these recordings, but they were all recorded at 75% speed.
Let me know if you have any questions about this.
Regarding route, I thought I'd chime in since I've done a lot of testing on console routes. I discovered an error in my spreadsheet a while ago that was giving some bad times, however I've resolved that and things are looking a bit better.
That said, from my tests, the best overall route applicable to this run seems to be go up red Brinstar, do WS, go to the gauntlet, then go down blue Brinstar, do Maridia and lower Norfair, and finally go up green Brinstar. I know it sounds a little awkward to go down blue Brinstar, but the difference in time is pretty minimal. This route has INSANE PB requirements, but in a TAS, I think it can be done.
One other thing I'd recommend doing is getting Ice Beam before leaving Norfair the first time, especially since you're focused on game time. The Springball glitch by itself saves quite a bit of time, more so if you count the extra door transition. Since it is game time-oriented, you don't care about having to change weapons either.
Nothing I can say will do justice to this TAS.
APM means actions per minute? How is APM determined?
I would suppose that a Lua script is used. In that case, I'd like to see the script.
Yes, it's calculated via a Lua script: http://pastebin.com/Pt4i6mS4
In this context, an "action" is a button press, or a frame in which a button is pressed following a frame in which that button is not pressed. Each button press per frame counts, so if you press A and B on the same frame, you get 2 actions for that frame.
Current APM is calculated by adding all the actions that happened in the last 3600 frames. Average APM is total actions/total minutes. Peak APM is the highest current APM encountered so far.
Trihex boasts an APM of 150 while playing this game. It's probably only 100 useful APM, with 50 showboating APM. :P
Similarly, the APM spikes in this run typically happen during the autoscrollers.
