If you look closely to my input, you'll notice I actually pressed down+right (actually only down button is necessary) immediately after he turns around.
Yeah the alternative method is easier to perform.
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
hah, that's funny. So if I had to guess, I would say we are using the same strategy -- using crouch to get further into the wall. But with your timing, you were able to cancel the crouch immediately into a standing jump. I'll try this next.
First, I used my slower crouch version to make this new video. Hopefully it will be obsoleted very soon!
Link to video
3,655 console frames = 60.816508 seconds. 413 time units = 0:59 game clock.
I think 412 time units will change the clock display to 0:58, so if I can get the better wall clip, I may jump straight to 0:57 on the game clock. We'll see.
ok, I figured it out. It's actually very easy with slightly different timing.
You can push down early, while the Prince is still in the middle of turning back around to face the wall again. Once the motion is completed, the Prince is ready to jump through.
So you tap right to approach the wall, tap left to turn away, tap right to turn back. tap down during the turn. then hold forward + jump to get the faster standing jump. very easy after a few tries.
I don't have time to record a new video right now, but I'll try tonight after work!!
In your case, I might suggest experimenting with this idea. Perhaps if you change your "down" input to an earlier time, and then do an earlier "forward" hold, you might be able to start the jump 1 animation sooner on the TAS. Let me know if you see a difference.
The game does have some strange and interesting buffer commands. This seems similar to the buffer where you push "down" while the Prince is in the middle of a standing jump. After the Prince lands, and completes the motion, he will suddenly crouch, even though you pushed "down" a long time ago.
In this case, there is buffered crouch, but it is cancelled by the standing jump. So there must be a transition period that brings the sprite forward very slightly.
Here's a quick shot of the 0:58 with the better clip. This is a 407, which is 6 time units better than the video above. Looks like it may be pretty tough to shave a couple more time units off, to get down to 0:57. We'll see if it's possible once I start grinding.
Yaa Hoooo!!!!!
Fantastic recording. Perfect wall clip. Better beginning. Perfect middle. Improved Boss. Perfect Perfect Perfect!
Note: my wall clip room was the exact same number of frames as the new TAS. 270 frames, 4.4926 seconds. :D
I think I may have been +2 frames on the Boss Music un-pause, but there's really no margin for error, being that close to the ground after a deep jump, so that might actually be perfect also. Very close.
Got time units down to 402, which is about 56.75294 game clock seconds.
Console frame count is 3,559 frames, equal to 59.219138768253968253968 seconds real time.
Man, I'm proud of this one. This is a 462 frame improvement over the old route, which will bring savings of over 7.5 seconds to the "All 20 Levels." I've got some homework to do. There's been a ton of improvements since I made my "compilation video" several months ago, so I've got a lot of new numbers to add up.
Edit: this brings the "All 20 Levels" projection for GT1 all the way down to 22:33.50425 = 81,344 console frames. I'll be taking another look at the GT1 setup for that route next, to see if any of the new strategies you found for GT1 can be applied to that route. (In earlier posts, I mentioned that it appears they work, but with the need for a button reset. I'll go back through, and see if that Zero^2 setup for instant GT1 on 2nd pass can be done here.)
Gate Thief #1 -- Level 16 -- 0:57 -- 402 -- 3,559.
Link to video
3,559 console frames = 59.219138768 seconds.
I made some excellent progress with GT1 setup for "All 20 Levels" category.
We can use the alternate GT1 setup you tested previously, with Single-UP at hole #2, which warps to Level Zero, then falling, and heading left 2 rooms towards hole #4 (where you found the shorter warp to Level 4).
As we mentioned previously, this route causes small button scramble, where B takes function of Jump + Kill + Crouch, and essentially requires Button Reset for humans.
Furthermore, extra lag is caused in Level 2. I counted about 52 extra lag frames with optimal play and Guard Lag Minimization, as seen in my previous videos.
Altogether, the net savings is about 3.114 seconds with absolute perfect play in the Level Zero segment thru Hole #4, and fast Button Reset in the final Level 1 segment where we jump into Hole #1 to acquire GT1.
This could bring the Projection down to 22:30.3927, but we'd still need to "find" 24 more frames to save in order to break 22:30.
I remember that when you did a GT1 Warp-4+16 tas test, your Level 18 was 40 frames faster, and since then you found 4 more frames saved with the alternate step and jump on the lower level to approach the ledge, after pushing the End Button. So the route definitely has the capacity to reach below 22:30 now.
I'm sure the TAS treatment would find several animation saves on each level, so it seems possible that another 5+ seconds could be saved throughout all 20 levels.
I don't think the other GT1 setup will be possible here. If Level 1 is beaten first, rather than doing a Guard Exit, the Zero^2 route goes to the room with the 2nd Guard, rather than room #2 (where we can do the Warp to Level 16). I don't think alternate movements can fix this.
We can disregard the idea of dying on Level 19. Despite the fact that the spikes are 1 unit closer, too much time is lost by "turning" while the front door closes, and more real time is lost by the room transition. Altogether, it's about 24 frames slower to die on 19. So the current strategy of dying at the beginning of level 17 remains best.
So we're down to one last improvement, that I know of. You can pause deeper inside the Boss Room on Level 16 to save somewhere between 2-4 animations. Of course, this difference will probably remove the "extra fast" princess scene before Level 20, but that should only be a 3 frame loss. I expect the better Boss on 16 to save 8+ frames, net.
Because of all the exit practice I did for the GT1 setup (exiting Level 2, and returning to Level 1), I've got a pretty good beat on the Level 17 reset. After a few attempts, I'm pretty consistently restarting the level only 18 frames slower than TAS, real time. I had to alter the technique a little bit. My preferred technique had been to use left thumb for Select + Down + Down. But in this instance, it's dangerous, because releasing Run too soon can cause the Prince to stop inside the spikes, and avoid death. Anyway, I'm finding it a bit more consistent to get top speed with Right thumb on Select + A, while left thumb does Down + Down only.
Well, I should have worked weeks before: http://tasvideos.org/userfiles/info/74378685126672182
Improvements: at the second room of level 16, it's faster climbing from the other side. I also improved two standing jumps (third and fourth room, respectively)
Well, I tried to improve the boss room and no success.
The "extra fast" princess scene before level 20 isn't present this time.
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
I'm not having any trouble completing runs. All the major tricks are pretty easy. I'm consistently getting a sub-7:10. The problem is, I keep getting stupid bumps on the easy gates! A little more practice, and I'll start getting some serious sub-7 runs recorded, and then post up a new record on speedrun dot com.
-------
As you could see from my new GT1 recording on Level 16, the other side is faster, but I guess it depended on how fast you could get there. I hadn't tested it out with the Guard Sprite. Good catch.
The "extra fast" princess is essentially hitting the lottery. The game gives and takes lag frames as the system rounds the timing for the output. It seems to happen every couple of rooms, but you probably have to have a lag save at the exact moment the Princess Scene is starting. So quite random, and very unlikely; but definitely dependent upon what has happened before then.
-------
if your frame count data for the new tas has the same setup as the previous tas, this looks like a 42 frame improvement, bringing the RTA time down to 6:50.02473, just 2 frames shy of a sub-6:50 !!
-------
How did you "improve" the two standing jumps? Did you previously forget to hold forward? lol.
There's one more big question that's never been fully resolved. Once the Warp Glitch has activated Kill Button, is it ok to use Debug Activation Code (Pause, B, Y, Up, Down, Left, Right, L, R, Unpause) to regain access to Sound Menu. Since Kill Button is already activated, Debug Activation does not "illegally" activate it. And as long as Debug Health and Debug Level Jump are not used, nothing unfair is really happening.
Prior to doing any Warp Glitches, Start + Select is a legitimate in-game command to access Sound Menu. Sound Menu has been used in other versions of the game to get things to behave differently.
I have decided to accessing Sound Menu during the GT1 20 Levels Speedrun is allowed prior to any Warp Glitches, to finish Level 1 faster (End of Level Music replaced by short jingle).
If we could access Sound Menu for Level 19, we could save 7 seconds by turning Jaffar's Intro into the short Jingle. As well as save a second at the end of levels 16, 17, 18, and 19. So that should be an easy 10+ seconds off, if we could just do the debug code to restore Start + Select functionality, and access sound menu!
I don't know how to exactly say about Sound Menu, but I'm fine anyway. Most because I aimed for in-game time too, pausing the game and lasting longer than my 2016 TAS (earning 00:29 from level 17 was close to impossible, for instance)
Akuma wrote:
How did you "improve" the two standing jumps? Did you previously forget to hold forward? lol.
Yeah I forgot it.
Akuma wrote:
I'm not having any trouble completing runs. All the major tricks are pretty easy. I'm consistently getting a sub-7:10. The problem is, I keep getting stupid bumps on the easy gates! A little more practice, and I'll start getting some serious sub-7 runs recorded, and then post up a new record on speedrun dot com.
This guard sprite is really a problem.
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
You may have seen my Game Genie record, where I used "No Gates" code, instead of GT sprites. I might do something like that again, for fun, after I get a good Legitimate record; in this case, the Guard Sprite can jump through gates. I think with "No Gates" code, this new route would be almost identical to GT1 on 16-20, but with a few improvements, such as where FGT was better. Such a run would probably be around 6:20.3536 + warp lag, so maybe 6:24.
----
I hope you can find a few more frames to save. Your latest WIP is only 2 frames away from sub-6:50 RTA. (I think the improved pausing strategy during Level 16 Boss should save 8 frames or more.)
----
I think this Guard Sprite is a blessing in disguise. I don't think it will be possible to find a "fast 16" warp with a GT sprite. Even though Guard Sprite is 25 seconds slower, that's better than having something like a normal Warp-6 Sprite, which cannot skip gates at all.
Plus, this adds another layer of difficulty to the speedrun category, so it will not be "too easy" for people to fly through. They will need to take time to master these tricks, and the timing of the game, before they can prove their worth on this route! :)
Maybe this will help you to fix the Level 16 Boss Fight. You can save 12+ frames by following these input instructions. I have done it by hand, and it definitely works.
There must be a period of No Input after the Select Button is finally released, and before the Forward Button is pressed. Otherwise, the Forward Input will take the place of the Select Input, and the game will continue to remain paused.
I know I said I probably wasn't going to submit a new World Record speedrun until I got below 7 minutes, but I changed my mind. This one is close enough, and I might as well start documenting the new route! Sub-7 coming soon, I'm sure.
This one clocks in at 25,304 console frames = 07:01.0399234031746. It's only 63 frames away from sub-7. The aggravating thing is, I wasted 1.6 seconds on the Level 17 Reset, because I lost track of my select menu cursor, while trying to Exit. Ugh! That right there cost me the first sub-7. lol.
Level 1 setup for warp was mediocre, about 1.3 seconds slower than my average best.
Level 16 was decent, only 1.12 seconds slower than TAS.
Level 17 Reset wasted 1.6 seconds on menu navigation.
Level 17 was only 32 frames slower than TAS, because I did triple crouches on the 2nd & 3rd logs, to be safe.
Level 18 was 4.4 seconds slower than TAS. This wasn't too bad overall, but the big bump off the first gate hurt, and the rest of the level was played a little bit conservatively, to make sure I didn't have any more major mistakes.
Level 19 was pretty good, only 1 second slower than TAS.
I made a really stupid mistake on Level 20. I hit the kill button about 20 frames late, and gave Jaffar time to initiate a magic projectile, which delayed the Arena Exit. So I was about 37 frames off target for Level 20. :(
But if I can clean these mistakes up, I should already be very close to a sub-6:55!
Link to video
It just occurred to me that maybe you've already optimized the Level 16 Boss fight by hitting Select (2nd time, to close Select Menu) "too soon" (several frames before Prince running input is allowed), and let the Prince continue his mid-air jumping animations "during the music," and perfectly timing it so that input is enabled just as the Prince is landing.
In this situation, perhaps it's the exact same thing, doing the "free motion" at the beginning of the music (my method), or at the end of the music (alternate method).
If that's the case, and you already matched my projected perfect boss time, then that's somewhat disappointing.
The only other thing I can think of is that perhaps if you do an abbreviated version of my "Select Hold," you may be able to save 2 frames, by getting rid of any lag associated with Window Closure (the window will already be closed). I feel like sometimes the game adds extra frames during Window Closure.
If not, we're gonna have to find 1 more improvement to break 6:50!
I have noticed that, depending on jump patterns, or other things, sometimes you can get Select Menu to pop up 2 frames early during "End of Level" music on Levels 16 & 17. This alternate timing might indicate a slightly earlier start of the music. So it may be worth looking at alternate jump lengths at end of 16 & 17, to see if one can save 2 frames.
Sorry again for my late answers. I've been busy working with some other games, and I forgot this game again (despite not having new ideas to improve warp glitch :\).
Akuma wrote:
It just occurred to me that maybe you've already optimized the Level 16 Boss fight by hitting Select (2nd time, to close Select Menu) "too soon" (several frames before Prince running input is allowed), and let the Prince continue his mid-air jumping animations "during the music," and perfectly timing it so that input is enabled just as the Prince is landing.
Yeah I exactly did that.
Akuma wrote:
The only other thing I can think of is that perhaps if you do an abbreviated version of my "Select Hold," you may be able to save 2 frames, by getting rid of any lag associated with Window Closure (the window will already be closed). I feel like sometimes the game adds extra frames during Window Closure.
I noticed it too, and I unpaused at the right timing to gain 1~3 frames.
Oh! Before going to Boss room, you have two jump choices:
While my TAS is from left, your recent WR video jumps a bit later (right).
I tested the "other jumping", but I only gained a frame, and lost it before restarting Level 17 :(
Akuma wrote:
I have noticed that, depending on jump patterns, or other things, sometimes you can get Select Menu to pop up 2 frames early during "End of Level" music on Levels 16 & 17. This alternate timing might indicate a slightly earlier start of the music. So it may be worth looking at alternate jump lengths at end of 16 & 17, to see if one can save 2 frames.
During my redos, sometimes a frame is lost at the end of Level 18 or 19.
As far I can tell, I already optimized those jumps.
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
I'm sure you've already checked the best spacing for jumps in the level 16 Boss room. Really, the key is to get the correct distance for stopping far away from the wall, but close enough to climb the wall. In that case, there should not be any difference in timing with an earlier or later jump after the gate.
However, we have to remember that some jumps just save time, while others hurt time. So we just have to find the best combination of 2 jumps from the gate to the wall, going across the Boss room.
I guess this must be the end. I really wish we could find 2 more frames, so the TAS's RTA equivalent is below 6:50, but I can't think of anything else. I think you've already incorporated every good idea.
The only other thing to do is go back and check each segment of each level, and make sure the best combination of jumps has been selected, and every standing jump is holding forward, etc.
Maybe check the long hallway before the level 19 jaffar battle, to see if any alternate jump patterns save 1-2 frames. I know you did a good job finding the best jump pattern for the level 20 hallway.
Did you ever notice it makes the sound of wall collision when the Prince is leaving the hallway to enter the boss room in level 19? I wonder if there is some interaction with the walls (curtains) in the boss room, from the hallway. Maybe there is a particular sprite position at the end of the hallway that is more conducive for best boss entry.
Besides this, we can save time with the sound trick, using debug code to enable start+select functionality, access sound menu, and use first sound for faster end-of-level transitions, and save 7 seconds on level 19 jaffar intro.
Challenger, I sent you a private message regarding an easy way to capture the final TAS at 1080p 60fps, for best preservation on youtube.
After I sent you the sample, I went ahead and captured a full video for my own reference. I won't upload it unless you want me to. I'll leave that up to you, until you think it's not possible to save any more time, and whether you want to make your own 1080p60 video for upload.
However, a little bit of bad news has arisen. Now that I do have this excellent 1080p60 video, which is actually captured and played back at 60.098476 fps (very close to SNES 60.09881389744 fps), I have confirmed that the RTA count must be considered 24,643 console frames, which is 1 frame slower than I had calculated previously. Thus, we are actually 3 frames away from RTA sub-6:50, as we must reach 24,640 to break the 6:50 barrier.
24,643 = 410.0413702 349206 349206 seconds (current TAS).
24,641 = 410.0080917 079365 079365 seconds (lowest 6:50).
24,640 = 409.9914524 444444 444444 seconds (highest 6:49).
I do have a small suggestion for cleaning up the TAS time (as opposed to the RTA time). For the "official" TAS time, the clock stops whenever you make the final input.
I would argue that it is not important to minimize the Level 20 time on this Route, because the Best Times board is already busted. This bad Guard Sprite always puts 0:14 for Level 19, making it impossible for the player's time to appear, since it cannot overwrite the very low garbage time.
For this particular Route, I think we would be justified in ignoring the Level 20 clock, and only use the Select Menu to confirm the RTA equivalent. As soon as the Select Menu has served its purpose (when the Prince crosses the finish line), immediately close the Select Menu 2 frames later, end the TAS there, and allow the "End of Level" music to play without further use of the Select Menu trick.
The end result will probably be 6 or 7 seconds appearing on the Best Times board for Level 20, but this shouldn't matter for this particular route. The only real interest we have in the Best Times board for this route is seeing that Level 17 was cleared in 29 seconds.
Anyway, this would cut several seconds off the official TAS time, which I believe is a justified concession.
Saw your private message. Your sample encode is good.
Well, actually I'm not done with this game yet, because I still want to test this game. I put this TAS on hold because I was very busy with the sequel game since September. Now this game was finally finished recently.
About the TAS time: actually I decided to aim for in-game time instead of real time, which is the reason why on Level 20 I didn't close the select menu as soon as possible, and it achieves 00:00 at the end.
And the very low garbage time of Level 19... at least we know the true clock time when beating this level, isn't? (01:14)
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
And the very low garbage time of Level 19... at least we know the true clock time when beating this level, isn't? (01:14)
Yes. We can decode the passwords to get a more accurate read on the time. Although the Best Times board is not helpful in this case, at least all of the necessary evidence is present in the video.
at end of Level 18, we see the password, 39JG!V9 = 5509 in-game time units.
at end of Level 19, we see the password, C2JYFVG = 6036 in-game time units.
The difference is 527 time units / 7.08333 = 74.400000 game seconds (1:14.40), and the displayed time is rounded Down to 1:14. The game clock is somewhat inconsistent in its rounding. I have seen instances where x:xx.3 rounded up.
One thing I've never been able to figure out is whether the clock actually tracks half steps. Each time unit is 2 animations, so it's possible to have nearly identical runs that get the same password, even though one is 1 animation worse than the other. I'd be curious to know if the next level's time starts at +0 time units, or +0.5 time units, depending on how quickly the previous level was beaten. My assumption is that every new level starts at +0 time units, but I'm not able to test this.
This is funny, because theoretically, you could have a 20-level TAS that is in the slower half of the time unit on every level, and then improve the TAS by 19 animations on the first 19 levels, and never get any better passwords, even though the TAS is 76 frames better.
----------
--
Challenger wrote:
About the TAS time: actually I decided to aim for in-game time instead of real time, which is the reason why on Level 20 I didn't close the select menu as soon as possible, and it achieves 00:00 at the end.
It may be possible to close the Select Menu a few frames sooner, end the TAS time sooner, while still getting the 00:00.
----------
--
Challenger wrote:
Well, actually I'm not done with this game yet, because I still want to test this game.
I wanted to remind you. You may see in some of my world record speedruns that I like to play with the Kill Button during the recap scenes in the outro/credits. When I made the 1080p60 video of your TAS, after the TAS ends, the emulator reverts to player control, so in the same video, I pushed the Kill Button during the Level 15 scene, to kill the Ghost, and send the Prince through the gate into the hallway. I think it would be fun if you did something like this in the official TAS video. Of course you can also use player controls to view the Best Times Board from Main Menu at the end of the TAS video.
I just uploaded a new Warpless Compilation video, which includes the wall clip updates from levels 3, 16, and 19.
81,344 console frames = 22:33.5042494984243.
I still have to work on the GT1 setup, as better methods have been found to save time. This may save up to 10 seconds. Once I've put together a better Setup Demo, I'll make another Compilation Video. For the previous Compilation Video, I spent a lot of time making sure each segment was spliced together to create almost perfect level transition time. I haven't gone to all that trouble this time, so a measure of the video from start to finish is not exactly 22:33.5, but it's still pretty close, within a second at each level transition. I'll probably make the next video perfect, once a better setup has been established. All of the individual level segments can be found on my GT1 playlist linked here:
https://www.youtube.com/playlist?list=PL2IdN8AnNyUS4Utl0kWUlqklDLk-rcKxS
And here's the new compilation video:
Link to video
Challenger, any thoughts on why the Biz-Hawk emulator creates different video corruption, compared to console? Here are some examples, my console speedruns on the left, your emulator TAS's on the right:
It's an interesting comparison, because I've actually found other warp glitch patterns on console that recreate these kinds of visuals we see here for the emulator examples. However, it seems that the emulator really loves those solid colorful horizontal lines, and almost always goes to that. (You may recall these colorful lines very early on, when you made a video of Warp to Level 6 Checkpoint.) It's not emulator-exclusive though, because I have found over the years a couple of very obscure warp glitch patterns that recreate these colorful horizontal lines on console. But it does appear that in almost all cases, console and emulator come up with different visuals for any particular warp glitch.
I'm also curious to know if any of the memory watching tools have given you a better idea of what's happening, and maybe using that data to author new glitch patterns?
Which BizHawk version did you use? Which SNES core did you set? If it's an emulation inaccuracy, we can ask the emulator developers to update the BizHawk cores.
However, there is also a chance that this glitch is causing the game to read uninitialized RAM data, and use it for drawing the background graphics. Please try the same glitch again on console, and see if you get exactly the same glitched graphics or not. A glitch that reads from uninitialized RAM would be impossible to emulate accurately, as that would make emulation non-deterministic.
I don't understand about those different video corruptions at all, but seems you can get a different video corruption depending of timing.
When I warped to level 16 for the first time, the video corruption was different from both screenshots:
The warp glitch TAS was made using BizHawk 2.4 version, BSNES core.
--------------------
Speaking of this game, I have another news: I'll work on a update of the published any% run too, since there are some mistakes I didn't fix at the time due to some problems and probably I can improve this game even more with TAStudio tool.
My homepage
--Currently not much motived for TASing as before...--
But I'm still working.
There must be a period of No Input after the Select Button is finally released, and before the Forward Button is pressed. Otherwise, the Forward Input will take the place of the Select Input, and the game will continue to remain paused.
While my TAS is from left, your recent WR video jumps a bit later (right).
I tested the "other jumping", but I only gained a frame, and lost it before restarting Level 17 :(
It's an interesting comparison, because I've actually found other warp glitch patterns on console that recreate these kinds of visuals we see here for the emulator examples. However, it seems that the emulator really loves those solid colorful horizontal lines, and almost always goes to that. (You may recall these colorful lines very early on, when you made a video of Warp to Level 6 Checkpoint.) It's not emulator-exclusive though, because I have found over the years a couple of very obscure warp glitch patterns that recreate these colorful horizontal lines on console. But it does appear that in almost all cases, console and emulator come up with different visuals for any particular warp glitch.
I'm also curious to know if any of the memory watching tools have given you a better idea of what's happening, and maybe using that data to author new glitch patterns?
The warp glitch TAS was made using BizHawk 2.4 version, BSNES core.
--------------------
Speaking of this game, I have another news: I'll work on a update of the published any% run too, since there are some mistakes I didn't fix at the time due to some problems and probably I can improve this game even more with TAStudio tool.
