new tas projection --
1a -- _710.
1at - _308.
1b -- 1936.
1bt - _344.
1c -- 2444.
1ct - _136.
1s -- 5878.
4d -- 3276. (thru first frame climbing steps)
4dt - _478.
4ds - 3754.
5a -- _772. (thru select at 4th frame final sprite)
5at - 1437.
5as - 2209.
5b -- _690. (right-facing drop)
5bt - _164.
5bs - _854.
16 -- _766. (improved lag)
16t - _342.
17 -- 1828.
17t - _346.
18 -- 7312.
18t - _346.
19 -- 5008.
19t - _414.
20 -- 1686 (thru first frame select menu on-screen, at finish line)
tot- 30,743
30743 tas = 8:31.5408775365..
Note, all of these can be +/- 1 frame. The game engine timing floats in real time, so there can be arbitrary lag frames added or saved in any room for absolutely no reason. Once the TAS is programmed, these arbitrary lag frames will be set in stone, as the game is very consistent in replaying TAS performances. Sometimes, there's nothing you can do about a lag frame, and then it'll get saved 2 rooms later for no reason. So these numbers for each segment may be different.
However, also note that real time can be bumped either way by 2 frames, depending on Select Menu timing. So always mess around with select menu exit time at the end of each level, to see whether the game will give you a faster exit. (For levels 16-20; not for the glitches.)
I tested my theory on level 16. The double crouch to the spikes does set up a favorable position for jumping deep into the boss room, so you can save 7 animations with this trick. Unfortunately, the double crouch and missing out on an extra jump costs a total of 12 animations, so it's better to just do the one jump over the spikes, and two jumps in the boss room. So I think once you fix the select menu trick, that's the best we can hope for. I'm out of ideas there! :)
----
The level 4 trick works because of the slow-down animation of this sprite corruption. When the "dead prince" animation occurs during the slow-down sequence, his collision detection box is very low, below the gate.
----
I confirmed the lag save for level 16. As mentioned above, the new right-facing drop in level 5b saves about 81 frames, but loses 22 frames on the transition (net 59 savings for 5b); for level 16 however, the lag savings is big in the first room, and also in the boss room. a total of 83 lag frames is saved.
So the new warp-16 drop saves a total of 142 frames = 2.3627754 seconds.
this further reduces the tas projection to 30,747 = 8:31.60743459..
(projection will be 30,743; i corrected 16 with 2 frames lower during level, and 2 frames lower during transition. see below.)
WIP looks great. here's a few notes...
level 1 --
Only thing you forgot on level 1 is the 2nd room in 1b and 1c. You can do the "stop running" at the edge to fall more to the right, then when you jump down, a single crouch, turn, and grab the left edge of the hole. Facing left saves a lot of time in the 3rd room each time.
Level 4 --
awesome first room. How did you do that? Is it a u-turn that slips through, and collides with the outside of the gate? -- if so, that's funny, because I just tried that exact trick on level 19, to see if there was a better way to get through the next gate, after jumping "through the top" of the previous gate. The fat gate thief starts a u-turn while running through the gate, collides with the backside of the gate, and gets bumped forward without actually turning around. It was a good idea, but it didn't pan out. The traditional way of running into the wall was still slightly faster.
At kill button for first guard of level 4, you forgot to hold forward before starting the "standing jump." This caused a slow standing jump, which cost 3 extra animations (12 frames).
Otherwise, Level 4 is excellent. I'm glad to see you found the right timing for the double jump past the gate, after the u-turn. This was my original approach, as seen in the 8:57 world record from a couple of years ago. By waiting the extra 8 frames, the next room is 31 frames better, so that's a net improvement at 23 frames. Unfortunately, back when I did my old world record video, I was overly cautious, and waited an extra 48 frames. lol.
I'll need to correct my tas projection: the better u-turn is 8 frames better than what I've been doing, not 12 frames better. So I will have to add 4 frames to tas projection before accounting for these new improvements.
-------
5a looks perfect.
5b, you got 1 animation better than me in the first two rooms (-4 console frames). Did you add a jump press at the edge in the 2nd room, or did you just run off?
the new right-facing fall on 5b is awesome! It saves about 81 frames over the previous "climb and fall" facing left. The transition time is slower (about 22 extra black frames), but did you notice how fast level 16 starts? Basically no lag! It was 48 frames faster to reach the ground.
Faster drop, slower transition, faster level 16 start, net improvement is about 107 frames. (1.77699 seconds!)
I'm curious to see if this translates into more lag savings through the rest of level 16.
Level 16 --
You'll need to fix the boss fight. You've got +5 frames from first frame of boss room until first frame of movement after unpause. This is because you didn't remove the select menu early.
You can hit select again during boss intro music to make the select menu window go away early, but you'll need to have some other button continuously pressed during the boss intro music, to keep the game frozen after the select menu goes away. I always hold the left arrow on the d-pad. Then, just before first possible frame for input, I release d-pad to officially "unpause", even though the select menu has already been gone. Then immediately repress d-pad left so the prince will not stop running. I also press X & B at the same time for immediate kill and immediate subsequent running jump.
----------------
overall, 53 frames improved on level 4, and 107 improved between 5 & 16. Also 8 frames improved from little things I noticed, this brings the tas projection down by 168 frames = 2.795 seconds!
30,782 = 8:32.1898088. wow!!
Of course, I should be able to do most of this by hand, so now my RTA goal may be going towards 8:35.9! -- this may be an unrealistic dream. I think a low 8:36 is definitely something I can shoot for though. :)
=================
i had another idea for level 16. I'll have to mess with it tomorrow, as it's way past bedtime now! I was thinking maybe after this better immediate jump, I could try a crouch or two, to mess with the spike spacing, and then maybe start a jump at the edge of the room. Depending on the spacing, may be able to save 5-6 animations during boss music, plus better spacing to opposite wall could save a few more animations by avoiding wall collision. It's just in my mind right now. I can't tell whether it's a good idea or if it's possible until I sit down and play it tomorrow.
A few preliminary tas notes for Challenger --
1a -- You'll need to use the exact same movements as from Level 1 in your "33:31" tas. The test has a later jump off the ledge.
1b -- My standing jump in the 6th room knocks the panel down faster than running to it. However, your jump/fall pattern in the 4th & 5th rooms seems to be best there. I noticed a slightly different jump used here for 1c. However, I believe both patterns result in the exact same time. 1c will always be 3 frames slower, because of new lag.
4 -- we've talked about the better pattern in the first room of level 4. Despite being slower in normal circumstances, lag reduction for this route dictates leaving the first room as quickly as possible, to save the most time universally. My movement must be considered only preliminary. 2 steps, stop on button, full run into gate, and standing jump forward. There may be a slightly better way to do it, but it is not practical for me to test it by hand.
I just found a better jump pattern at the 1st guard, this is another 4 frames saved, compared to the World Record video. I'm currently working on a new Blind Level 4 video to document all these notes.
I have found that my way of doing the long drop (take damage at lower gate) is best, as this later jump at the 3rd guard sets up a quicker panel fall. That jump causes equivalent time loss in the previous room (with the 3rd guard), but I think there is a consistent lag frame save there. You may be able to double check that for me.
I've noticed that your U-turn at the crusher is better, saving 6 frames in the room before the crusher, and another 6 frames in the crusher room.
However, it must be observed that you need to stop running in the crusher room, and allow the gate-waiting time to elapse here, rather than in the room with the gate. The gate cannot cause extra lag while it is off-screen in a different room -- even in the Blind level, it is still considered on-screen if you are in the same room.
5a -- I've optimized the movements in 5a. A standing jump to the button, a quick triple jump to clip the ledge, a double-crouch to the wall, and a quick double jump to the gate.
5b -- it appears that I have not yet published the better 5b jump pattern to clip the ledge. I'll do that shortly. For reference though, the World Record video sets up the first jump correctly, while the second jump was executed 1 frame too early. The next later jump provides the good clip for a double-crouch.
16 -- Today, I found a better way of doing level 16, which saves 38 frames, so there's the savings we need to get TAS below 8:35. Basically, you just have to hold run + jump before 16 starts for an immediate running jump. The same tactic used at the beginning of Level 17. The spacing is perfect to fall below the opposite ledge, and have time to start another running jump at the spikes. This saves a full 38 frames over my previous method of simply running off the edge at the beginning.
ok, so I've crunched some numbers, and here's what I've come up with so far as a potential projection for the TAS, with a comparison to the totals for the new World Record.
1' - 5992 wr, 5882 tas, -110.
4' - 3826 wr, 3807 tas, _-19.
5a- 2239 wr, 2209 tas, _-30.
5b- _952 wr, _912 tas, _-40.
16- 1239 wr, 1200 tas, _-39. (new strategy found after WR recording).
17- 2202 wr, 2174 tas, _-28.
18- 7670 wr, 7658 tas, _-12.
19- 5434 wr, 5422 tas, _-12.
20- 1688 wr, 1686 tas, __-2. -- tas [-292 = 4.85866494 seconds better than WR video]
tot- 31242 wr, 30950 tas = 8:34.98520508...
With this, 31k is broken, and 8:35 is broken. (At least by RTA counting.)
All of my personal best segments add up to 31,112 = 8:37.680765765,, which is 19 full frames clear of 8:38.00. I would have to learn the better 4th and 5th rooms in 1b and 1c if I were to have a reasonable shot at 8:37.x.
Here's a closer look at the numbers: ("t" = transition segment; "s" = sub-total.)
1a -- _714 wr, _714 tas. __-0.
1at - _325 wr, _308 tas. _-17.
1b -- 1955 wr, 1936 tas. _-19.
1bt - _378 wr, _344 tas. _-34.
1c -- 2468 wr, 2444 tas. _-24.
1ct - _152 wr, _136 tas. _-16.
1s -- 5992 wr, 5882 tas. -110.
4d -- 3348 wr, 3329 tas. _-19. (thru first frame climbing steps exit)
4dt - _478 wr, _478 tas. __-0.
4ds - 3826 wr, 3807 tas, _-19.
5a -- _773 wr, _772 tas. __-1. (thru select at 4th frame final sprite)
5at - 1466 wr, 1437 tas. _-29.
5as - 2239 wr, 2209 tas, _-30.
5b -- _790 wr, _770 tas. _-20.
5bt - _162 wr, _142 tas. _-20.
5bs - _952 wr, _912 tas, _-40.
16 -- _895 wr, _856 tas, _-39.
16t - _344 wr, _344 tas, __-0.
17 -- 1854 wr, 1828 tas, _-26.
17t - _348 wr, _346 tas, __-2.
18 -- 7323 wr, 7312 tas, _-11.
18t - _347 wr, _346 tas, __-1.
19 -- 5020 wr, 5008 tas, _-12.
19t - _414 wr, _414 tas, __-0.
20 -- 1688 wr, 1686 tas, __-2.
tot- 31242 wr, 30950 tas, -292.
30950 tas = 8:34.98520508...
(notice -- Challenger has found improvements to levels 4 & 5 since this post of written. A new tas projection list to account for the new discoveries can be found below.)
YaaHooooo!!!
Magic happened last night!
I made a small mistake on 5b, while recording a full setup demo. I missed the 2nd jump, which caused a bad clip off the ledge. I had to adjust and do a triple-crouch to reach the next wall, instead of the better double-crouch (as seen in Challenger's TAS for Level 8, after the 2 chompers).
I was about to quit, because I've been trying very hard to have a perfect proof of concept for the improved setup route, incorporating several of the tricks Challenger added in his setup test.
But I knew levels 1, 4, and 5a were very good, so I switched gears, and turned it into a World Record attempt. I played Level 17 conservatively, as usual, losing about 30 frames, because I did triple-crouches at the 2nd and 3rd log, just to make sure I didn't get hit. Before that, I had lost 20 frames on 5b, mostly from that missed ledge clip. (+4 on the ledge clip, +12 for the extra crouch, and +4 mistake in final room to knock loose panel late).
But after that, the run was basically perfect. Only single-animation mistakes here and there through levels 18 and 19 cost another 24 frames, but I saved 24 frames on the transition between 19-20, relative to my projection, because I forgot that the 19-20 transition is faster than some of the other Princess Scenes. It only takes 412 frames here, and in Challenger's 33 minute TAS, as opposed to the 440 frames I've been projecting. So I went into Level 20 with a comfortable buffer below the 8:40 mark, and played Level 20 almost perfectly, only losing 2 frames (less than 1 animation), probably lag.
So the new human World Record is 31,242 console frames = 8:39.84387. This is the perfect way to celebrate Challenger's upcoming TAS! ;)
I'll provide a breakdown of the individual level times soon, with better projections for the new TAS. Also next will be more individual level videos, now that the full setup has been officially documented.
A few fun notes about this video --
at 11:25, I used Kill Button to kill the Ghost during the "Highlights" video after the end. This killed the Prince too, but since he was stuck in his running position, he continued to run through the gate, and down the hallway. This will happen with any Sprite Set, including original sprites. It's not a Gate Thief trick. Here, you can also do something like this by putting in a password for Level 17 with 0 health bottles. The Prince will start the level running, although dead, and continue to run until the Spikes catch him.
at 11:40, I used Kill Button to kill the Level 18 Boss. The timing was just right so the Boss died during his sword fight animation, so the Prince put his sword away, and got killed by the dead Boss. haha! -- Also interesting, this Warp-16 setup gives me Invincibility Mode, which doesn't have any useful purpose for shortcuts in this route. But this video here shows that Invincibility Mode does not continue to be active into the Highlights portion, since the Prince was killed by the dead Boss.
at 12:10, I show off my Console + TV setup, and show all the cables.
at 14:52, you can see me say hello to the camera!
Prince of Persia - SNES - ANY% - 8:39.84387,
Link to video
31,242 console frames.
new sum of best is down to 8:38.995. I'm working on recording an official video for it, and then I will update the projection stats. I have a pretty clear picture of where the TAS will be, so stay tuned for that as well. Once I have my official human & tas projections, the challenge will be issued, to see whether Challenger can find any improvements to lower my tas estimate!
In the 2nd segment of the sub-9 setup, where you drop down and approach the "warp-8" hole, Challenger preferred a double-crouch to knock down the loose panel, whereas I have been doing a 1-step run. I've found that this double-crouch is easy enough to implement into my routine, so I will start doing that, and update the projection accordingly. I'm not sure if this improvement will give me quite enough room to realistically shoot for an official sub-8:40 speedrun, but it may be close!
I think it will be more instructive, statistically, to break up the setup segments the way I account for full levels. This will draw a better comparison between the human limitations, and the TAS projection. For instance, while playing the level, the performances will be much more consistent for comparison, whereas the "transitions," including time to navigate the Select Menus and exit, will fluctuate more, as the human hand is slower than the TAS inputs to rapidly navigate the select menu. The next time I release a video, I will provide times for each segment, each transition, and comparisons to TAS time.
ok, so my absolute best setup (3x level 1) recorded so far (which is the basis for the projection) is about 54 frames slower than TAS on select menu & main menu, and another 38 frames lost between 5a and 5b select menus, and main menu.
That's 1.531 seconds.
Also, Challenger put together some slightly better movements in level 1 to save 54 frames in 1b, and these same rooms produce another 46 frames saved in 1c. So I'm 100 frames behind TAS playing level 1. At some point, I might try my hand at these alternate setups, but for now, I'm going to continue to grind with what I know, and give TAS those 100 frames.
That's another 1.664 seconds.
My recordings and movements for Levels 4, 5a, and 5b are all better than Challenger's test. So assuming no other slight improvements are found, I'll be tied with TAS after the Warp to Level 16, assuming no mistakes.
As I've said many times before, I'm doing all of this by hand with a controller, a phone camera, and a calculator. So I can't say for sure that Levels 16-20 are perfect. I would imagine that once Challenger starts putting together the official TAS, he may find a few animations here or there. Each animation saves 4 console frames. And certain setups can save half steps (2 console frames), or reduce lag (1 frame). So I wouldn't be surprised if the TAS precision could save another half second over the length of the run.
I made a 60 fps reference video from Challenger's test, so I've been able to track all of the tas times for menu navigation, and will be able to make note of these differences when I have a good video to upload. I will start grinding in earnest now, and see if I can get a WR closer to 8:41.
I'm trying very hard to shave 1 more animation off the Blind Level 4 time projection. Otherwise, the projection is pretty much done, in terms of human expectations.
So, if the human projection can come down to perhaps 31,221 = 8:39.494445, we could say the TAS will likely end up around 30,999 = 8:35.800529.
Of course, that would be counting from first frame of level 1 until first frame that select menu is displayed on-screen during the "end of level" music at the end of level 20. As I understand it, TAS rules provide for a slightly different accounting of the time. Nevertheless, I will be striving to help Challenger figure out how to break the "31k" mark by speedrun standards.
Lol, I beat the world record today with a pretty bad run, while messing around. I've been grinding a full "Setup thru start 17," as a connected proof of all the warps in the route. On this particular try, I made a couple of mistakes that made it unusable for a demonstration of best possible time. On a whim, I decided to run the whole route thru end of 20, and came out with an 8:44.9, so already, 8:45 is officially broken. I may not publish it, because of a few stupid mistakes.
Nevertheless, I am on the heels of something massive.
In the meantime, a thought occurred to me, regarding the potential difference in time between me & a TAS. Really the only key difference will be how quickly the TAS navigates the select menu for each "exit" for the warp glitches. My thought is this: At best, my menus are about 21 frames each. Regardless of how much faster the TAS is, there's not a ton of room there. The TAS would be at least a few frames each, which means it cannot be more than 2 seconds faster.
With this in mind, I must start giving the TAS a projection of only 2 seconds or less, better than my human projection. I will analyze Challenger's test video tonight, and get a more exact figure.
My projection will probably stay very close to the current mark of 8:39.628, so perhaps the TAS cannot do better than 8:37.628 (for example), unless some new techniques are discovered.
ugh. This game is so stupid!
I just realized that if you go into the room with the guard at the beginning of the setup -- YES, you save 22 lag frames in Blind Level 4 -- BUT, it also creates a bad Warp-16 with super lag and continuous time remaining box popup.
So, in order to get the preferred Warp-16, and save a billion lag frames, you must pay the 22 lag frames in Level 4.
That's disappointing. But now we know!
I tried several alternatives, and different timing, and different sprite positions and animations, but it's always the same. If you go into the guard room first, it ruins the Warp-16 pattern later. haha. :(
I went back and edited the previous comment to remove inaccurate information. I'll update the rest later.
Currently, because of the new 5a recording, and the upcoming improved setup recording, these extra 22 frames will not hurt too much.
new breakdown --
6022 -- setup
3823 -- blind level 04
2247 -- blind level 05
_936 -- F-GT level 05
_893 -- F-GT level 16
1824 -- F-GT level 17
7312 -- F-GT level 18
5008 -- F-GT level 19
1686 -- F-GT level 20
1478 -- transitions 16-20
31,229 = 8:39.62755959365,,
Late tonight, I finally got a best run for level 5a during speedrun with video corruption, so this will be official for the projection.
It was completed in 2,247 console frames, from first frame 5a to first frame 5b after intro reset. This is an improvement of 19 frames, and brings the projection down to 31,226 console frames = 8:39.5776418.
Earlier today, I thought I had finally got a great recording of Blind Level 4, but I discovered a funny issue that causes extra lag. Now that it has been solved, I'm glad to report that an official Blind Level 4 video should be recorded very soon to confirm the projection.
It's late, so I will post the video for Blind Level 5a tomorrow.
So, here's what I figured out about Blind Level 4 and the lag problem.
At the beginning of the setup in Level 1, your first segment is to "visit" the first guard, then exit the level. If you access the select menu in the hallway, and exit the level on the last frame before the screen transition, the game will load the room with the guard for 1 frame after the exit, and you will "see" the guard. This method causes 22 extra lag frames at the beginning of Level 4.
In order to remove these 22 lag frames, you must do something else that ruins the Warp-16 pattern. Therefore, these extra 22 lag frames are mandatory for the route. Very disappointing!
(The solution for the 22 lag frames would be to go all the way into the room with the guard before exiting. This makes level 4 faster, but ruins level 16, with huge massive lag, and continuous flashing "time remaining" box.)
Therefore, I must alter my projection accordingly, to include these extra 22 lag frames. At the moment, this increases the Level 4 projection up to 3,816 console frames.
After the new 5a recording improved the projection to 31,226, this discovery must increase the projection to 31,248 = 8:39.9437056.
Fortunately, it's still below the 8:40 mark, but just barely. I have been improving my execution of the menus during the setup, and am looking for a time around 6,032. This 11 frame improvement in the setup would bring the overall projection down to 31,237 = 8:39.7606737,
woohoo! got an almost perfect run for the new Level 4 route.
3,416 console frames from first frame level 4 to first frame select menu at end.
3,416 + 346 level transition + 32 lag frames projected for the first room after doing the Warp to Level 4 = 3,794 = 63.1293656888... seconds from Start Level 4 to Start Level 5 real time, during the Sub-9 full speedrun route.
I picked up an extra animation or two during the U-turn under the Crusher. I may try to record this one more time, and see if I can nail the earlier jump, and save 4-8 frames. Everything else was perfect.
For now, this improves the Sub-9 Route Projection to 31,285 console frames = 8:40.5593583492. Only 34 frames away from a sub-8-40!
---
Sub-9 Route - Clean Level 4 - 0:56 - 396 - 3,416.
Link to video
Here's a pretty good full level 4 recording, for the Sub-9 Route. I managed to do everything perfectly up until the end. I accidentally waited too long for the last gate, and wasted some time there, and also a bad End Door climb. I should be able to save another 24-28 frames will perfect play, and favorable lag.
This is 3,439 + 346 + 32 = 3817.
projection update: 31,308 = 8:40.94206.
Now we're just 57 frames away from a sub-8:40. Assuming I can get everything I'm trying to do on Level 4, that just leaves 30 frames to be shaved off for levels 5a & 5b.
---
Sub-9 Route -- Level 4 -- 3,439 console frames
Link to video
399 time units.
I've saved another half second on Blind Level 4 by reducing lag. The key is to get out of the first room as soon as possible, even if it would mean a slower overall route under normal conditions. By using a standing jump to get under the gate as soon as possible, and immediately get into the 2nd room, critical lag time is saved. The traditional method of speedrunning the beginning of Level 4 is to do a u-turn for optimal running/jump speed under the gate. But here, where lag is a major consideration, some time can be lost with less optimal performance, to seriously reduce lag.
I will see if I can splice a video together showing the method used. It's only 4 frames slower (1 animation), but saves about 36-37 lag frames. So that's a full half second improvement that I wasn't expecting!
Edit: Here's the spliced video. This net 33 frame improvement may be the key to breaking 8:40!
MAJOR EDIT -- This video has already been made obsolete, as I just found another improvement. It is actually 8 frames faster to run directly into the gate on the 2nd run, rather than doing a 1-step run without gate collision. This improvement of 8 frames might translate in another 1-2 frames of LAG being avoided as well. I will have to confirm this.
---
Sub-9 Route - Blind Level 4 - Improved LAG in 1st Room.
Link to video
---
For Blind Level 4, I'll be recording the route under normal conditions, without lag or video corruption, to document exactly what the intended route may be. One thing I'll be focusing on is finding the absolute fastest way to get out of the first room of Level 4, to avoid extra lag, since there is no lag in the 2nd room. If I find a slightly better way to do this, I'll also record a warp-4 counterpart to document the improved lag in the first room.
So far, it appears I may be able to record something around 3419 + 346 transition. To this would be added known lag from the first room, to calculate the best possible time from start level 4 to start level 5. It would be something like 3419 + 346 + 73 = 3838 = 63.8614933 seconds.
This would be an improvement of 124 frames, and would bring the overall projection down to 31,329 = 8:41.291485943.
It's hard to say how much savings can be done in 5a and 5b, but it seems doubtful I can save enough to break 8:40. Assuming another 30 frames saved, that'll only get me down to 31,299 = 8:40.7923. At that point, we might project that TAS could get down towards 8:33.300.
WooHoo! Got a great video recorded today.
I set a new record for the Warp 4 + Fat Gate Thief Setup + Warp 16. I also documented a best Level 16 Segment, and documented LAG for Levels 4, 16, and 17.
Level 4 lag is limited to 73 lag frames, and only in the first room, with best setup. For Level 17, lag appears to be limited to 2-3 frames, in the first room only.
This new setup clocks in at 6,043 console frames, from first frame level 1 thru first frame level 4. The TAS will be faster, because of insane fast select menu exit navigation, but as far as level navigation goes, I think this maps out the perfect route for the best warps with minimal lag.
This covers the Setup Segment + the Level 16 segment. Using the above Fat Gate Thief videos for Levels 17-20, and the old World Record video for the Level 4 and Level 5a & 5b segments, the current projection is down to 31,453 console frames = 8:43.3547546 for a human. As I said earlier, I think a TAS could push this down to 8:35, or 8:34, with faster select menu navigation.
Here is an updated accounting of the Sub-9 speedrun projection:
Setup' ---- 6,043 console frames.
Level 4 --- 3,962 console frames (to be improved).
Level 5a -- 2,311 console frames (20+ frames to be improved).
Level 5b -- 936 console frames (to be improved slightly).
Level 16 -- 893 console frames.
Level 17 -- 1,824 console frames.
Level 18 -- 7,312 console frames.
Level 19 -- 5,008 console frames.
Level 20 -- 1,686 console frames.
Levels 16-20 transitions -- 1,478 console frames.
Projected total without further improvements -- 31,453 = 8:43.3547546.
---
Fat Gate Thief Sub-9 Route Setup - 1:40.55107 - 6043.
Link to video
6,043 console frames = 100.55107 seconds.
I'll post some videos tomorrow. I got a great setup segment recorded today, and a good level 5a segment. between these, and the level 4 segment from the World Record video, and the 5b + 16 segment from the World Record video, I have a tentative projection to share. I don't think I'll be able to save much more time, as level 4 was pretty good in the old video.
Setup -- 6,066 console frames = 1:40.93377, (should be 6,048. I noticed a silly mistake, and will need to re-record this segment before I upload a video.)
Level 4 -- 3,962 console frames = 1:05.92476,
Level 5a -- 2,311 console frames = 0:38.45334,
Level 5b + 16 -- 2,182 console frames = 0:36.30687,
Levels 17-20 -- 16,962 console frames = 4:42.23519,
projected total -- 31,483 console frames = 8:43.85393252,
It's nice to see that it has gone below 8:45! Perhaps there is enough wiggle room where I could seriously attempt a sub-8:50 speedrun.
------------------
A reminder on Challenger's test. You can see a link on Page 8 of this thread. May 17th, 2019, he posted a replay video of this route.
http://tasvideos.org/userfiles/info/55284868098526612
For comparison, using the speedrun timing standard, from the beginning of Level 1 until the beginning of Level 16 Checkpoint, roughly 3:36 elapsed. There were several examples of inaccurate play, since it was a quick test, so one might guess another 2-3 seconds might be removed from this for the TAS. We might project a time of 3:33 for TAS during this portion.
My "best segments" so far add up to about 3:40.5867 for the same portion of the run, about +7.5 seconds. The vast majority of this time is saved by the TAS's insane speed in navigating the select menu for fast exits.
I suspect my limit will be around 8:42.500, putting TAS around 8:35.000,
I've decided to also do a quick projection for the "sub 9 minute" route. I just recorded Levels 17, 18, 19, and 20 with the Fat Gate Thief. Next, I'll go back and record a better setup for Warp 4 + Warp 16, and see just how low we can take it. Now that we have good totals for the last 4 levels, it's fairly clear that this route can be done in less than 8:50.
Given the frame counts below, plus 2x346 plus 1x440 for transitions, the last four levels can be done in 4:42.235187. Simply adding the first half of my first World Record speedrun of 8:57.331736 thru the beginning of Level 17, we have 4:08.115 + 4:42.235 = 08:50.3502. That included a 1.5 second mistake on Level 5, and a grossly inferior Level 1 setup. So now, I'm feeling like I can get a "sum of best" down below 8:45. It'll be fun to see just how low it can go.
---
Fat Gate Thief - Level 17 - 0:29 - 209 - 1,824.
Link to video
1,824 console frames = 30.3500166095238 seconds.
---
Fat Gate Thief - Level 18 - 2:01 - 858 - 7,312.
Link to video
7,312 console frames = 121.666294653968253968 seconds.
---
Fat Gate Thief - Level 19 - 1:15 - 530 - 5,008.
Link to video
5,008 console frames = 83.329431568254 seconds.
---
Fat Gate Thief - Level 20 - 28.053798 - 1,686.
Link to video
1,686 console frames = 28.053798247619 seconds.
---
I just finished editing together a compilation of the Gate Thief #1 "All 20 Levels" route. aka, Warpless.
I tracked the 346 and 440 frame times for each level transition as closely as possible, and the final product comes in at 4.358 console frames slower than the official projection. But still safely under the 23 minute mark.
official projection: 82,927 console frames = 22:59.8442036063492063492.
compilation video: 82,931.3579845 console frames = 22:59.91671725873903492063492063492.
Now that the full demonstration is available to view and study, I will continue to grind out an official speedrun, hopefully around 23:15 - 23:25.
edit: there's a small video error during level 16 that causes the video to freeze for 5 seconds, although the audio continues to play without a problem until the video catches back up. This was caused by youtube, as I used its auto-generated 720p60 videos to create this compilation (due to software limitations on my part). I didn't realize the 720p60 version of youtube's level 16 video had corruption in it until after I made the compilation video. All the original segment footage is of course available in 1080p60 in the videos I've already been posting in this thread. So this is a minor annoyance, but overall, it will suffice for the demonstration.
--
Gate Thief #1 - Warpless, All 20 Levels - Full Route Compilation.
Link to video
"22:59.8442"
I noticed something fun about SNES and old CRT TVs today.
The high frequency noise an old TV makes is related to the resolution and frame rate, and this numerical frequency translates into a real sound of the same frequency.
Usually, you will see information relating to the NTSC broadcast tv signals. It'll say the frame rate is 29.97, and there are 525 scan lines. So you just multiply them together, and you get 29.97 x 525 = ~15734 hz.
(actually, it's 29.9700 299700 299700 299700 299700... x 525 = 15734.26573426 573426 573426 573426...)
But today, I was looking at some audio from my Prince of Persia recordings, and noticed the sound on the spectrogram was lighting up around 15,746 hz. I wondered, how could that be?
Well, the math confirms that it should be different. Different frame rate, different scan lines. So these old TVs actually make different sounds, depending on what they're playing!
In the case of NES and SNES, the content is played at 39375000/655171 fps = ~60.09881389744 fps. And the signal is only half the number of scan lines.
525 NTSC scan lines is for "interlaced" video. That's 262 scan lines for frame A, and 262 scan lines for frame B, and 1 leftover line that's used to tell the TV whether frame A or frame B is next. Frame A uses only odd numbered lines, and Frame B uses only even numbered lines.
That's an average of 262.5 scan lines per frame, and an effective frame rate of 59.94 fps.
59.94 x 262.5 = ~15,734 hz.
So what about the SNES playing Prince of Persia?
Well, it's not an interlaced signal. It's progressive signal, at half resolution. So the extra line used for switching between A frames and B frames is skipped, leaving 262 scan lines per frame, and a frame rate of 60.09881389744 fps.
60.09881389744 x 262 = 15,745.88924112928 hz, exactly the frequency being displayed by my sound editor!!
WOW!! I have a giant update for Level 9!! I expected to save 4 frames when I redid it, but instead, I saved 73 frames!!!
I finally figured out the spacing to do a running jump across the first pit, after the conveyor belt. This saved massive amounts of time. I also saved time with a few more jumping optimizations throughout the level, and figured out an excellent way to save more time at the boss fight, by adjusting how I deployed the Select Menu.
The new total is 3,037 console frames, and this deals a heavy blow to both the "All 20 Levels" projection, and the "Warp 8" projection.
The "All 20 Levels" projection is reduced to 82,927 console frames = 22:59.8442036063492063492063492063. The 23 minute mark has finally fallen!!
The "Warp 8" projection is reduced to 60,951 console frames = 16:54.1797491047619047619047619048.
I got things down to 316 time units, and then found a way to save one last animation with a better jumping pattern in the moonlit hallway with guard. That last animation was able to drop the game clock down to 315. However, it is clear now that a 314 would be necessary to drop the displayed time to 0:44. And since I don't think there are another 2 animations to be saved, I think a 314 is unreachable with this sprite set. So it shall remain a displayed "0:45."
I'll update the stats in the previous posts later. It's time to celebrate!
edit: all stats updated.
--
Gate Thief #1 -- Level 9 -- 0:45 -- 315 time units.
Link to video
3,037 console frames = 50.5334432253968253968253968 seconds.
Epic best Level 3 recorded! 9 frames improved (2 animations improved on end button u-turn = 8 frames; 1 lag frame better also).
3 lag frames saved by "Killing" Skeleton on first pass. This was already cooked into the previous recording.
5,506 frames = 91.6157847873015873 seconds.
"All 20 Levels" projection updated above. 82,999 = 23:01.0422305777777777.
Still got 1 animation (4 frames) to save on Level 9, and need to test Level 13 for possible savings in conveyor belt area at end. Levels 9 & 13 improvements will also apply towards Warp 8 projection.
This run of 649 time units is 1 time unit shy of lowering the displayed game clock time to 1:31. Unfortunately, it appears there is nothing left to save!
--
Gate Thief #1 -- Level 3 -- 1:32 -- 649 time units.
Link to video
5,506 console frames = 91.6157847873015873 seconds.
I made a new video today for the Gate Thief #1 Setup + Warp to Level 8. It represents an improvement of 26.85 seconds over the previous Warp 8 Setup video.
From the beginning of Level 1 to the beginning of Level 8, the official Setup time is 6,832 console frames. Level 8 was performed in 4,667 console frames, from the beginning of Level 8, until the first frame of the Select Menu during the "End of Level" music at the end of level 8. The level transition was 437 frames. This gives a total of 11,936 console frames to reach Level 9 from the beginning of the run, which equals 198.606249 seconds (3 minutes, 18.606 seconds).
Level 8 was 269 frames slower than my best recording. I estimate this includes about 237 lag frames resulting from the Warp Glitch, and another 32 frames lost for conservative play.
I included a run through Levels 9 and 10 in this video also. So for this run, we have an official time of 20,542 console frames to beat the first half of the game, which is 5 minutes, 41.804 seconds.
From here, I can say 46 frames could be saved by using relevant segments from the Best Setup and Improved Level 1 Segment videos. I would also say another 16 frames can be easily saved with slightly less conservative play on Level 8, although it would be quite difficult, given the partial blind status. Assuming these, and combining time data from the remaining levels, I have a Warp 8 projection of 61,007 console frames, which puts us within striking distance of a sub-16:55, assuming I can manage to save a couple of animations when I re-record levels 3, 9, and 13.
--
Gate Thief #1 -- Setup + Warp 8 -- 113.679 seconds.
Link to video
6,832 console frames = 113.67944817777 seconds. (Setup)
4,667 console frames = 77.65544271746 seconds. (Level 8)