Posts for Akuma


1 2
10 11
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I got it! A very clean recording that's 10 frames better than my previous best. The previously published best was 2,444 frames. This one is 2,434 frames = 40.499967 seconds. The main takeaway is that before I implemented this new technique of using my Right Thumb to hit Select during Exits #2 & #3, I was usually not even able to get 2,444. The old method of using the Left Thumb to hit Select, and then come all the way back to the d-pad, was very inconsistent, and full of delay & hesitation. I very rarely came within 5 or 6 frames of my best. But now that I'm using the new technique, it is very consistent, and I'm within 1-3 frames of my best on almost every attempt. So, in essence, this represents much more than a 10 frame improvement for actual speedrun attempts. More like a full quarter of a second, which will go a long way towards helping me reach my ultimate goal of a 6:52.x in this category. Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I am having some success with the new Select Button technique on the setup. So far, I've recorded one that's 8 frames faster on all parts combined. But I think there's a little more I can squeeze out of the last menu exit. So I'm gonna keep going, and see if I can get it up to a 10 or 12 frame total improvement before I upload the next video, and update my sum of best.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I've mentioned previously that I have been using an alternate method with my hands, when doing the Level 17 reset. Once my right hand is no longer needed during the approach to the spikes, I use my right thumb to hit Select, and as my left thumb has remained on the d-pad, I can much more quickly begin scrolling down the Select Menu towards the exit selection. Under normal circumstances, I've generally used my left thumb for Select, followed by a slight delay as my left thumb physically moves back to the d-pad. I've realized that I can use this same technique when doing the exit glitches on level 1 and level zero. The difficulty here is that I'm using the A button to hold onto the ledges, so my right thumb is preoccupied until the Prince begins the climbing animation. Therefore, I have a tight window to release A (no longer needed once the Prince begins climbing), and move my right thumb over to the Select Button during the 3-4 climb animations. But with correct timing, I am once again having the advantage of being able to keep my left thumb on the d-pad when it's time to navigate the select menu. Therefore, I am, on average, saving 3-5 frames on each of the 2nd and 3rd menu navigations, for a total of 6-10 frames of savings during the "Setup" segment of the sub-7 speedrun. If I'm able to record a new demo at 2,434 frames (10 frames better than my current demo), and able to consistently save this time during speedruns, this will be a much needed boost towards reaching my goal of a 6:52.x for RTA timing of a full speedrun.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
RTA update! The MODs on Speedrun Dot Com agreed with my thoughts on the timing rule, and have set the runs to begin on the First Frame of Level One.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I got a new World Record! -- 6:54.5672499. In honor of Suigi's infamous pause during the final bowser fight of his recent 70 star world record speedrun on super mario 64, I have allowed my own embarrassing pause moment to be published. I had a completely inexplicable brain fart that's never happened before. Early in Level 18, my mind just decided to go hit the Select Button. Fortunately, I was able to react quickly and unpause after just half a second. Altogether, the blunder cost 39 frames in real time, without which this run would have been low enough to break even another Second Barrier. My previous WR was a 656, so this 654 breaks 2 second barriers, going below 656 and 655 in one go. And this is, of course, in accordance with the Old Timing Standard. I have decided to disregard to recent rules change on Speedrun Dot Com, which starts timing literally when the player gains control, on the frame the Prince stands up straight after falling through the first gate. This new rules change flies in the face of the precedent set by many previous versions of the game on other systems which start out exactly the same, with the Prince falling through the first gate, landing and standing up. There, timing has begun on the First Frame of Level One when In Game Timer is not the official standard (not always available or reliable, depending on the version). Furthermore, due to random Lag Frames throughout the SNES version, the Prince sometimes stands up after 73 frames rather than after 72 frames. Following this discrepancy, the game may offer more lag to the 72's, and less lag to the 73's from that point forward, causing a discrepancy of up to 3 frames in just the first couple of rooms. Not at all fair or consistent for speedrunning, in my opinion. Therefore, for the foreseeable future, I will be submitting my times based on the First Frame of Level One, in protest of the rules change, and I will attempt to convince the MODs to revert the rules change. SO, this is a 654.5672499. Without the 39 frame pause blunder, this would have been a 653.918. I can see from this run that a 653 is easily within reach. However, it will take a lot more work to get a 652. But I'm encouraged by the progress thus far, and look forward to seeing what's next. I think once I get a 652, I'll be done with this category, and get back to work on my Warpless Sub-23 minute runs with GT1. Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
In other news, the "RTA" equivalent for Challenger's latest TAS has been reduced by 73 frames for now, given new wording for the RTA rules on the Speedrun Dot com website. Previously, the wording had been that RTA timing begins when the Player Gains Control of the Prince. I had been ignoring that language all these years, and simply began timing on the first frame of Level 1. It made sense to me, since timing for every other Level begins on the first frame of that level. Furthermore, several versions of the 1989 edition did the same thing where In Game Timing was not available, or was not official. In those instances, RTA timing also began on the first frame of level one. And in several of those versions (various consoles and computer formats), the Prince fell from the gate just like in the SNES version. However, many players have been interpreting the language of the SNES rules chaotically, with some starting time when the Prince touches the ground, and others starting timing when he first stands upright after landing. In an attempt to clear up all this confusion, I brought it to the MODs attention, and for now, they have latched onto the first frame the Prince is standing upright after the fall, rather than first frame of the level. I am attempting to discuss the matter further with the MODs, but in the meantime, the current RTA rules say Challenger's latest TAS is 73 frames shorter now, giving a new total of 24,359 console frames = 6:45.3158194 031746 031746 031746.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Another fun discovery today. "Invincibility Mode" is actually not invincibility at all. It's just a whole bunch of extra hidden bottles on the meter. What I've been referring to as "Invincibility Mode #1", which is described as being able to survive 1 fatal hit without dying, is actually 160 extra bottles (off screen), which function exactly the same as normal bottles. Including the 15 bottles meter on screen, this gives a total health of 175 normal hits. IE, after you stand around and let a guard hit you 160 times, that 161st hit finally makes a change on screen, turning the 15th bottle dark, and leaving you with 14 more hits. Similarly, "Invincibility Mode #2" is actually 240 extra bottles off screen, for a total health of 255 normal hits. IE, if a guard hits you 240 times, or you fall down a 2 level gap 240 times, the next time you do it will make a difference to the on screen bottles. The important discovery is this: fatal hits don't have a special parameter that tells the game whether or not to kill you. They simply do more damage. 100 hits! So when you're in INV#1 with your 175 health, the first "fatal" hit does 100 damage, and leaves you with 75 health, not enough to survive another 100 damage. Whereas, if you have 240 hits, you can survive 2 separate "fatal" hits, and still have about 55 total health to work with. I have shown some setups where you can knock an overhanging loose panel, land on spikes, and then the falling panel hits you and returns you to a playable position (instead of being stuck on the spikes). In a case like this, you take 100 damage from the spikes, and an additional 1 damage from the falling panel, for a total of 101. If you do this process twice with INV#2, you have survived in a playable state, taken 202 damage, and have 53 remaining health, and can take another 38 normal hits from a guard without noticing any loss of health on the original 15 on screen bottles. The Crusher, the Chomps, the Spikes, the Sparkles, and being hit by a guard when your sword is put away, all do 100 damage. And falling from a height of 3 levels or more is also 100 damage. Fire Walls on Level 11 do 100 damage. Being stomped by the Monster on level 17 without your sword drawn is 100 damage. One possible exception to this rule, which cannot readily be tested, is Jaffar's attack when he lifts you in the air, and destroys you with a fireball. When I very quickly access Debug Menu, I can select "Boost Meter" to restore all of my health, and stay alive for another hit. But eventually, the hits overtake too quickly for my health reset attempts, and I die. I think if Challenger was able to run a test tas, you could enable debug menu, jump to Level 8, do the quick warp glitch off the ledge after the first guard to acquire Invincibility Mode #1 (160 invisible bottles), then jump to level 20, play jaffar, and let him do his air fireball killer move, and see if you can access Boost Meter on every frame, to see how long the TAS can stay alive. lol! I made some fun illustrations for the meters, below: INV#1 (160 extra bottles, 175 total health points): INV#2 (240 extra bottles, 255 total health points): .. I've also confirmed that in both cases (+160 and +240), large bottles, debug health boost, and merging with ghost will only refill existing meter to full status. These do not create extra meter count. Presumably, these are only programmed to increase meter count when there are less than 15 bottles onscreen. I can also report that in the past, I have done glitches where the Prince survives a fatal hit, and loses onscreen health meter. Previously, I thought it was some sort of broken setup, but now we can use new knowledge to suppose there is a setup which gives something like +95 invisible bottles, so there is enough total meter to survive the 100 hit points, but not enough to conceal the effects on the visible meter.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
During a recent speedrun attempt, I tied the TAS exactly on Level 16. Counting from first frame of the Level through the first frame the Prince is touching the finish line: 3,997 console frames. I made a fun split-screen video, showing Challenger's most recent TAS on the left, and my Perfect Level 16 speedrun on the right! Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I did a fun video experiment today. I turned my phone the other way around, so the video camera would scan the field upward, in the opposite direction as the TV, which draws lines in downward succession. This allowed for a much more exact timing method, as a very sharp line appears along the bottom of the tv's progress on the first frame of any new screen (after a black transition). Knowing the Camera scans upward at about 84 fps with this spacing I use, and the TV draws downward at 60.0988 fps, I can create extremely accurate timestamps for each frame at the beginning of any new level or new room. The following stats were created for the casual run I did today -- 01a -- 1043.0658602792248130640702961517 (1,043) 01b -- 1426.9198387923588977970200582312 (1,427) 16o -- 5327.9707817010477308323214201222 (5,328) 17a -- 0411.9668600694649619281509276644 (0,412) 17b -- 2349.9809092260973525559412297379 (2,350) 18a -- 3238.9765618005506523160692835483 (3,239) 18b -- 2520.9379530442107044246027809098 (2,521) 18c -- 2619.0024596632024311210355769718 (2,619) 19a -- 2863.9643307679543639315973561539 (2,864) 19b -- 2608.9013248476897089244094835002 (2,609) 20o -- 1706.1017647682818683977160161242 (1,706) You see for the most part, the timing (frame count) is slightly off, thanks to the inaccurate variable fps of the video. But it is close enough to say with certainty how many frames were in each segment. The raw total (time based) was 26117.78864496 frames, and the adjusted count was 26,118, which matches the pattern of "rounding up" that I always do in my videos. I think I will continue to record the way I always have, as this upside down method is more tedious, but definitely proves or reinforces that my timing has been correctly calculated in the past. I had to do a little bit of in depth observation and calculation to figure out exactly where the intersection line was for the final frame, as you can see in the picture below:
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I've discovered an interesting difference between my video recordings of my TV, compared to BizHawk videos captured during TAS playback. Whenever the Prince goes to his final position upon touching the finish line on Level 20, the BizHawk videos initiate the temporary soundtrack silence either 0.4 or 1.4 frames BEFORE then. This silence lasts for a couple of frames, as the system is preparing to load the "End of Level" music. For example, if the Prince enters his final position at 7:13.1001 on the TAS video, the silence already started at 7:13.0768 according to the audio track, some 0.0234 seconds earlier. However, for all of my video recordings of my TV & Console, the silence begins later, either 1.4 or 2.4 frames AFTER the Prince has already made his final movement and touched the finish line. I'm curious to know why there's a difference. I might run a couple more short test videos of my TV and Console tomorrow, to see if anything different happens.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Today, I did 1 attempt at new world record, and luckily got it. World's First Sub-7 Minute RTA Speedrun on original hardware. I turned around 1 frame early on the 2nd gate of Level 18 to bonk and lose 1.76 seconds, and I jumped 1 frame early through the 1st gate of Level 19 to come up short and fall through the loose panel to lose 2.71 seconds. Also, for whatever reason, I thought I had previously done a lag test which concluded that it didn't matter when you kill the 1st guard of Level 16. However, I was very much mistaken, and picked up 24 extra lag frames on Level 16 by waiting to kill the guard later. Very silly! Altogether, about 4.87 seconds lost on those 3 silly mistakes. Everything else was fairly clean and consistent. 25,199 console frames = 6:59.2928007365. And without those mistakes, would have been around 6:54.434. So I think a 6:54.x is definitely in the cards soon, if I can get a good run without such mistakes. Then I can start working on doing all double-crouches on Level 17 instead of the more conservative triple-crouches, and clean up a few spots and fast menus, which should get down into 6:53.x territory! Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I'm getting back into practicing for a 6:55 run or better. As a part of that practice and warming up, I made a fun meme video of beating the game in 6:32 with the Game Genie Code for No Gates. For the entire route here, every segment, every room, every gate skip, is a combination of previously used strategies for either Gate Thief #1 or Fat Gate Thief. So in a sense, this is kind of a fun "What If" video, considering the possibilities for the main speedrunnning & tas'ing category, if only this particular Warp Glitch had given us an optimal Gate Thief sprite, instead of our Guard & Skeleton sprite! From first frame of Level 1 until first frame of Select Menu at the end of Level 20, we have approximately 23,613 console frames = 6:32.9029288380952380952. I counted about 1.9 seconds worth of mistakes & inaccuracies. Lost 25 frames on setup, 19 frames on Level 17 Reset, 22 frames on conservative log crouches, and 47 frames bonking the wall after Level 18 boss. So it looks like a 6:30.x is doable, but not entirely sure if there's a full second worth of savings elsewhere to bring this meme route down below 6:30. Would be fun to see, but I'm sure Challenger has more important things to do than make Game Genie TAS's! XD Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I'll mention quickly that I prepared a 1080p60 video from the most recent TAS file (with no Level 17 reset). BizHawk encoded the video at 60.098476 fps, so it's very close to running at true console speed. From first frame of Level 1 thru first frame of Select Menu after Prince crosses the finish line on Level 20, there's a total of 24,432 frames, which gives the latest TAS an RTA equivalent time of 6:46.53049. I noticed that you were able to get the faster 4 frame pull-up of Select Menu there at the end of Level 20. So the new route eventually saved a couple more RTA frames that way. It also appears that the alternate route saved 1 more frame on that first Level 19 gate, compared to the previous test TAS, which had improved level 19 but not yet adding the Level 17 improvement. So the Level 17 improvement has rearranged the various lag errors, and created an optimal situation for levels 19 and 20, for RTA purposes. This sure is looking pretty solid now! Unless you come up with any more ideas, I think this is really close to being ready for publication!
Experienced Forum User
Joined: 12/26/2018
Posts: 269
A quick note today. For the time being, I think I will at least get GT1 as low as possible before transitioning to GT2 on the "All Levels" (Warpless) Category. I'll of course keep my primary focus on Any% until I can post up a good 6:53 or 6:52. But after that, I will try to make a good final version of the GT1 setup, and update the Spliced Compilation Video, and possibly give GT1 one last good run, before I start to learn the new route with GT2. One main point of concern for GT1 was that the better setup left the partial button scramble. But I remembered today that I can reset buttons without wasting any time by waiting until mandatory Select Menu pause. So I can roll that into the route at the end of Level 2, or perhaps more conveniently, during the death music at the Level 3 checkpoint clip. Since I need to re-record the Level 3 clip video anyway, I might go ahead and do button reset during the Level 3 video, for documentation. We'll see how motivated I'm feeling! :P
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Challenger wrote:
Even if the second one could work, unfortunately the height is high enough for an unavoidable fall death as soon as he lands on that raised floor from the room below.
Are you saying that when the Prince clips through the floor in this pattern, he does not actually touch the first floor, so there is no resetting of the falling height? Since I only did it once, I couldn't remember exactly what it looked or sounded like. I guess I assumed there was some kind of hit, like on Level 3 when you clip the edge and take damage while falling into the next room below. But if there's no hit on the first floor, I guess that makes sense that it would end up being at least 3 levels falling, which will kill you. So it would only work with Invincibility, and very lucky circumstances in the next room to fall next to a wall for a pop-up, or landing on a loose panel below, etc.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Challenger, I was just thinking about something weird I did once. I can't remember if I asked you about it. I wonder if you would be interested in testing this. I feel like there must be somewhere in the game where this trick could line up for a big skip. In the picture, RED arrows show what I actually did. I jumped off the ledge, landed in the pit, near the wall, and clipped through the floor to fall below. But I fell into the next pit, and died, so the shortcut was unsuccessful. I have never been able to repeat it, so I haven't been able to test whether, in this case, the GREEN arrows might show a viable shortcut option. I think this could come in handy somewhere. For now, I can't remember whether it is Gate-Thief-Only. I also can't remember for sure if it was a running jump or a standing jump. Obviously in this case, if it was a running jump to get the perfect spacing, it could not be done from the 2nd pit here. But I still think it's worth looking at, in case there is some clip somewhere in the game that can be exploited.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I had a new idea today. Speedrun dot com just updated the pop snes category labels again, with slightly better logical organization. The Any% category has a "No Exit Glitch" sub-category, which is the original glitchless category, and a "Standard" sub-category, which is the sub-7 minute route we're working on here. Separately, my "All 20 Levels (Warpless)" work with GT1 has been given its own stand-alone category called "All Levels," denoting that this is a separate condition. This sets up a good template for the completist categories to come. 100% could be divided into glitch vs no glitch categories like Any%. And a separate category for "All Life Bottles" could mirror "All Levels" for once again denoting a separate condition not specifically related to game %. So my idea is this: I had been begrudging the possibility of having to redo all of my GT1 individual level videos to make SLOWER, INFERIOR individual level videos with the new GT2. But, the new GT2 sprite could definitely find some immediate use as the basis for an "All Life Bottles" run. I would make quick work of the setup, and be off & running to the bottles in no time. And the new routing to include the bottles would provide a new challenge that's not specifically related to undoing my GT1 work. So this may be yet another project that I take up in the near future. But I do still want to try to make a good Sub-7 RTA record first, before I get too distracted.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
ok, I've got some fun updates today. For now, it looks like the alternate Zero^2 setup for GT1 is not useful, as it requires exit at first guard. If you do the guard skip, and save progress to level 2, the setup sends you to the room with the other guard, rather than to the room directly above. In other words, you'd have to do the guard exit, the full setup, and then play level 1 all the way through again. So it'll be 10+ seconds slower. As for the alternate Gate Thief Sprite Set, it does seem to have promise. The setup is very cooperative, and I've been able to use it several times in a row, without a single mess up, or freeze. It also works just fine after completing Level 1 and saving progress to Level 2. Furthermore, as soon as Alternate Gate Thief is acquired, and Level 2 begins, there does appear to be much less lag than standard GT1. So Alternate GT gets off to a great start with 38 seconds saved on setup, and slightly faster level 2 from better lag. Levels 2, 3, 5, 10, 15, 16, 17, and 20 appear to be identical to GT1 in terms of abilities, and no limitations to cause slow-downs. That leaves levels 4, 6, 7, 8, 9, 11, 12, 13, 14, 18, 19. We already have a good idea of the potential for levels 18, and 19 from your TAS test, which loses about 78 frames altogether. So that just leaves 4, 6, 7, 8, 9, 11, 12, 13, and 14 to stay within 36.5 seconds of GT1 to be better, averaging 4 seconds per level. Early scouting reports reveal that it should be well below this mark on each level, with only a few delays here and there. After playing the other levels, it feels like each one only has about 2 seconds worth of delays, due to slower movements in approaching gates, etc. So I think we're gonna find that this would be about 18-20 seconds faster than the best GT1 route. I noticed in your full TAS for this alternate GT sprite, that you intentionally collided with the wall to fall short of the gate before the boss on Level 18, so you could get a full jump into the boss room. This will be a 26 frame improvement, plus the 1 animation save in the hallway below the end button, that's 30 frames, or a full half second. So I can tentatively put the GT1 full projection around 81,344 - 190 = 81,154 = 22:30.34278943492. (160 for alt setup, 30 for better Level 18). I know there's another 8 frames to be saved elsewhere, which would bring it down to 22:30.209675327 with 13 more frames needing to be saved in order to reach a sub-22:30 for GT1. So if the new Alt Sprite is 18 seconds better, looks like the Warpless Route may have a potential for 22:12 or better. I think it may be appropriate to refer to this alternate GT sprite as GT2 or Gate Thief #2, as it certainly seems to have priority over any other alternate gate thief sprites in the context of the Warpless Category, (like wobbling sword GT, or Fat GT).
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Despite appearances, and the overwhelming lag caused by the newer setup, using the lag reduction trick does still work on the first guard on level 2!!! The jump pattern is 24 frames slower, but the trick saves 67 lag frames (net lag differential, as the falling panels cause extra lag). So 67 - 24 = a net savings of 43 frames by continuing to use the lag reduction trick (killing the guard while he is swinging his sword). Furthermore, I have refreshed my memory on the better setup from one of your most recent GT1 test videos ("Prince of Persia - GT1 TAS with more optimizations"), (8:35 TAS time), and I think a human can do something like 132 frames better than my previous setup projection (including extra lag from Level 2). I will continue to work on a good video for this, and maybe get it up to 160 frames with tight movement & menus. So far, my best segments add up to 132 frames saved, including all extra lag, so the GT1 projection should definitely be considered below 23:31 now. You may recall that I had, at the time, said I was unable to duplicate your pattern where the Prince simply climbs up & back down through Hole #5. It had been freezing the game, and so I had been trying alternate exits, like turning and running off the ledge. However, today, I implemented all of the main movements, including the crouch to stop before the gate, and doing a standing-jump through the gate to the spot underneath the loose panel. Today, I've had no issues with the game freezing on this pattern. The only remaining problem is the partial button scrambling. It concerns me that B is changed to Crouch, because there may be a possibility of me accidentally holding it too long, and buffering a crouch at an inopportune moment! (It's not a huge problem, because the game is hard-coded to recognize the B button as a Jump Button first and foremost, before considering any changes made to the controller settings.) This problem is fixed with the other GT1 setup you tried ("Prince of Persia (SNES) - another GT1 setup"), (similar to Warp-16, but climbing up to the top of the screen and jumping from up there, instead of immediately turning and jumping to the Warp-16 spot). However, this alternate setup is almost exactly the same length (what a funny coincidence!), so I'll need to tighten up both for a further comparison, and see if the Zero^2 setup causes less lag than the Zero^1 setup. At some point though, We'll have to consider the alternate sprite you found, which was acquired in just 7 seconds, hahaha. It will be hard for GT1 to make up 38 seconds over the course of 19 levels. From the final TAS you made for the alternate gate thief ("Prince of Persia (warp glitch TAS - fastest setup)" (7:53 TAS time), it does look like this guy has a few limitations on spacing to jump through closed gates, so I did notice some places where you had to slow down quite a bit, like doing a U-turn on the Island on Level 4, or jumping through the 2nd gate on Level 18 to grab the ledge, instead of just running through. So who knows, if there are a couple of seconds worth of problems for this alternate sprite set on each level, perhaps GT1 can catch up! It'll be fun to find out eventually! @__@
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I decided to take a day off from this sub-7 minute route, and take another look at GT1 setup for the Warpless category. So far, the findings are disappointing. The alternate setup you made for the GT1 warp-4 route improves things by about 350 frames. But, very sadly, it appears to cause grievous lag, much worse than the other setup, especially in the final segment of level 1, where the Prince jumps into the hole to initiate the sprite scramble, and even worse on Level 2. It looks like there's upwards of 250 extra lag frames, in addition to previous lag, and also the lag reduction trick used in the other setup by killing the first guard on level 2 while he's swinging his sword, does not appear to work at all this time. So, initial tests appear to show that the final GT1 setup you cleaned up, may only have a potential savings of around 100 frames for the Warpless Route (about 1.66 seconds, instead of 3-5 seconds). I have more testing to do. Once I clean up my attempts, I may find that the lag situation improves slightly. I wonder if you would be interested in helping me make a small test video with your TAS, to see what the best potential is. To refresh your memory, it would involve segments from multiple different GT1 videos: segment 1, complete Level 1 as quickly as possible, and use START+SELECT Music Menu to bring up password box sooner (skipping full "end of Level" song). Exit Level 2, and restart level 1. segment 2, Single Up at Hole #2 to warp to Level Zero, climb down, go to the left 2 more rooms, climb up & down Hole #4, and exit. restart Level 1. (remember you jump to the button to open the gate, crouch to stop short, then do standing-jump through gate underneath loose panel, knock panel & crouch to avoid hit.) segment 3, jump into Hole #1 & exit. CONTINUE to Level 2. segment 4, complete Level 2 as quickly as possible, probably following my Level 2 video. At beginning of Level 3, confirm Best Times board (hopefully no Low Time garbage) and Button configuration (hopefully no button scramble). ALSO, FYI -- maybe you didn't notice, but I did finally get the Speedrun.com categories updated to include our Sub-7 route, AND the GT1 Warpless Route. I had previous posted a video here of my nice All 20 Level Compilation, which is down to 22:33.5. At worst, it would appear I've already confirmed enough savings with the alternate route described above to bring the route projection down below 22:31, which is nice for sure. Anyway, now that the Warpless category exists, perhaps you will finally find some interest in thinking of my GT1 Warpless Route as a good candidate for a future TAS project, after we have finished the Sub-7 Minute Route, and you've had time to work on other projects....
Experienced Forum User
Joined: 12/26/2018
Posts: 269
That's very interesting! A couple of questions -- At the end of Level 16, I hear some very fast menu navigation. As far as I know, there is no reason to make any changes to settings, or controls. So why is there any extra menu noise here? Surely this is adding to the frame count. (Well, maybe I can guess -- Are you resetting sound from Mono back to Stereo during music replay, so no real time or game time is lost, and this is theoretically improving playback quality? lol) Sound is changed from Stereo to Mono as a consequence of the Warp Glitch. This is an indication that it successfully enabled Kill Button. There are other Warp Glitch patterns which do not activate Kill Button, and in these cases, sound is not switched to Mono. Also, I noticed you are no longer pulling up the Select Menu at the end of the Jaffar fight on Level 20, inside the Arena. Did you determine that it did not save a frame in this instance? Or did you forget it? ======= I will be curious to see if this Level 17 death improvement will be consistent for human RTA. Who knows what may cause the skull patterns to change? lol.
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I've discovered another strange potential time save for TAS, this time on Level 17. So, whatever I did differently in the Boss Room + the Final Room allowed me to bring up Select Menu 7 frames faster than TAS in the final room. One would assume this means the End of Level music did, in fact, launch sooner, and should bring up Level 18 sooner. However, this particular spacing and timing may also require the movements I did in the Boss room, which was 4 frames slower than TAS. Thus, it is a net improvement of 3 frames in the last two rooms, compared to current TAS. Now, the reason I say this is a strange improvement is because it is possible that this is a fluke, which TAS will not be able to reproduce -- similar to saving 3 frames on a Princess Cut Scene, when coincidental lag save allows you to stop the cut scene 1 animation (3 frames) sooner than usual. The fluke is noticeable here because I was able to bring up Select Menu after 3 frames of Prince touching the "finish line," instead of after 6 frames. However, I would recommend giving it a try, as perhaps the alternate spacing can reproduce timing that launches the music 2 frames sooner. I now have 2 different runs that are clocking in at 1,829 frames, which is 6 frames slower than TAS. The one I am showing here is the older video, with the worse Game Clock time of "0:30." It was due to inaccurate jump patterns in the first half of the level. These jump patterns were improved to match TAS in the 2nd video, which was able to acquire the "0:29" game clock result & password. However, it's still 6 frames slower than TAS, mainly because of the boss room, but also an unlucky lag frame at the beginning. The boss room difference might be a jump pattern, or something to do with the Select Menu. I do not know. I was in the same position as TAS at the finish line, and I had Select Menu open during Boss music for the exact same number of frames as TAS. So I'm throwing my hands up on this one. At least my second video was able to improve the game clock execution, and get a low PB time without a possible fluke. I will be uploading the new video with the improved "0:29" game clock time soon. But in the meantime, here is the former video, which I think you should have a look at, especially at the end, to see if TAS is able to reproduce this fast Select Menu press at the finish line. Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
More Level 19 stuff. I matched TAS exactly (!!!) on Select Menu open & close + first kill. Here's the interesting part. Overall, my boss battle was 5 frames slower than TAS, but this included 3 (!!!) late kills, which should have been +12 or worse. Therefore, I must assume I did something different to spawn guards sooner than TAS to make up for lost time. I would suggest taking another look to see if TAS can make up this phantom difference of 7+ frames. Perhaps TAS inputs or movements were delaying spawn in some cases? I picked up extra lag in the first section, so only 2 frames faster than TAS instead of 3 or 4, but it still shows that the alternate jump & run through the first gate is faster than TAS. I'm kind of burned out on this level for now. I think I've done all I can do. I'm getting the "1:14" on the clock pretty consistently now, but I keep picking up +2 half animations in various places, so it's frustrating, but at least this video is officially a new PB at 5,039 frames, and the important "1:14" on the game clock. (529 time units is the highest possible password to still display 1:14 before bumping up to the next second.) It feels like I could get another 2 or 4 frames knocked off the top, since other former PBs have some better rooms, but I just can't get it to work! Levels 18 & 19 have boosted my confidence in trying to match TAS on Boss Select Menus, so I'm going to go back and run Level 16 again, which still has 12 frames to save on boss. Also, I can save the last 4 frames on wall clip, which I did in a subsequent video for the Warp Setup. lol! If I can actually tie TAS on Level 16, that'll take my PB projections all the way down to 24,770 = 6:52.1545567, with a decent amount of wiggle room to shoot for a 6:52.x in a full run. Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
I am in the process of closely examining the TAS on Level 19, to figure out where my mistakes are. My posted video is 43 frames worse. However, I'm actually better on the first section. How funny is that? My approach to the first gate, after the first wall clip, is actually 1 animation better than TAS. I jumped sooner, and had 1 extra length of running to stop after the loose panel. But the wall clip room is 1 animation faster, and the first climbing room is tied. So net savings of 3 or 4 frames, depending on lag / cpu float. I will post the video here, so you can see what I did. This should make the TAS faster, and will possibly have an impact on Level 20 later, perhaps allowing for the faster select menu pause at the end, which may save 1 more animation in the hallway, and allow for a sooner unpause to end the TAS faster on TAS timing. Link to video
Experienced Forum User
Joined: 12/26/2018
Posts: 269
Yes, I see it's only saving you 8 frames. I completed 4 "perfect" runs so far. One of them, posted above, is 1669 frames, getting the Select Menu to come up after the final animation of Prince's movement is only on-screen for 4 frames. In the other two "perfect" runs, the time was 1671, because the final animation was on-screen for 6 frames before the Select Menu appeared. And in one other "perfect" run, the time was 1670, because the final animation was on-screen for 5 frames. There is some fluctuation of on-screen time for other things as well. For example, in all of my videos, the time remaining window appears on-screen after the final white strobe frame of Jaffar's death, and is on-screen for only 70 or 71 frames in every case, followed by 62 frames before the fade-out pixelation begins. In your TAS playback, the Time Remaining window is onscreen for 73 frames, followed by 63 more frames before the pixelation fade-out begins [+4]. I think the game adjusted by having a faster transition for your playback. I think you can manipulate the timing of the Select Menu at the End of the Level by adjusting the time and duration of your Select Menu manipulations during the Jaffar Intro in the Hallway, and after the Jaffar death. I would suggest trying a few variations, and see what happens. Under the right circumstances, the game may allow the final Select Menu window to open 2 frames sooner for you, which would give you a lower RTA time, and possibly allow you to unpause sooner at the end, for shorter TAS time. (For RTA time, your TAS playback is currently tied with my two worst "perfect" runs, at 1,671, because the last animation is on-screen for 6 frames instead of 4, before the Select Menu launches.) ==== For the other thing, I think it doesn't work because the Walls don't have the same collision circumstances as Gates. After the approach and buffered crouch on Level 16, you could stop inputs (no jump), and be standing there in the wall, with no wall collision. And if you started running, you could run in place without hitting the wall. None of that is possible with Gates, because of forced Gate collision. Too bad!
1 2
10 11