Okay, so.. I'm stumped more than ever with the 2x damage glitch.
I think that you can buffer an attack with regular sonic as long as the hitbox is disjointed, before the hurtbox is available (Not finish Jelly tonight, too tired, and can't confirm this either). Not sure if this can be used for boost since it's not largely disjointed.
Anyways. Touching CJ during the delay of the third stage explosion does a couple of things... sometimes:
1-a) It drops Speed and X-Velocity to 0.
1-b) You can't change X-velocity even by using a first or second stage explosion
2) CJ become invulnerable
3) The explosion is sped up considerable.. 10~13 frames
But, sometimes you CAN touch him during the delay and 1a/b and 3 will not occur, but he'll still become invulnerable.
Touching Jelly with the innate hitbox that's attached to burst will erase any charge. Touching Jelly with a level one explosion during the delay of the level 3 counts as a stage 3 for both triggering invulnerability and damage... but sometimes it won't, and CJ won't even be damaged by the level 3 explosion's large hitbox if that happens. It seems to depend on how close burst is to CJ.
EDIT
Definitely no buffering for boost, I was just really tired xD. It's possible to buffer a stomp to hit the first frame the hurtbox comes out, but I've been doing that since 2010 <_<.
Ended 15 frames faster. I lost two frames because of dropped input frames during the stomps which I couldn't help :/.
Uploads later today.
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
So i got your message on twitter, and just found the maps and have uploaded them. There was a time where i thought i'd lost them and would have to re-run my script to get them again.
progsoc.org/~cruizer/z51.png
progsoc.org/~cruizer/z52.png
Those are the maps locations for the time being. They will have a permanent folder location soon, just not yet as I'm sprucing up that webpage.
So far in SC1 I've saved 45 frames at 288 frames in. Improvements come from:
A) Actually using booster's slowed decay at the beginning of the level (I discovered it for myself in my old WIP at the end of this very level)
B) Only staying on the rails long enough to hit boosters and get a x-velo boost from jumping as soon as the rail curves.
C) Keeping speed stored when hitting first rail
D) Not using homing attack to get Rocket (The most important)
Homing attack "was" used because the Rocket is too low to drop to the ground and jump back out (It's setup so you are forced to use the Rocket you just got to exit the pit).
This time around, I found and utilized a trick to hit a capsule with boost while in mid-air to gain another mid-air boost in SM1. This is important here because breaking that capsule with boost give a small boost of negative y-velo (going upward). This idea worked (baaaaaarely) and I saved about 20 frames from just that alone (With homing attack, you're unable to move after hitting something until almost beginning to fall back down)..... I'll fix this last paragraph so it makes sense tomorrow, I'm in a hurry atm lol.
EDIT
Wanted to do a little bit more before bed. 91 frames faster so far at frame 731 of SC1. Relying on Rocket to go up instead of the launcher should be faster in the next section, so that should save quite a bit of time too.
Came across a glitch at the result of another glitch, made me laugh so I took I dumped a video of it.
Link to video
Not sure how useful this glitch is going to be anyways and I don't really understand it yet. Somehow landing on a booster pad with x-velocity above 3840*16 and boosting and sliding all on the same frame results in having undecayed x-velocity. Unfortunately the game breaks if you don't do anything to exit the glitch when entering a loop... and results in this video.
Also, sonic is technically running backwards in this clip.
EDIT
I think it's just an effect of the booster. Boosting is required, as well as holding forward the frame of landing on the booster.. at least in the scenario I'm in now. Sliding isn't necessary though. I don't know of any other applications of this in the run since getting x-velocity over 3840*16 is rare, but even more so near a booster pad. There is one occurrence in TR1, but boost isn't available yet lol.
EDIT Again (Sorry IRC lulz)
I think it might be pretty useful in SR or SRA however since getting over 3840*16 is always happening. It won't be useful for straight paths since backwards airdashing is faster, but something like inclined slopes definitely since x-velocity stays constant regardless of slope until it is "too steep", i.e.; a loop, it results in being stuck.
Just finished optimizing two different routes for traveling through a multi-railed section. I'm happy that the crazier of the two is faster (by only 3 frames) since it looks so ridiculous. It requires a lot of setup and it even overshoots the upper rail, so I have to maneuver back just to land on it.
I do have an idea that alters the faster route, but I wanted to get these two different routes optimized tonight and call it quits.
To stop you having to quintuple post, and so I can post something other than 'good luck bro', if you tick "minor edit" whilst editing a post your edits do not spam IRC.
Also; Good luck bro.
To stop you having to quintuple post, and so I can post something other than 'good luck bro', if you tick "minor edit" whilst editing a post your edits do not spam IRC.
Also; Good luck bro.
Thanks for the tip and thank you :D.
I've got the same problem I had playing back the run in 0.9.6 that I had in 0.9.7. I don't think I've tried to reproduce it since switching over (or maybe I have and forgot, it's been a long time). The RNG is the same the whole time in both playbacks (although that wouldn't be an issue) as is everything until an extra lag frame during the first playback. Maybe adding in a frame would fix it.
Although IIRC, it was only a problem I encountered lol <_<.
As for progress:
Almost done with SC1. I haven't calculated my savings in awhile, so I can't post that just yet. I'm currently stuck with multiple ways of progressing through a section again, so I'm optimizing all of them to figure it out. I need to progress downward after a slope, but there is a ramp in the way. I need to find the optimal method of keeping both X and Y velocities high. Jumping from the slope early (while under leftover slowed decay from boost) gives a short-term boost in x-velocity in trade for a lower short-term Y-velocity. By the time Y-Velocity starts to build up to a respectable number, X-velocity has suffered a couple of frames of regular decay. edit Due to some dropped input frames, this isn't an option unless I can overcome it.
Staying on the slope has two options itself (and from these two options, there are multiple options concerning using boost). Jumping and falling over the ramp, which is going to suffer x or y velocity loss (I can't imagine if x or y is more beneficial to keep high since there are spikes just past the ramp). Or I can land on the top of the ramp and keep a high x-velocity but at the cost of nearly 0 y-velocity for 15-20 frames. Then as I said, boost is the big key factor here and adds multiple options to each of those two options depending on when it's used.
Then to make matters more complicated, going too far in the air is going to cost getting that higher-than-cap x-velocity that is given from a straight rail. Although the rail curves shortly after getting on the straight rail (and cuts speed to 3840*16 and x-velocity to <2900*16), that extra x-velocity isn't something to ignore.
Bleh lol.[/b]
Finished 397 frames faster than my old wip with a final time of 1'02"37 which is over seven seconds faster than the real time record over at TSC. I tried every option I could on this level and I'm pretty proud of how well it came out. Having laser would cut this down by over 10 seconds (skipping the first rocket at the beginning alone would save 3-4 seconds), but it's unavailable due to aquarium park being unavailable :/. Just for fun, I might TAS this level over after unlocking laser.
2/19
I went back and tried to improve TR2 today. Using the curved rail, it's possible to boost x-velocity by 40% (so instead of being something like 13000*16, it would be 18000*16). The setup on the rail takes too long and the barrier loads... so no improvement could be made (I tried for two hours lol) :(
I used to love rails. They're in every level and always provide a good boost in speed/x-velo. More than anything though, they are pretty straight forward and don't require that many rerecords if you're actually riding them. Starlight Carnivals rails suck. SC2 has more rails than any other level and is the only level (off the top of my head) that has loop rails. The loop rails reset your speed to 3840*16, so all the extra built up on the rail is gone. The general rule when it comes to rails is that if you jump off, it won't let you back on unless you touch something else (ground, another rail). So jumping off to avoid the loop rail isn't a bad idea, except they completely ignore the rail rule. Avoiding a rail with a loop isn't a bad idea, but not always possible. It's possible to jump over the magnetic section of the loop (it snatches you right out of the air when you get too close). You've got to jump pretty high to actually do so and x-velocity is going to take a hit.
Debating what to do during a part when I ride the same rail for three seconds. The boost sound can be annoying, but just riding the rail looks plain and awkward. Maybe I'll time it with the music or something <_<.
2/23
I went through 62 out of 123 RNG values from the results screen last night for Orcan manipulation. I only saw five different routines. Each routine is different with what Orcan drops during his first two attacks. He drops three objects during each attack, so six total. One is always the Orange Rocket during the second attack.
There is always at least one spike out of the six objects, if not more. So the best routine out of the five is:
Attack 1: Barrel, Barrel, Barrel
Attack 2: Rocket, Spike, Barrel
So the first part of the strategy for Orcan is to knock the three barrels up, then break the Rocket wisp. Then I'll delay transforming for the best RNG manipulation of Orcan's next attacks (Since this is determined by when Orcan is hit by Orange Rocket).
Orcan has a frame rule and cannot be hit with Orange Rocket before frame 726. If it is, it'll go back to dropping stuff instead of the missile routine. There are two different routines that yield four total barrels and the rocket wisp early in the second wave. Either of them can be used because of the rule.
It's possible to get 2x damage by transforming very close to Orcan. When I tried, it lead to a time of 22"64. Being completely unsatisfied with that time, I decided that I was going to find another way. After a day of testing, I found it.
By use homing attack to travel faster [1] and setup perfect positioning, Orcan is hit right after transforming into Orange Rocket. Why is this faster than doing 2x damage? Because it gets rid of that awful Orange Rocket start-up delay. The new time is 21"84.
[1] - I'm thinking that this faster homing attack (about 2.2x faster than in normal stages) might be included in other boss fights. Although it doesn't warrant a restart of either Globotron or Jelly since it's not useful or even used against them.
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
VanillaCoke wrote:
Orcan has a frame rule and cannot be hit with Orange Rocket before frame 726. If it is, it'll go back to dropping stuff instead of the missile routine
.
And my frustration with Time Attacking this level now disappears! I'll test this in real time myself to see if this is true.
VanillaCoke wrote:
By use homing attack to travel faster [1] and setup perfect positioning, Orcan is hit right after transforming into Orange Rocket. Why is this faster than doing 2x damage? Because it gets rid of that awful Orange Rocket start-up delay. The new time is 21"84.
[1] - I'm thinking that this faster homing attack (about 2.2x faster than in normal stages) might be included in other boss fights. Although it doesn't warrant a restart of either Globotron or Jelly since it's not useful or even used against them.
I might have to see it in action before i fully understand what you mean? Words are being a bit confusing right now. However 21 is a really good time for Orcan, 2 seconds faster than the realtime record.
I might have to see it in action before i fully understand what you mean? Words are being a bit confusing right now. However 21 is a really good time for Orcan, 2 seconds faster than the realtime record.
Yeah, I was a bit tired when I wrote that post. I originally wanted to upload all of SC together in one video, but I'm not in the mood for learning how to make AviSynth scripts work haha. I'll upload the .dsm and a video to Youtube in about an hour.
I believe that the strategy I use can be replicated like this:
1) Jump at the lowest part of Orcan (without boost).
2) Transform the frame before hitting him (obviously this will take a handful of attempts in real time)
3) Y-velocity is carried over to the first frame of Orange Rocket and it's just enough to hit. That skips the long wind-up and continues with the two other hits.
I'll test it out to see if that actually works tomorrow. It might not since I use a homing attack instead of a jump (extra y-velocity). We'll see *shrug*. EDIT: I couldn't reproduce it without homing attack after a half hour of trying with tools. The extra y-velocity is probably necessary, if not then I'm not sure how to do it (I don't plan on spending any more time trying it out, it would be slower anyways).
I don't see any benefit to picking Aquarium Park over Planet Wisp, so I've chose to go to Planet Wisp and get Yellow Drill out of the way.
.dsm: http://dehacked.2y.net/microstorage.php/info/907842189/Sonic%20Colors.dsm
Youtube:
Link to video
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
Ah okay yeah, i see what you mean now. Seems to work as efficient as possible :o
Great SC section also, those speeds were pretty damn impressive. Also good luck with the drill sections in Planet Wisp. It's going to be fun to see how optimal your routes are. Also Drillinator should be close to real time record if stomp-boosting won't work, like in Globotron and Captain Jelly.
That 2nd wip is FANTASTIC, exciting, fast-paced and I still admire the amount of time you put in to optimise each level as much as possible, I fully look forward to watching more :)
That 2nd wip is FANTASTIC, exciting, fast-paced and I still admire the amount of time you put in to optimise each level as much as possible, I fully look forward to watching more :)
Thank you very much Hutch.
Found a couple tricks so far on PW:
- Ramming into the wall just before exiting a drill section will have those values carry over when exiting. Depending on positioning, it will cause the game to reposition Sonic above the gravel.
- Boosting the frame when landing on the tight rope launches Sonic immediately with the same Y-Velocity as well as being higher by avoiding the "dip".
- Boosting while under the effects of a vent gives 3840 or -3840 x-velocity depending on the direction boost is used
- It's possible to skip the hitstun when breaking rocks by holding down and forward one frame, then adding boost the next.
Then there is one hindrance:
- Trying to use Drill while having more than 4000 x-velocity results in ejection.
Finished in 1'12"25. Haaaaaaaate the drill.
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
VanillaCoke wrote:
- Trying to use Drill while having more than 4000 x-velocity results in ejection.
Finished in 1'12"25. Haaaaaaaate the drill.
That sounds like 'fun', from my memory of the stages anyway. Drill is pretty annoying to control in real time, so i can imagine how hard it is. 1'12"25 is a really good time for the first stage, knowing that drill sections can't go faster than what we get it at.
Final time for Planet Wisp 2 - 1'03"23.
Taking the drill tube route is much faster than using Rocket to go up and then down. I didn't optimize both routes to compare, but it was easy to come to that conclusion.
It's possible to go between vertical sets of mines and saves about a second by avoiding a U-Turn. That trick doesn't work on horizontal mines though :(.
EDIT
- RNG doesn't effect Drillnator since it only has one attack routine (same as Globotron)
- Boosting into Drillnator's arm doesn't knock it into it like Globotron because its drilling fist is underground.
- During Drillnator's invulnerability frames, Sonic can run through it's arm.
I originally had a time of 0'09"01. I figured that because of the second trick listed, it would be faster to defeat Drillnator by boosting every 9-10 frames. It took anywhere between 9 to 19 frames to get a desirable pass through Drillnator's arm, so it was a little more complicated than that.
So instead, I went back to stomping. I found a way to get Drillnator into its defeated animation (when it takes one more hit to win) faster. By delaying the last boost hit for 19 frames in the run, Drillnator was sent into its defeated animation 9 frames faster. I think that it has something to do with the life gauge catching up, but I'm not sure.
Finished with 0'08"77.
3/7
.dsm:
Youtube:
Link to video
*shrugs*
I have no idea >_>. I couldn't reproduce it on a second trip through the level.
3/8
- The Crystal launches laser out at an angle that has 2715 x-velocity (slower than using plain old boost). Plain old boost is faster at 3072 and so the Crystal is skipped and the laser wisp was held on to.
- I'm trying to find the best place to actually use Laser. I might just hold on to it until just before or after the first underwater section.
- I'm optimizing two different routes right now (which are both rather tedious). One using Laser to enter the drill section and the other is the default, lower route. I know the first route is faster, but I'm not sure if it's worth using the Laser on that route.
Laser notes:
- Shooting laser horizontal has an x-velocity of 3840.
..... not much else to say about Cyan Laser. It's a more sophisticated version of Orange Rocket.
3/10
AP1 is finished and I'm leaving the times for it and the rest of the run a secret until submitting. Three levels and three bosses left :O.
3/11
Major desync, have to redo AP1 from about halfway in -_-.
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
That 'Touch The Wisp' section was my favourite. It wasn't even on the touch screen XD
But really spot on with that entire zone there. Pretty much every route seems to be optimal, and well executed. I wouldn't be able to find any qualms on it unless I had the ability to go down to frame by frame optimisation. But i don't due to lack of time and resources. Keep up the great work.
Joined: 10/11/2009
Posts: 52
Location: Sydney, Australia
NrgSpoon wrote:
So uh.. I don't suppose there's some way you can hack out the boost start sound effect, is there?
While I am able to extract files from within the game, I haven't figured out, or haven't found a way to convert the .sdat sound data into something readable and familiar to program's like Media Player or Audacity. If you're talking about removing that particular sound, there would have to be a program that could edit out that sound without leaving a data inconsistency that would crash the game.
Just a few additions for Laser so I don't have to remember them later
- Hitting Laser objects like Crystals will set the gauge's value to a predetermined value if current value is below it.
- Bouncing off a surface freezes the value left for Laser temporarily. Multiple bounces result in subsequently smaller freezes.
- Laser lasts exactly 60 frames once movement starts.
I'm done with AP1 for a second time. Saved 9 additional frames in-spite of having to redo half the level. Figured I might as well make an improvement.
3/12
Sooooo, I was testing each frame to find when to jump and backwards boost for 3900~4000 temporary x-velocity off a slope (result depends on degree of slope and speed). Instead of running into the outside wall like every other loop, Sonic sneaks through the bottom left, skipping going around the loop.
I tried getting it to work with every loop encountered in the run (I've indirectly "tested" some already) and didn't find success with any of them >_<. I didn't expect it to work with regular loops and I like I said, there are a handful of loops I've already "tested" indirectly when finding the best frame to jump boost from a slope.
3/13
- RNG doesn't effect Admiral Jelly.
Found a glitch with Laser in the Admiral Jelly fight. By transforming the same frame as hitting A.Jelly and holding away (if jelly is left, hold right), Laser gets the bounce from hitting the boss and moves during the start-up delay. It's a boss only glitch like the moving rocket in the Orcan fight (that isn't used in that fight though, it breaks the fight lol).
3/14
Void is the only wisp who you can't immediately touch during it's Touch-A-Wisp, so I tried to jazz it up. I spent way longer on this than I should have lol (about two hours). Getting the coordinates correctly with the mouse is a pain, so I tried to edit the .dsm. It didn't work at first, but I eventually found that the value at the end of each line represents enabled and disabled.
Only two more levels and two more bosses.
3/19
Sorry there hasn't been any updates from me in a few days. I'm not sure what the issue is, but I've been dizzy the past few days. I finished AC1 prior to the dizziness and will finish up AC2 after I've recovered.
- For the first and only time in this run, I have had to carefully plan out boost usage in AC2.
- Boost doesn't work with <5% remaining (I should have known this, but keeping boost high was never important prior to AC2 lol).
- The bottom route is faster after the first roller-coaster (40 frames faster), but would be taken anyways unless it was drastically slower just for Laser.
- By using the gained Laser, it's possible to skip the second roller-coaster because the loops are actually solid and can be walked on.
