With the addition of three new shortcuts, Sonic and Tails finish the game about 45 seconds faster than the previous run. Although neither Nitsuja nor Upthorn were actively involved in this improvement, they are listed as co-authors because their input comprises over 90% of the movie.

Game objectives

  • Emulator used: Gens 11a + S3KCamhack + SolidityViewer + (partial)HitboxDisplay + LUA HUD for Genesis Sonic games
  • Ignores delays caused by bonus effects
  • Aims for fastest (primarily in-game) time
  • Takes Damage to save time
  • Abuses death
  • Abuses programming errors
  • Counts time spent dying
  • One player controls two characters

Comments

Marble Garden 1, Carnival Night 2 and Launch Base 1 are all improved. Marble Garden 2 was redone to fix desynchs, with no actual improvements. There were also some minor corrections due to lag and luck differences.
According to Upthorn, luck, in this game, "is strictly due to frame rules in the universal frame timer". Short pauses can be added (or, in this case, removed) just before the level that desynchs to correct the issue.

Comparison Table

ActNew TimeOld TimeSavings
Marble Garden 10:30:290:37:210:06:52
Carnival Night 20:42:111:08:530:26:42
Launch Base 10:37:410:56:190:18:38
Total Savings: 0:52:12 in in-game time, or 0:46:24 in real time.
The table shows the in-game timer values. The value after the final colon indicates the in-game frame counter, and is restricted to the 0-59 range. It is updated every frame (except during lag frames), and increments the in-game second when resetting to zero.

Stage by stage comments

Marble Garden 1

Using a trick to get inside the spring by Mr Sweed, the level is cut short by several seconds. A few frames are lost because Tails can't be manipulated to reach the boss faster.

Marble Garden 2

While visually nearly identical to the previous run, this level has been redone almost entirely, using little input from the previous run, to correct desynchs due to different (sub)pixel position. The exception is the boss fight, which uses essentially the same input as the previous run.

Carnival Night 2

Using the shortcut from the 100% Sonic Run, but 42 frames faster thanks to Tails.

IceCap 1

The level has 3 less lag frames due to luck differences when Sonic is hit (the placement of rings).

Launch Base 1

Tails pushes Sonic inside the spikes, which put him inside the wall. No more waiting for that pesky Flybot! The trick is thanks to Orkal.

Sky Sanctuary

There are fewer frames "lost" in pauses to manipulate luck for Death Egg 1 at the end of this level.

Other comments

For more comments about the run, see the comments from the previous run, as well as the runs that it obsoletes.
I started the run from the previous run, keeping the rerecord count and splicing in the unchanged input as needed.
I wish to thank Nitsuja and Upthorn for making the previous run(s), as well as improving the tools for making runs, as well as Orkal and Mr Sweed for the tricks.

Possible improvements

If one were to replay the portions that desynched due to luck differences, instead of adding pauses to manipulate luck, some 30-35 frames might be saved in real time, but none in in-game time. At this time, I am unaware of any other improvements.

cpadolf: Accepting as an improvement to the published run.

upthorn
He/Him
Active player, Emulator Coder (387)
Joined: 3/24/2006
Posts: 1802
bombman wrote:
when angel island act 2 started why did you come to a full stop at the spikes past the starpost instead of rolling under them while charging for the spindash
The only way under those while they are extended is to glide under them as Knuckles. Also, ditto moozooh about reposting. It is pretty impolite, especially when your question was phrased so vaguely. After all, it doesn't actually say Act 2 on screen until a fair bit after the point you're talking about. And at first you didn't even specify act 2 of which level.
How fleeting are all human passions compared with the massive continuity of ducks.
Player (50)
Joined: 4/10/2009
Posts: 226
bombman wrote:
when angel island act 2 started why did you come to a full stop at the spikes past the starpost instead of rolling under them while charging for the spindash
if I understand what you're saying sonic is too tall to roll under those spikes, only tails can.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
Is this game particularly hard on the h264 format? Is the unusually large amount of visible compression artifacts to be expected?
Publisher
Joined: 4/23/2009
Posts: 1283
Warp wrote:
Is this game particularly hard on the h264 format? Is the unusually large amount of visible compression artifacts to be expected?
Yep, there is a lot of action and all the layers really makes it hard to encode.
Senior Moderator
Joined: 8/4/2005
Posts: 5769
Location: Away
Lots of dithering does contribute.
Warp wrote:
Edit: I think I understand now: It's my avatar, isn't it? It makes me look angry.
Joined: 11/4/2007
Posts: 1772
Location: Australia, Victoria
If we had the patience of a saint, and a few super-computers, we'd use one of them scripts that encodes at 1fps rather then 6fps. But I am sure Grunt, Aktan, Raiscan and I (Along with other active encoders) only have one computer for encoding with, otherwise I would have run this game through a more advanced script. Other TASs need encoding too you know. Long story short, if we wanted to keep the bitrate and filesize while increasing the quality, the run would have taken two weeks to encode.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
Maybe people could donate computing time for the cause? (IIRC this idea has been thrown in the past, but I don't remember the discussion)
Publisher
Joined: 4/23/2009
Posts: 1283
At the bitrate of the published file, I doubt it would still look good with the most intense settings, but then again, my definition of HQ differs from everyone else. The encode I posted in this thread was done at 1 FPS, and, to me, it is only borderline HQ despite having a huge filesize.
Post subject: Raidisencode
Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
Flygon wrote:
But I am sure Grunt, Aktan, Raiscan and I (Along with other active encoders) only have one computer for encoding with, otherwise I would have run this game through a more advanced script.
I have 4 Computers at my disposal, and I've been trying to work on a distributed system in which chunks of the capture are sent to each of them to encode. I had only mild success, the biggest hindrance being I couldn't work out a good way to stitch them together without doing it at the container level (i.e, having 20 or so different chunks of the movie playing one after another).
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
Post subject: Re: Raidisencode
Publisher
Joined: 4/23/2009
Posts: 1283
Raiscan wrote:
Flygon wrote:
But I am sure Grunt, Aktan, Raiscan and I (Along with other active encoders) only have one computer for encoding with, otherwise I would have run this game through a more advanced script.
I have 4 Computers at my disposal, and I've been trying to work on a distributed system in which chunks of the capture are sent to each of them to encode. I had only mild success, the biggest hindrance being I couldn't work out a good way to stitch them together without doing it at the container level (i.e, having 20 or so different chunks of the movie playing one after another).
Hmm, what's wrong with stitching them together at the container level? For some longer runs, I usually do split the movie in half to max out both cores and then combine them. If chosen at the right points like a scene change, the output is usually very close to not splitting at all.
Post subject: Re: Raidisencode
Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
Aktan wrote:
Hmm, what's wrong with stitching them together at the container level? For some longer runs, I usually do split the movie in half to max out both cores and then combine them. If chosen at the right points like a scene change, the output is usually very close to not splitting at all.
Some people like to demux or transcode these things for use on another device. I am also unsure how some of the streaming sites react to multiple chunks (though I haven't tried). I also had problems where with a large number of chunks, the audio desynced, though I'm not sure why. Regardless, I've put the project on hold and will resume it at a later date (I have a programming to-do list at the moment) One thing that made me lose interest is the lack of a working h264 encoder for the PS3. At the risk of derailing this thread further, if anyone knows of one, let me know :)
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
GGG
Banned User
Joined: 12/11/2009
Posts: 57
Location: USA
I've seen the publish page of Sonic 3 and Knuckles and had not seen any page but if I saw a video on youtube without parts. http://www.youtube.com/watch?v=fkQ0WaLFF94 I have also seen one with three parts uploaded by Aktan in DailyMotion Part 1: http://www.dailymotion.com/video/xb9l16_genesis-sonic-3-in-321938-by-nitsuj_videogames Part 2: http://www.dailymotion.com/video/xb9jjv_genesis-sonic-3-in-321938-by-nitsuj_videogames Part 3: http://www.dailymotion.com/video/xb9iqk_genesis-sonic-3-in-321938-by-nitsuj_videogames
Joined: 11/4/2007
Posts: 1772
Location: Australia, Victoria
This is due to the fact that any attempts of uploading Grunts encode to DailyMotion failed, that's why there was a lack of streaming media link other then Archive.org. In fact, I actually linked to Aktan's encode inside the description but not the publication simply for consistency sake. However, I'll add the YouTube link now, thank you.
Joined: 7/19/2010
Posts: 1
Nice run... One suggestion though. Wouldn't collecting less rings or taking damage at the boss to lose rings in place of Tails delivering the final blow save a few frames when scoring? Prime example, GHZ 1 finished with 76 rings and it took around a second to decrement those 7600 points to zero. As long as Tails is hitting whilst Sonic is in recovery, little time should be lost in favour of how much time could be saved.
Joined: 7/2/2007
Posts: 3960
The Sonic runs on this site typically ignore the bonus tally, aiming instead to optimize the amount of time used to play the actual levels. Paying attention to the bonus tally would, on many levels, mean standing in front of the goal line until the time bonus is less than 50000 points. As a result, rings are only ditched when doing so doesn't cost any time.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Player (50)
Joined: 4/10/2009
Posts: 226
while everything that derakon says is true, a better answer would be that since the time bonus and ring bonus tally at the same time, and the levels in this run are done so fast the ring bonus is never an issue. for instance if you beat a level in 1:29 (giving you a 10,000 point time bonus) you'd need to have over 100 rings before having rings would make the bonus tally take any longer.