That was really interesting. The most shocking fact, of course, was that you almost caught up to the bat TAS.
If you optimize it, skip Richter and make cleaner per-room toss routes, you may actually beat the bat run. Wanna do it?
In regards to the vid I posted above, it was played on the PS2 version of DoDonPachi DaiOuJou, which is actually a very manipulable game. If you set the practice mode (which is how it was played) parameters in a certain way and keep using them to program the key input, it will likely not desync much enough to not let you record a complete video at least once.
Here's an example of another emulator-less TAS on a PS2. Made by Morimoto himself, if titles are to be believed. He used "SuperCommand Controller" to program a path between the bullet sprays, abusing all the blind spots on the way. Yeah, some people dodge that shit unassisted.
The last known 1-player version was played on the U ROM, and it's only a second slower. While switching to J ROM saves about 22 seconds by itself. That's not to say that the current published run uses 2 players and is about 35 seconds faster.
You're right that 640x480 carries more pixel information. However, the original claim was, citing, "using a larger resolution with the same bitrate does not decrease the quality of the video". Which was something that had not happened in the example. It was the reason I compared the videos at lowest common denominator resolution, where pixel information and quality of scaling couldn't be factors in comparison.
There is more to that, though. Your example was a fragment from anime, with very little details in most frames. It concealed some of the compression artifacts which would inevitably have appeared if most of the image wasn't made from gradients and/or entirely monochromatic areas. 2D (and software-rendered 3D) videogames, however, carry useful information at pixel level most of the time, so a 4:1 increase in information doesn't make it "more of the same". Which is why it is best to encode videos at a resolution the main amount of pixel graphics is displayed in: pixel graphics scales horribly, because it's integer by nature. Any attempt to resize it by a non-integer factor will inevitably produce blurring and rapid image degradation.
There's an option that's supposed to calculate the amount of added noise objectively (as close to it as possible, anyway), it's called PSNR. Add :psnr to your mencoder string and compare the values.
Btw, the 640x480 encode looks slightly worse at 320x240, as expected.
When you climb down from your cross, note that it was you who:
a) suggested something from the height of your experience;
b) argued that Johannes (who has already proved his experience with public encodes) was wrong, while the only proof of your claim was your experience; and
c) never showed anything to back up your claims of experience.
Do you really believe it's alright to claim that people are wrong without proving it and expect everyone to take your words for granted? If it is so, I'm afraid you need to grow up a bit.
Yes it has. Now you're not just storing the information about original pixels, you're also storing the information about interpolated pixels. The amount of useful information has not increased, but there's redundant information now, which also has to be stored. The only reason it doesn't require proportionally higher bitrate is that the information density is lower now. (Similarly, downscaling a picture to 1/4 of its resolution will commonly require a higher bitrate than original/4.)
May I require an example, together with resulting video files? That will alleviate all the possible confusion (and don't you dare cop out after starting this debate).
It's searching for shapes in raster image. Because the image is still raster, not vector graphics. They are absolutely incomparable in efficiency, speed, compression ratio, scaling quality, everything, and no matter how advanced motion recognition algorithms are, they won't ever be close to actual vector graphics.
Warp, what you're talking about might not be a quality increase due to upscaling, but a quality decrease due to improper stretching of a downscaled video to fullscreen. If you use common sense you'll see that 320x240=76800, while 640x480=307200 pixels, which is 4 times more pixels shown on each frame of a video. No, of course it doesn't explicitly tell that such a video will necessarily consume four times more bitrate, it just means the codec has to compress 4 times as much data overall. There's no way it's going to be more efficient than otherwise, regardless of your experience. Math doesn't work on personal level.
Now there are multiple ways of upscaling a low resolution video on decoding level, which may or may not look better than using an encoding filter to upscale one.
It's very cool of you, actually.
By the way, what would you think about doing a glitchless S3&K run with Sonic solo? That kind of run has been in high demand since who knows when, but Upthorn isn't up for it.
I would have written it even better had I proofread the result once more. There are a few typoes here and there, as well as a couple things that could be worded differently.
But hey, thanks. :D
Following the recent upsurge in emulator development, there are some interesting news on the PS2 front as well: apparently, development teams of the two most advanced PS2 emulators have merged to form one solid program, which will assume the name PCSX2.
If you check the compatibility list on their site, you'll note how many various games are already playable using the recent versions. In fact, it is not unlikely that we receive tool-enhanced PCSX2 already in 2009 if we ask nicely. (AngerFist? :D )
All in all, it seems the 2009 is going to be a floodgate opening year for tool-assisted speedrun platforms, kinds of like 2004 has been.
[EDIT]
So, AngerFist convinced me to create a feature request on the project's googlecode page. Which means you can go ahead and star it:
http://code.google.com/p/pcsx2/issues/detail?id=16
Thankfully, it only continues for less than 4 more minutes.
I'm not attacking Comicalflop, and thus not sure why you even need to support him in the first place. Also, are you arguing the entertainment value I derive from different games? Just don't.
Lastly, the notion of Mischief Makers as "more of a platformer" is rather ironic in the light of the amount of time actually spent interacting with, you know, platforms (rather than just sliding right, flying over obstacles altogether, or doing some unrelated things).