As always, I have provided a release package on my my googlecode project.
It includes the movie file, submission text, screenshots, a spreadsheet of frame counts, and Ram Addresses.
You know Mario can't let his sidekick take all the glory!
IMPORTANT: This uses DeSmuME 9.6 with Advanced Bus-level Timing OFF (Config -> Emulation Settings). It will not sync without this setting off.
This is my long overdue completion of v2 that I started soon after my 1st publication. I took a lot of time off due to luck manipulation issues World 5. During that time, TRT finished his movie. Most of the content up to 5-G was done prior to his movie (but map sections edited to sync with 9.6 instead of 9.2).
Because TRT's movie is the currently published, I will compare my improvements to his movie.
This movie is 11918 frames (3 minutes 18.6 seconds) shorter than TRT's published movie. Most of this, however, is due to Desmume9.6 having significantly less lag than 9.4. It is hard to estimate the true improvement due to some improvements in the map movements I made. However, the in-level improvements total 723 frames (12.05 seconds). I had overall better luck on map movements too, but I have no idea how to calculate a lag-reduced value. I don't value map manipulation improvements much anyway since the entertainment value of this game is with the levels.
Some tricks not used in previous published movies:
Sliding at the bottom of a hill
By running down a hill and sliding just before the bottom, you can get a speed boost. If you hold down you can continue to get extra speed for many frames after the hill. This is used in 1-1 and several levels in world 8
Fast vertical swimming
By alternating B and X for swimming, you can retain 31 vertical speed as opposed to turbo firing one of them which results in an alternating speed of 31 & 30. It also goes so fast that the swimming sound is canceled out (which is kind of pleasant compared to the sound of turbo fire swimming).
Most improvements come from various strategy improvements, more corner boosts, and various optimizations.
Also, not so relevant overall, I must mention, the luck manipulation for getting the 2nd blue shell for 8-3 is ABSOLUTELY EVIL. I spent about as many hours getting good luck there as making the rest of the TAS. You have so many elements working against you for luck manipulation: The previous level is a boss level so no flag, you are mini-mario so no killing enemies to change the RNG, the previous level is the most brutal in the whole game to optimize, and the hammer bros appears immediately in 8-3 so nothing in 8-3 can affect what he drops. I think TRT mentioned struggling in this spot too. Evil :|
Joined: 10/28/2007
Posts: 1360
Location: The dark horror in the back of your mind
The map screen lag reduction makes this worth publishing by itself, but the rest of the gameplay is solid, too.
I think this should be published, but I also think that waiting for the proper 0.9.6 release is a good idea (as I understand it, this hasn't been done only because the person responsible for 64-bit builds hasn't made a build yet).
Nevertheless, yes vote.
Always nice to see a new Mario submission. :)
What was that weird delay right before the flag in 5-1 for?
I assume that getting the top of the flagpole is slower than hitting it partway up, and that spinning in the shell is slower than constantly jumping. That latter would also explain why you lose the blue shell -- so you can run on the moving platforms in the castle.
I also assume that getting Tiny Mario and his improved walljumping into the penultimate level is impossible, or at least not helpful. Is it not possible to get him to the checkpoint, die, and then come back at the checkpoint with the wandering platform ready?
All in all, this is everything we've come to expect from an Adelikat submission. Nice work!
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Emulator Coder, Site Developer, Site Owner, Expert player
(3600)
Joined: 11/3/2004
Posts: 4739
Location: Tennessee
Derakon wrote:
What was that weird delay right before the flag in 5-1 for?
To avoid a certain time on the timer that would cause fireworks. Some extra frames were also spent there for some map screen luck manipulation.
I assume that getting the top of the flagpole is slower than hitting it partway up
Anything under the flag itself is going to be equally as fast. So anything up to 1000 of a score.
and that spinning in the shell is slower than constantly jumping.
Spinning with the shell is 44 speed. Constantly jumping is 48.
That latter would also explain why you lose the blue shell -- so you can run on the moving platforms in the castle.
Among other reasons, yes. I lose it in that level because that is another level where I must wait for the timer. And the timer runs while taking damage. Thus most of the time loss is negated. The benefits of regular mario aren't just for moving platforms. It is for corner and super-corner boosting, and sliding as well.
I also assume that getting Tiny Mario and his improved walljumping into the penultimate level is impossible, or at least not helpful. Is it not possible to get him to the checkpoint, die, and then come back at the checkpoint with the wandering platform ready?
Unfortunately mini-mario is no match for that level. Even getting to the check point isn't possible IIRC. I think TRT even tested this and found it would be significantly slower anyway. I'll be honest though, I didn't look into this concept with much depth.
Watched it. Liked it. Yes vote. I had a feeling an improvement over my not-so-optimized run was due especially with the approach of v0.9.6.
I remember the agony that was 8-Tower. My result looked sloppy, unfortunately :(.
Just three question though.
1) 5-2 has Mario not facing left all the way. Is that for luck manipulation?
2) 5-3 has Mario land on an upward slope. Does he lose speed for that one frame?
3) How were you able to compare your run with mine effectively (since I remember my run desyncing quite often for many people)
Anyways, this certainly deserves to be published :P.
Emulator Coder, Site Developer, Site Owner, Expert player
(3600)
Joined: 11/3/2004
Posts: 4739
Location: Tennessee
TRT wrote:
I remember the agony that was 8-Tower. My result looked sloppy, unfortunately :(.
GOD, I hate that section.
5-2 has Mario not facing left all the way. Is that for luck manipulation?
Probably not, but it apparently doesn't lose time. I was going off frame count for level endings and I had no loss of time there.
2) 5-3 has Mario land on an upward slope. Does he lose speed for that one frame?
1 frame on an upward slope is going to be a cost of about 1-2 speed. 1 speed is 1/16 of a pixel, hardly enough to cost a frame anywhere.
3) How were you able to compare your run with mine effectively (since I remember my run desyncing quite often for many people)
I got your run to sync, though it was tricky in a few spots. Just had to match your firmware settings, use the right emulator version, and some other issues I think. But for the most part, I didn't have trouble.
Anyways, this certainly deserves to be published :P.
Joined: 9/2/2008
Posts: 289
Location: United Kingdom
That latter would also explain why you lose the blue shell -- so you can run on the moving platforms in the castle.
The other reason why the runner loses the blue shell is because you can instantly change into a Fire plumber right after picking up the Fire Flower, otherwise it would go into the reserve slot (see 5-Ghost).
Very nice run, this deserves a yes vote, M.Bision style.
In before the Weegee fanboys cry for not using Luigi on this run :P.
Joined: 8/26/2006
Posts: 1139
Location: United Kingdom
Voting yes. The lag reduction makes a huge difference to the entertainment factor here. This is the first run of this game that I have particularly liked. Also, this game has so many beautiful levels and worlds that an all-levels run of some sort should be considered.
I've watched it 3 times already. First two times was because it was just fun to watch. Third time was to look at the run in more detail and look for any improvement that can be seen at first glance. So far, everything looks optimized.
I must say that I agree that lag reduction makes it more enjoyable (and make TASing it more enjoyable as well). There were so many other subtle improvements in the levels that I surprised that it is only a 12 second in-level improvement.
adelikat wrote:
I also assume that getting Tiny Mario and his improved walljumping into the penultimate level is impossible, or at least not helpful. Is it not possible to get him to the checkpoint, die, and then come back at the checkpoint with the wandering platform ready?
Unfortunately mini-mario is no match for that level. Even getting to the check point isn't possible IIRC. I think TRT even tested this and found it would be significantly slower anyway. I'll be honest though, I didn't look into this concept with much depth.
I haven't tried mini-Mario in that level, but I like the idea of mini-Mario walljumping to the checkpoint.
To be honest, I went through the checkpoint shortcut idea quickly pretty quickly as well when I was making the movie (few times on v0.9.4 which is laggy and few times on the actual game). I wasn't exactly able to do an accurate comparison for the two approaches, though.
By dying at the checkpoint, you gain time by skipping a portion of the platform's route. However, you lose time from map movement, the loading screen, and power-up times. In 0.9.4, this turned out to be slower. I haven't tested it in 0.9.6 yet, but I think that less lag from 0.9.6 certainly makes it an option to consider in the future.