Topic by JXQ

Posted: 11/19/2005 12:03 AM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

Yeah I'm going to hold off a bit on Marble 2 because I hope that the lava path can be improved as well. I didn't get much time to work on it today but I'll try more tomorrow. I'm also going to see if the ending path I took is possible at the end of your v2 run. If it's not, then that may be another piece of randomness puzzle that needs to be considered. Edit: It's not possible from your v2 run, Sonic is one pixel (I think) short of the jump onto the spiked chandelier thing. On an earlier post you mentioned that you could get the moving objects to appear in a four-frame window or something like that; this is probably a similar effect of reaching different areas of the lava part at different times. Daniayaw: I don't understand the question...are you referring to the ROM, emulator, or something else? I'm using the ROM "Sonic The Hedgehog (JUE) [!].bin" and emulator Gens Test 9f.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/19/2005 12:36 AM

Quote

daniayaw

Former player

Joined: 7/29/2005
Posts: 459
Location: Brazil

there is a REV00 and a REV01 rom version^^ well, so, that's why.. i dont know what the difference between both, i thought you knew and used REV00 for any reason.. but its ok then

My big signature was cleared by admin; i should read <a href="http://tasvideos.org/ForumRules.html">forum rules</a>. But... who does?

Posted: 11/19/2005 1:00 AM

Quote

Quietust

He/Him

Emulator Coder, Former player

Joined: 7/14/2004
Posts: 250

Er, the only Sonic ROM I'm aware of having multiple revisions is "Sonic Compilation"...

* Quietust, QMT Productions P.S. If you don't get this note, let me know and I'll write you another

Posted: 11/19/2005 1:41 AM

Quote

daniayaw

Former player

Joined: 7/29/2005
Posts: 459
Location: Brazil

im talking about "Sonic the Hedgehog" ^^

My big signature was cleared by admin; i should read <a href="http://tasvideos.org/ForumRules.html">forum rules</a>. But... who does?

Posted: 11/19/2005 2:51 AM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

daniayaw wrote:

im talking about "Sonic the Hedgehog" ^^

I think Quietust knew that. But there is a ROM known as "Sonic The Hedgehog (W) (REV 01) [!]" which is different from the one JXQ is using, and a ROM known as "Sonic The Hedgehog (W) (REV 00) [!]" which is identical to the one he's using.

Posted: 11/19/2005 2:55 PM

Quote

daniayaw

Former player

Joined: 7/29/2005
Posts: 459
Location: Brazil

thats what i said o.o in the previous post i mean to say that im not talking about Sonic Compilation, im talking about Sonic the Hedgehog^^ which is quite different

My big signature was cleared by admin; i should read <a href="http://tasvideos.org/ForumRules.html">forum rules</a>. But... who does?

Posted: 11/19/2005 4:53 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

I did some more comparisons with that lava segment, and here's what I get: I am fifteen frames slower when falling down to the platform just before jumping with the lava, but I have a bit more velocity because of different jumping, I think. This causes me to gain back 11 of those 15 frames by the time we both hit the next wall. So, if the original lava bounce works, it should save at most four frames. I could be wrong on this, so if anyone else wants to compare, feel free. Four frames or not, the original way still looks cooler, so I dove into some Marble Zone 1 lava fun for a while, but still no success at this point. I got bored with that for now and did Marble Zone 2. The WIP is here. I'm not saying Marble Zone 1 will stay the way it is; I have no problem redoing levels, and I've gotten pretty good with a hex editor since Super Demo World, so redoing may not even be necessary as much as cut-and-paste.


Marble Zone 2          Xebra    -4.01    JXQ      Diff
-------------------------------------------------------
First jump             3.37.52  3.33.51  3.31.09  2.42
1st switch             3.42.49  3.38.48  3.35.56  2.52
Past lava              3.46.56  3.42.57  3.40.03  2.54
Done pushing           3.50.51  3.46.50  3.44.00  2.50
Push right             3.53.54  3.49.53  3.47.03  2.50
Edge before swings     3.57.23  3.53.22  3.50.33  2.49
On platform (all)      4.00.13  3.57.12  3.53.28  2.44
2nd switch             4.02.05  3.58.04  3.55.19  2.45
Break block            4.16.24  4.12.23  4.09.38  2.45
Land on lava block     4.21.54  4.17.53  4.15.00  2.53
Highest step           4.24.20  4.20.19  4.17.30  2.49
End                    4.31.58  4.27.57  4.24.54  3.03

I gain some frames in the beginning, but lose them in the area with the swinging platforms because they were positioned differently and I had to slow down a bit (which I did by jumping lower). I gained 21 frames in the last segment of the level. I'm not positive where, but I think that avoiding hitting the spiked-worm enemy near the end helped. I also tried waiting on the last spiked chandelier to jump directly to the top grass, which worked, but was slower. The area after the second switch has some real rapid-looking jumping. I did this because in some cases, holding the jump button longer keeps your speed just slightly faster, and that gained me a frame here or there. Finish time is still 0:53.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/19/2005 10:47 PM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

JXQ: Try playing your movie until the frame counter says 2:37:48, then stop playing the movie and hold the jump button without any directional buttons. You'll go through the next section much faster (at least 11 frames faster is possible for that very short section, if you wait enough time before holding right again). Maybe in this case it wouldn't actually help because there are timed crusher objects coming up after it (then again I don't know how timed objects work in this game, maybe reaching them earlier affects them), but look out for this in other places with steep hills / similar setups.

Posted: 11/19/2005 11:15 PM

Quote

Highness

Player (70)

Joined: 8/24/2004
Posts: 2562
Location: Sweden

I will probably check the WIP tomorrow and try to give some input on it.

Posted: 11/20/2005 12:23 AM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

Nitsuja: I'm unable to reproduce what you are telling me. If I play the movie until the counter says 2:37:48, then stop movie, hold B and frame advance, Sonic doesn't jump, instead he lands on the first floating piece of land, and I have to jump from there. If I press jump on 2:37:47, then Sonic makes a really big jump that almost reaches the third floating piece of land, but pressing directions in the air doesn't extend this jump. I tried avoiding spinning, but I still can't make the jump, and any adjustments I've tried to make the jump gets me to the shield at 2:41:14 or later, which is a 10 frame delay. Any idea what I'm doing wrong?

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/20/2005 1:21 AM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

JXQ wrote:

Any idea what I'm doing wrong?

I must have changed the timing of your jump down the hill beforehand without realizing. Here's a GMV demonstrating what I meant: http://www.filespace.org/nitsuja/sonicmarblehugejump.gmv.zip Everything before 2:36:48 (possibly a bit later) is the same. Note that I didn't test all the possible frames of jumping down the hill or what effect doing more rolling would have or when to jump after landing, so it could probably a little faster. Note also that the way I did it made the pillars actually 2 frames slower than for you, but maybe it's bad luck from the frame timing I happened to end up with.

Posted: 11/20/2005 10:44 AM

Quote

xebra

Joined: 5/3/2004
Posts: 1203

That's a nice find, nitsuja. JXQ, your handling of the pistons and the falling off the block you need to push are not optimal. For one, you shouldn't land on the block at all. If you perform a series of smaller jumps towards it you can fall right into place to the left of the block (and it's faster.) If anyone has a copy of my WIP from when Ouzo and I were messing around with Marble 1, the strategy is in there. My comp is out of commission at the moment, so I don't have a copy.

Posted: 11/20/2005 4:57 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

I redid Marble Zone 1 with the new improvements, thanks to Nitsuja & Xebra. Unfortunately, the improvements end up hurting the time later on, as I can't jump to the spiked chandelier as I did in my first version. The techniques gained 16 frames in total, but it is irrelevant because of the timing of the moving platforms near the end. In this run, the platform on the right is visible on frame 3.06.46, three frames later than my first version. By messing around near the lava, I was able to make it appear on 3.06.45, but that's still slower than the first version. Honestly, I don't have the patience to try different pauses all over the level to see where the platform can be manipulated to the same time as my first version because then the frame count would just be the same as what I already have. I also tried doing the original lava bounce with this new timing for quite a while and was still unable to do so. However, I'm pretty sure that these ideas will have time-saving applications in later levels when there is less randomness to worry about. Anyone who's curious to watch or mess around with this can find Marble 1 partially redone here.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/20/2005 7:30 PM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

JXQ wrote:

Unfortunately, the improvements end up hurting the time later on, as I can't jump to the spiked chandelier as I did in my first version. The techniques gained 16 frames in total, but it is irrelevant because of the timing of the moving platforms near the end.

It looks like you got terrible luck with the first set of pistons coming down - it must be faster to manipulate the first one so you can just barely get through it when you get to it (unless I'm wrong that it's randomized, but it appeared to be when I tried it). And the same for the chandelier, I think just changing the frame it gets loaded affects its timer. It might not matter if even saving 15+ frames wouldn't change the level time, but that still seems like a lot of frames to abandon considering how the other levels are being handled.

Posted: 11/20/2005 8:03 PM

Quote

xebra

Joined: 5/3/2004
Posts: 1203

JXQ, how many frames improvement is the new WIP versus my version 2 video? When Ouzo and I were messing around, I got (I think) 50 frames of improvement from the beginning of the level to pushing the block. Also, why do you insist on using that bad lava jump sequence? In my v2 video, if you can just do that 1-2 frames faster, you will knock 5-6 seconds off the run. Also, I don't even think your slow lavajump sequence is optimal. When you jump down into the shaft, wouldn't it be better to continue on into the wall of the step you are landing on for instantaneous deceleration? You start slightly farther back, but I'm 99% sure it's faster than decelerating yourself, because I've used the strategy in testing other parts of other levels. Nitsuja, the pistons appear in the same position, always, whenever you reach a particular point in the level, I believe. In my WIP that no one has a copy of, you would have seen me roll and jump in a really weird spot (the second to last burning platform, I think) that slightly accelerates Sonic for a few frames, but slows you down a buttload while you make up the lost speed when you land. That was done to spawn the pistons 4-5 frames early, and the lost time is meaningless since you can never slip under the second one.

Posted: 11/20/2005 9:31 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

The new WIP vs. your v2 is 28 frames faster to the block push. About the lava jump, it's not that I insist on using the bad one, but I haven't gotten the good one to work once on my run. Even if I did, and I got it a few frames faster than your v2, I don't see how I would save 5-6 seconds. Do you mean if I can get to the moving platforms early enough to jump on the first one as it is moving into the wall? That would definitely save some time if it still allowed jumping onto the top of the spiked chandelier. I will also test the deceleration you described while jumping down the shaft. (Edited for clarity) Unless the position of the platforms can be manipulated to our advantage, all the timesavers in the world aren't going to improve the time, they will just make for more waiting near the platforms. Any time gained before the platform is loaded into memory is only good if it doesn't mess up the position of the platforms, and to get this 5-6 seconds you are talking about (if I'm interpreting correctly), enough frames must be gained after the loading point of the platforms, wherever it is.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/21/2005 6:34 AM

Quote

xebra

Joined: 5/3/2004
Posts: 1203

JXQ, don't bounce off of anything in Marble 1 prior to the pistons, it's at least 5 frames too slow. For the first bee, you can either miss it entirely or fall through it without bouncing. For the second caterpillar, you can just jump straight over the caterpillar and the bee directly onto the raised platform before the long slope.

Posted: 11/21/2005 3:00 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

I redid the beginning and had some trouble figuring out exactly how to correct the part. It seems that Sonic slows down a bit from jumping, or bouncing, but not in every case. Xebra, do you know of a good way to determine this, or is it just something to test on each jump? I did improve the beginning a bit more messing around with different jumping strategies, but again, the platform in the wall was coming out later than in my original run. I'm going to leave this level for now - these improvements that end up not improving the overall level are frustrating me, and I think my energy is better spent on upcoming levels. When I come back to this level, I will try and figure out where the two spiked chandeliers and moving platforms are loaded, and figure out a "timing map" of sorts to play the level by.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/21/2005 7:04 PM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

JXQ wrote:

I'm going to leave this level for now - these improvements that end up not improving the overall level are frustrating me, and I think my energy is better spent on upcoming levels. When I come back to this level, I will try and figure out where the two spiked chandeliers and moving platforms are loaded, and figure out a "timing map" of sorts to play the level by.

Well, do you know if this game is hex-edit-friendly enough for you to be able to return to this level without redoing everything after it? I don't think it will become any easier to work on it later than it is now, in any case. But maybe you're right to take the best you've been able to get and move on, as there's no complete TAS of this game yet and you could try improving on it later...

Posted: 11/21/2005 9:05 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

I haven't done any hex-editing tests yet, so I'm not sure if it would be easy at this point. I don't have a problem redoing levels, however, if improvements are found in earlier parts of the game. To be clear, I'm all for putting effort into improving Marble Zone 1 as a whole, but only if it's actually going to improve the entire level. What I'm not all for is improving Marble Zone 1 up to that point just to match the time I already have, because it doesn't look sloppy to me (except for the lava jump, which I haven't been able to do once), and also because every improvement I've tried has actually caused a 2-second delay later in the level. I have tried getting to the platform in a lot of goofy ways with delays in different spots and the earliest I can get it to appear is 3.46.03, as was in my first version. Getting on the platform on the way in is at least 40 frames away from possible by the most recent improvement, and these 40 frames must be gained after the platforms are loaded, which at the earliest is just before the pistons. Even if the other lava jump was pulled off, there would still be over 30 frames to gain in a very small segment that is already quite optimized, and that does not seem possible to me. Nevertheless, I'm still willing to try improvements, because I am not the most knowledgable about this game, but at the same time, I won't stay on this level forever trying what seems impossible. I hope I don't seem lazy or like I'm rushing this; that's not my intention. And I also don't mean to seem ungrateful for the improvements, because learning the game more now can help me in later levels. I just want to spread out my energy on this project efficiently.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/21/2005 11:45 PM

Quote

xebra

Joined: 5/3/2004
Posts: 1203

In my experience the game is not hex edit friendly, which is why I am kind of wary of you just moving on. I haven't tested this extensively, but for example, two hex edits that just don't work are trying to paste my Marble 1 v1 or Marble 1 v2 onto my v3 run. This, alas, is one reason I quit the run entirely. I didn't want to redo Spring Yard 1 and 2, and I especially did not want to redo Labyrinth 1. Also, I think it's just a little ... I dunno, undedicated/sloppy/uncaring not to want to optimize Marble 1. There is a way to do what I want to do, I guarantee it. Another reason I stopped the run is because I wasn't willing to half ass it. I didn't suggest that you pick it up in my stead only to half ass it for me. That's why I may seem disappointed with what I consider to be poor choices on your part, whether they be concerned with style, sloppy play, or poor path optimization.

Posted: 11/22/2005 12:11 AM

Quote

Ouzo

Former player

Joined: 8/15/2004
Posts: 422
Location: Minnesota

Xebra, I'd say if you really want it done right, you should just do it yourself. I know it's a pain in the ass to optimize this, but spending time to watch and criticize other people's attempts instead of working on it yourself (which you've had more than a year to do so) seems rather uninspiring. Maybe you can create and compare movies with JXQ instead of just telling him what he's doing wrong. I wouldn't call JXQ's attempt half-assed, since he did gain 47 frames in the Green Hill Zones compared to your v2 (which I had thought to be perfect). Though, I do agree JXQ should spend some more time on MZ1, despite being a level from hell. JXQ: I hope you keep this up, as it has been looking wonderful since you started.

Posted: 11/22/2005 2:52 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

Xebra, to be clear - is this the ideal run for Marble 1? 1) Optimized beginning up to block push 2) Lava jump from your v1/v2 3) Moving platforms jumped on early as possible (earliest I've found so far is emerging from wall at 3:06.43) 4) Able to jump onto spiked chandelier immediately as in my first version If I am missing something, please correct me. This scenario shares some of the frustrating properties of Super Demo World. In that run, entering levels on different frames (starting states) can subtly change the amount of lag in a level. This means some levels were improvable from that standpoint, if it had been tested to enter and play through them with many different starting states. But...I'm only human. And as you were saying in a WIP you had worked on, you slowed down in a certain part which loaded pistons at a different spot. Perhaps some of the places that are not fully optimized in my first run helped me unknowingly set up timesavers in the later parts of the level. I'm not saying that means it should be kept, because it could still be improvable, but it may mean that your ideal run is not possible. I just don't know. I plan to test it more (as I tried some hexediting myself and was unsuccessful) before moving on, but whatever the fastest run I end up with after my patience is done is what I'm going to go with.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)

Posted: 11/22/2005 6:47 PM

Quote

nitsuja

Emulator Coder, Skilled player (1300)

Joined: 12/21/2004
Posts: 2687

JXQ: I find it difficult to believe that you got so lucky the first time you ran it that even 1 more frame of improvement anywhere always causes a longer delay later, and even if that's the case, it should at least be possible to add in the right delays to get the same time you got before (that's why I feel not enough options have been tried to know that it can't be improved, because you haven't even gotten the same time as you got before by different means). So... good luck with figuring this out, anyway. BTW, I'm starting to wonder if I understand what this "lava bounce" is - I remember seeing Sonic hit the lava and lose some rings in this stage in one of Xebra's WIPs, but it didn't seem especially faster than what you're doing now so maybe I'm not thinking of the same thing. I also find it difficult to believe that whatever the stunt is can't be reproduced, unless maybe it depends on how many rings you are holding and you simply don't have as many rings as Xebra had before.

Posted: 11/22/2005 7:45 PM

Quote

JXQ

Experienced player (750)

Joined: 5/6/2005
Posts: 3132

nitsuja wrote:

JXQ: I find it difficult to believe that you got so lucky the first time you ran it that even 1 more frame of improvement anywhere always causes a longer delay later, and even if that's the case, it should at least be possible to add in the right delays to get the same time you got before

The 1 frame of improvement may not cause a delay, but if it doesn't load the moving platforms earlier, it won't give any net improvement either.

I remember seeing Sonic hit the lava and lose some rings in this stage in one of Xebra's WIPs, but it didn't seem especially faster than what you're doing now so maybe I'm not thinking of the same thing.

There is a difference of a few frames - mine was slightly slower because I landed to the left of a block while Xebra had open space to the right, so the first few frames of his jump allowed for more horizontal acceleration.

maybe [the lava bounce] depends on how many rings you are holding and you simply don't have as many rings as Xebra had before.

I think you are right, except it looks like I had too many. I watched Xebra's v2 and he had 2-3 less rings than I did, and so I kept that in mind and was able to do the better lava bounce for the first time on the first try. Hopefully that problem stays solved. Busy today, but I have lots of time tomorrow so I will try to hammer out a new Marble 1.

<Swordless> Go hug a tree, you vegetarian (I bet you really are one)