1 2 3 4 5 6 7 8 9
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Very interesting MUGG, some might have some use, especially that weird goofy walking around 1:54 to 2:00. I haven't worked on this much in the past week, but I'm still planning on playing around with it some.
Living Well Is The Best Revenge My Personal Page
Player (119)
Joined: 1/30/2007
Posts: 82
Ah, I must remember to check these boards more often. I watched your WIP and as a matter of personal preference, I think TheAxeMan mostly had better stylistic choices than your run did (I would probably be one of the few to notice such nuances, however). Then again, I think the run would look best if Mario was airborne as much as possible though even though most would find that repetitive and annoying. On another note, I found that Mario behaves unusually when auto fire down is used so it could make for some interesting downtime and in 3-2 it is one frame faster to get the second mushroom before the pipe cannon (either that or TheAxeMan lost a frame somewhere else). Other than that I can't think of any more improvments nor have I resolved the bonus screen mystery. Anyway, good luck with your run if you so choose to continue.
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
Btw, I made a small demonstration of how a run of the hard level could look like (I didn't care about bad bonus screen set-up, though). http://dehacked.2y.net/microstorage.php/info/147986503/theaxeman-sml%2Bmugghardleveldemo.vbm You'll have to fast-forward as I kept TheAxeMan's run in instead of starting from savestate. The run ends on the submarine level.
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
I have been making some progress on this, and while I haven't figured out the method behind the madness that is the bonus screens, I have discovered that the WIP that Xenos posted before managed to make the first three stages (1.1-1.3) as close to perfect as I can find. By delaying 4 frames in level 1 (as counted by sub-second RAM clock that determines when the main clock decreases by 1 second), he was able to maintain a perfect 1.2 and 1.3. The 46 frame rule seems to only have an affect on timing in 1.3, and even delaying one more frame in that stage will cause it to activate. But by delaying those 4 frames in level one, the frame rule is averted, while still producing the best possible bonus stage set-up. Xenos, thank you for blindly stumbling into the most optimal way possible of doing the whole freaking first level O_o. However, this means that the subsequent levels may be an aggravating sequence of trial and error, to see if delaying x number of frames in the first stage will produce the most beneficial bonus stage. But hell, at least there's progress. As for entertainment value, I prefer to have Mario take as little excess movement as possible, I think it makes things look a lot more sharp and focused, instead of making Mario look like a kangaroo by jumping whenever he can. But I am willing to change that if the audience thinks it's more entertaining. So, to anyone following this thread, which method would you like better?
Living Well Is The Best Revenge My Personal Page
Player (120)
Joined: 2/11/2007
Posts: 1522
A happy medium: not constant but you should try to interact with stuff where it doesn't lose time, like in SMB where random 1up blocks are hit etc. But you better go all out on the autoscroller stage :)
I make a comic with no image files and you should read it. While there is a lower class, I am in it, and while there is a criminal element I am of it, and while there is a soul in prison, I am not free. -Eugene Debs
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
alden wrote:
A happy medium: not constant but you should try to interact with stuff where it doesn't lose time, like in SMB where random 1up blocks are hit etc. But you better go all out on the autoscroller stage :)
The interacting with everything part was already taken care of, I hit everything that is possible to hit. Xenos had said that he preferred to see Mario constantly jumping, even when there was nothing to hit, though. That was more my question.
Living Well Is The Best Revenge My Personal Page
Player (120)
Joined: 2/11/2007
Posts: 1522
Yeah well I guess I should inform my opinions with facts or something... in that case, I would side with less/no extraneous jumping
I make a comic with no image files and you should read it. While there is a lower class, I am in it, and while there is a criminal element I am of it, and while there is a soul in prison, I am not free. -Eugene Debs
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Got it all worked out. At the end of World 1, I am: * 8 Frames ahead * Have 100 more points * And have 4 more coins (cuz Mario gots to have some cash to pay for his ho, once he gets to her). This won't be a very noticeable improvement (other than obvious stylistic changes), but that's what frame wars are for, I guess.
Living Well Is The Best Revenge My Personal Page
Player (119)
Joined: 1/30/2007
Posts: 82
I made a test run of this game (link removed) that finishes 21 (or 20?) frames faster than the current run. I borrowed some input from TheAxeMan for the final three levels to save me some time in making this. I'm not sure if any more time could be saved by manipulating the bonus worlds some more but I suppose there is still this potential. Also, don't watch this movie unless you find faster transition cutscenes or framecounts exciting. Edit: On second thoughts, ignore the previous. I found a faster route that saves 30 frames over the current run. Additionally, I made the beginning of a run that shows the first part of this route up to 2-3 which can be found here. There also exist some places to gain more points as summarised below. 1-1 200 1-2 300 1-3 100 2-1 100 2-2 100 2-1 was completed 6 frames faster but at the expense of 1800 points.
Ambassador, Experienced player (697)
Joined: 7/17/2004
Posts: 985
Location: The FLOATING CASTLE
Very good. Looks like on 2-1 you still have to wait a little bit for that seahorse but not as much. Too bad that kills the fish-jumping streak. It looks like you have this down pretty well.
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Are you working on this in anticipation of doing a full (ie submittable) run, xenos? I've been working on this off and on for a while, but have no desire to get into a frame-war competition
Living Well Is The Best Revenge My Personal Page
Player (119)
Joined: 1/30/2007
Posts: 82
I've already finished, as a matter of fact. Here it is, 72 frames faster than the current run. An additional 2 frames were saved via better lag reduction in 4-1 and 40 frames in 4-3 by defeating the two bosses faster. Coincidently, the score lost by taking a faster route was made up for by gaining points elsewhere leading to the two runs having the same final score.
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Guess I don't need to finish mine then :D
Living Well Is The Best Revenge My Personal Page
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
Two things: - Your movie died at the end on me... - Did you use that trick to push Mario off-screen? I didn't really pay attention so I might have overlooked you doing it. Otherwise, would be cool if you included that. Aside from these, I'd say it's pretty good. :)
Experienced player (822)
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
Movie finished fine for me. I do wish you'd had some more varied activities on the autoscrollers, particularly with the large set of blocks at the end of level 2. Mayb taking a little more time with those and hex them in? Then I'd say this would be an excellent replacement to the oldest GBX publication :)
Living Well Is The Best Revenge My Personal Page
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
Is there a particular reason why v1.1 is being used instead of v1.0?
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
The out-of-bounds bug featured in my walkathon run that is currently submitted works only on v1.0 and not on v1.1. Since I want to improve on xenos' run I took v1.0 since it appeared to be better. Xenos used v1.1. I got 1300 points additional points that xenos had missed, however, the bonus screen is troubling me now. The best result I found after hours of research was ~18 frames behind xenos' run. When I put in idle frames where xenos did (1 frame in 1-1 and 3 frames in 1-2) I get 2 lives on the bonus screen instead of 1. This means that the bonus prizes, or the RNG in the game behaves differently between the versions which is unfortunate! Either I use v1.0 at the cost of 18 frames and possibly more or I use v1.1 at the cost of the OOB glitch. This OOB glitch has not been proven to be a time-saver yet, I don't think it will be one. Personally, I would lean to v1.1 since the final time is what counts for me.
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
I'd like to attempt the hard mode run next. I uploaded the test movie I posted a while back: http://www.youtube.com/watch?v=eTRENWQxdMw It seems that the randomness behaves the same in the hard mode. Putting in 1 idle frame in 1-1 and 3 idle frames in 1-2 is still the best solution, so maybe this will be an easier project to do than the normal run. But here is a new problem and I need opinions in order to make a decision. Certain combos cannot be done without the occurence of lag frames. In other words, skipping combos will be faster by a handful frames. I can do one of the following: [1] Fastest possible: I will not go for combos that have lag frames [2] Fastest possible with entertainment/speed tradeoffs: I will go for combos at the expense of time, but I will not do any other kind of tradeoff (like shooting fireballs where it can be avoided). In other words, lag will be ignored but the run will otherwise be the fastest possible. [3] Fastest possible with more entertainment/speed tradeoffs: I will go for combos and I will shoot fireballs even if it costs time 1 is a good solution, but takes away from entertainment and will look questionable. 2 is another good solution. 3 is arbitrary and I'd rather not do this. Vote please!
Joined: 5/9/2005
Posts: 748
2, would be my choice.
Joined: 11/4/2007
Posts: 1772
Location: Australia, Victoria
I agree with Paused, I vote for option 2. Toothache summed it up best:
<Toothache> do the combos that dont add lag <Toothache> best of both worlds
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
http://dehacked.2y.net/microstorage.php/info/995195391/Super%20Mario%20Land%20TAS%20HARD%20MODE.vbm I'm going with option [1], since I found a solution to my original problem. This doesn't earn me the most points, but still enough to not be considered "missing opportunities". Unfortunately, this seems to get carried on in the movie. In other words, I still sometimes can't combo some enemies since it would mean lag. It seems that no idle frames need to be put in in order to make the bonus screen problem work out in 2-3. I'm now at the beginning of 2-3.
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
Hello, in my hard mode submission text I said different numbers of coins may have an effect on lag behavior, so I was going to study this. I had discovered that collecting one less coin throughout the hard mode TAS will remove 1 lag frame in 4-2, after leaving the passage with the pipes. I don't have movie files at hand to demonstrate, but I might make those later. 0xC0A0 (32bit) = Points 0xFFFA = Coins, increases by 7 whenever the coins counter hits a new ten-digit number, which is weird... Changing any of those before the relevant lag occurence in 4-2 doesn't affect anything, so I'm a little confused as to what the problem is with the lag.
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
In the ending sequence with the fake Daisy in 1-3, I tried looking for RNG or frame rule addresses. FFAC scrolls 0~ff infinitely. After the Fake Daisy is offscreen, when FFAC reaches an odd number, it fades into the bonus game (if the address is frozen at even numbers, the game is locked). Therefore, by being 1 frame slower (to manipulate luck), you can end up tying your time. Or by being 1 frame slower, you might end up losing another frame. EDIT: After completing the bonus game, DA1B has to become zero. If you take the fire flower, it starts decrementing at a later time if you've been small. If you are already big, the counter starts decrementing faster (69 frames faster). If you're already fire Mario, it starts decrementing even faster (additional 64 frames faster). I didn't know this when I was working on my regular mode TAS, since I take damage when killing the dragon in 2-3. If I kill the dragon like in the hard mode TAS and remain big, I should be able to get the 5000 points from the dragon. Thus the regular TAS is improvable by 5000 points, unless the points change messes with lag or luck behavior.
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
The bonus game luck seems to be dependent on something really complicated and unreliable. If I make the 1-up sound effect play 100 frames before the bonus game starts, the outcome doesn't change. But when I let it play 30 frames before, the outcome changes, and it changes again if I let it play 10 frames before. During testing I found that the outcome changes if these addresses are written to: ff9f, ffbd, fffe and e0a4. So since I can't understand the mechanics behind this, I'd stick to what I used to do when I made my SML TASes; simply play through a world, see if the luck is good, if it isn't, insert idle frames in the levels. more RAM addresses:
FFA4 - map x-pos
C202 - mario x-pos
C201 - mario ypos
C203 - Mario's state
DFE8 - music ID handler
FFB3 - if changed, may lead to the end credits
FFFE - if written to, coins increase by 1. If Mario touches a coin, this address is written to.
C0A3 - If written to, lives count changes.
           FF makes it decrease by 1
           1~FE makes it increase by 1 
EDIT: Another points improvement: In 1-1, chain the two flies near the end of the level to gain 400+800 instead of 400+400 points. EDIT: Probably I won't finish the regular mode improvement, because it's too laborious work for me.. There's VBA21 and VBA24m. The latter is more accurate but slower by an estimated 250 frames. Using the former is against the site's sense to be as close to the real console as possible. There's versions 1.0 and 1.1; the latter has better luck over the former on VBA21, hence why it was used. Now I really don't feel like playing through all the worlds again, on VBA24m on both 1.0 and 1.1 to see which has better luck, and then inserting idle frames one by one to see if the luck changes to my benefit. And then coming up with a new run which is actually 250 frames slower...
Editor, Expert player (2315)
Joined: 5/15/2007
Posts: 3856
Location: Germany
I don't know how this didn't appear to me when I initially compared v1.0 and v1.1 when I made my TASes, but v1.0 is now confirmed to be undoubtedly faster. After beating a boss and the gate opens, Mario runs to fake Daisy's chamber whereas on v1.1 he only walks there slowly, losing about 60 frames even with better luck (only one idle frame compared to v1.0's four). (EDIT: This turned out to be misinformation. You can run into the chamber by having landed on the pedestal at running speed beforehand. The v1.1 movie I compared to lost running speed to manipulate luck, and so I assumed it was a version difference. Still, it means I'll gain some time over the published run from this knowledge. Also, I updated the tcrf page in light of this.) World 1 is now completed on v1.0 in the normal mode category and I might continue with this WIP sometime. Also, since it's v1.0, I can use the fancy out of bounds glitch later. :) Compared to the published run, the WIP is 39 frames slower due to extra LCD lag (by better emulation). In fact, both movies are on par. The WIP is 400 points ahead because
Another points improvement: In 1-1, chain the two flies near the end of the level to gain 400+800 instead of 400+400 points.
1 2 3 4 5 6 7 8 9