Behaviour of N on gaining an extra frame
Sure you're all familiar with 151-0 speedrun, a run that van recently took by cleaning a frame of a seemingly maxed run by adding a jump, which only on rare occasions works. I've been fiddling around with the trick for some time today, and I'd like to open a thread to discuss and research about how it works. I'll start by mentioning what my first impressions were, and a couple theories on how it might work, including disquisitions about N handling fractionary frames, etc.The last frame touch theory
After thinking that it was just a random frame that you gain when plain jumping on rare occassions, I checked out many runs and they all had the same in common, the jump ends right next to the door, that is, the next frame after hitting the door, the ninja would fall on the floor, if it didn't finish the run. Either that or just touching the door really low when falling, but before falling to the floor or it wouldn't work, probably because his feet are more to the right when he's in the air, when landed his feet are much closer to his body, and hence further away from the objective, in this case the door.Both scenarios are just the same to replicate so which one is the cause is not really a problem. I had been trying it out unsuccessfully until I noticed it, and then managed it a couple times in no time, so I assumed this to be true. Naturally, I headed towards Debug mode to test my theory and eventually try it out in 59-0, as any normal Metanetter would.
This seemed to be right, however I eventually got two MBD's with this scenario happening, but one of them was 152f and the other one 153f, so this screwed the theory up. These are the MBD's, which I made as similar as possible in their endings.
153f:
152f:
Note that I finished writting both runs in frame 152, that way by using Display Mode you can verify that the second one finishes, while the first one requires another frame that has no input, and hence the arrow vanishes before finishing the level. Use fbf if you wish.
My final conclusion on this theory is that, even though it doesn't seem enough, it's a necessary condition for the extra time to be gained, as I couldn't manage a run without it. Didn't try too much, and I haven't seen all runs on the boards, but all the ones I saw were like that.
The one pixel apart theory
That being discarded, I noticed another interesting thing in the last frames of those MBDs I made. The ninja was damn close to the door, in fact, he was just barely grazing it. In N, you evidently don't advance a pixel per frame, that depends on your speed (at least), this means that every frame you hop from place to place skipping most pixels, however the hitbox is more precise, pixel-perfect, which means most times the last frame of the runs you're "far away" from the door, but since in the next one you passed it already, you're done.Now, why do I say all this you say? Well, as I said, I naturally hurried to try this in 59-0, tried to get a perfect MBD and emulated the ending I explained above, falling right as the door starts. I checked the last frame, and guess what, the ninja doesn't happen to be grazing the door this time, it is really far.
In this conclusive picture you see the last two frames of all 3 MBD's, the first one being 59-0, the second one being the fast 151-0, and the third one being the slow 151-0. Start studying it!
Note how on the second one, on frame 153 it has already touched the door, it has passed the hitbox, meanwhile on the third one it hasn't yet.
From this picture we deduce that this jumping trick is not a one frame trick, but rather it saves way less, but enough to cut it to the next frame.
Based on this I also can, unfortunately, foretell that this implies the trick will not work in 59-0, because you don't happen to be close to the door on the last frame at all.
Fractionary frames rounding
This is just another way of rephrasing the previous theory, I guess, but its worth saying because it provides another point of view of study. Because of the ninja skipping many pixels, it is clearly very unlikely that the ninja will hit the hitbox of any particular entity exactly, but rather one frame it'll be far apart, and the next one it'll already be inside it or past it.It is then natural to ask oneself the following question: Does N, in the insides of itself, predict whether some hitbox is going to be hit inbetween two frames, how does it anticipate it if so, how does it handle that, which frame is the hit attributed to? How is all that adjusted, is it rounded, maybe truncated?
This time, we are in particular concerned about the exit door hitbox. You guys have sure noticed how after finishing the level, the clock takes a bit to finish rolling numbers. Have you even been attentive enough to notice that time varies depending on the run? Sometimes it stops suddenly, sometimes it takes some more time to adjust. This just leads to thinking that N indeed detects the ninja has hit the hitbox inbetween two frames, the closer it is to one exact frame, the more suddenly it stops, the further away you're (up to a max of 0.025, of course), the longer it takes to go to the frame that it should stop.
If you're attentive enough you've sure noticed how sometimes after finishing a run the clock is painfully close to stoping at, say, .725, but finally it slowly rolls down to .700. Some other times it stops so suddenly that you think "wow, that was close, I almost don't make that one more frame!". Some rare times, it even stops at odd times, like .749 rather than .750, which indicates you were lucky as fuck to be "awarded" that extra frame, it almost went down to .725. I can't remember a converse situation, i.e. .751, but it might just be I didn't notice it. It would mean you were unlucky as fuck. Anyways, this brings the topic of fractions of frames, to which precision this is detected and how it is handled is very important, and I have no clue of course.
However, if all this is true, what we can deduce from the picture above that what happens in 151-0 is that in optimal conditions you happen to be massively close to the door on the last frame, it just needs a little help, a little push, to be past it on that same frame. This leads to thinking, again, that the jumping trick only saves a tiny fraction of a frame, just enough to make the cut to the next frame in the special case of 151-0, but not enough to do it in general (namely, in 59-0 ;_;).
Conclusion
I've exposed some behaviours and knowledge that sure many of you already knew, mainly for the completion of the post. I've also exposed a couple theories that I think might explain how this works, however nothing is conclusive, and hence the conclusion is that there's no conclusion yet and it requires more research. Unfortunately for now I'd say this trick won't help in 59-0, but it's still interesting to study how it works.You might have a little doubt yet. In the picture above, why is one of the runs 152f and the other one 153f? They're both very similar, the difference between the distance to the door in both is clearly way less than what the ninja travels in a frame as we can see in the picture of 59-0 in which the ninja is much farther yet its also its last frame; they fall the frame after touching the door, and even the 152f run ninja is a bit higher than the other one, so what is going on here? Well that certainly deserves to be studied, because I'd like to know too. Here's a more precise look at those final frames:
And here's the best zoom I can get in Flash:
You can clearly corroborate in this last one that, indeed, the foot of the ninja is closer to the door in the first picture, just barely. If that makes the difference, just imagine in 59-0, where the foot is like 10 times farther away from the door. The why, though, needs to be studied, we need your help!
Discuss.
) on 11/9/2009; 400 on 12/2/2011; max of 598 on 2/3/2013; tied for 0th in top 20 rankings on 2/26-27/13
