View difference between Paste ID: KQ31UMDc and JgugCHFw

-[From mraglar (3-Jan-2013)]
+[From mraglar (23-Jan-2013)]
 
-I've done some short research of the supposedly "random" factors that determine whether you'll pull off the second part of "the underwater clip with Yoshi" or not. I should start by saying that unfortunately there's really nothing you can do about it in order to increase the success rate, but I thought sharing the knowledge of why it works the way it does is never a bad thing.
+Ok, there were more factors that determine whether the trick will work or not. I have only tracked down one of them but at least I've found the actual chance that it will work.
 
-Before I took a deeper look into this (I have no earlier TASing experience with this game) I was quite certain that this was an issue of having different vertical subpixels everytime you're going into the wall in the first place, but this turned out to not be the case at all. The reason it works the way it does is because of framerule based vertical physics when being under water. This means that your vertical speed can only decrease at certain frames, in this case every fourth frame.
+It turned out that the success rate is also determined by the vertical subpixels of Yoshi. In this game, there's 16 subpixels per pixel and are stored as an unsigned byte meaning that they only come in multiples of 16 and can thereby are expressed by the numbers 0, 16, 32, ...,240. Here is a list of the chances to succed for every subpixel that Yoshi can have, after executing the first part.
 
-So let's say that your vertical speed can only decrease when the frame counter modulo 4 is 0, for simplicity. If you're now standing on the ground and push the jump button for 1 frame at frame 1 it will take 3 frames before the vertical speed decreases while if you push the jump button for 1 frame at frame 3 it'll only take 1 frame before the speed decreases, and thus in the second case you'll reach a lower height than in the first one and your vertical speed will also reach 0 at an earlier stage in the jump. And reaching 0 as fast as possible is the important thing when executing the trick since you won't be able to mount Yoshi if your vertical speed goes upwards.
+0-16: 0%
 32-64: 25%
-My tests showed (by holding down/left while unmounting Yoshi, don't know if you do it differently) that the trick works if you unmount Yoshi when the framecounter modulo 4 is not 0 - that is in 3 out of 4 cases. In the fourth case you'll reach that 0 speed too late into the jump so Yoshi has reached too high up when re-mounting him, otherwise Yoshi will be low enough to still be inside the wall. So the chance of succeeding is 75%.
+80-96: 50%
 112-240: 75%
 
 This explains why I first thought the chance was 75%. Since in all my tests, before I looked up the subpixels, I seemly always got values between 112 and 244 and thereby thought they didn't matter, oh well. If my math isn't completely out of order the chance of succeding is:
 
 (2/16)*0.25 + (3/16)*0.50 + (9/16)*0.75 = 53.125%
 
 But that's only correct if all subpixels have an equal chance of occurring, which might be the case if you miss it once and then have to try again. However as for the first attempt, granted that you succeed with part 1, the chance is always 75%. At least for FoI2.
 
 Why? Well Yoshi's subpixels are a bit special, they only change value when Mario isn't on Yoshi. At the start of the level Yoshi vertical subpixels are always 0 and will thus stay that until you try the glitch (granted that you don't unmount him before that). And all my tests show that from the time that Yoshi is unmounted to the time when he is mounted inside the wall, his subpixels always change by:
 
 (current_subpixels + 128) modulo 256
 
 Meaning that the first time you you perform it you will land inside the wall with a vertical subpixel value of 128 which is in the range for the 75% chance.
 
 It doesn't end here though, because if the water physics and Yoshi's subpixels were that only things that determined the outcome then, for all subpixelvalues from 128 to 244, the chance of succeding would be 100%, this isn't the case though and I honestly don't know why. Could maybe be that Mario is at a too high position when Yoshi is mounted but if this where the case then I can't see why it always works every 3 out of 4 frames since Mario's vertical subpixels changes value every 4th frame (according to the water physics) and thereby I would expect some inconsistency in the success rate.
 
 
 [previous message: http://pastebin.com/JgugCHFw]