I think Happil 2 has the elements of a really good game, and I like the insanity too, but it's gotta be humanly possible or else we can't appreciate it. No one would know what cool ideas are inside if we all get stuck at the beginning.
I'm curious what difficulty Star has made this. I'm assuming he nerfed it significantly and fixed the impossible parts, but what was his goal difficulty? Level 90? 95? 98? I can accept the game being harder than Kamilia 3, but not by too much because that would be pushing it.
Ok Hector gave me some key binding advice that helped me to finally test cactus jumping and figure out what it is. A lot of you already know how to do this, so like usual I'm mostly giving the technical details, which will be helpful for anyone interested in the more sophisticated applications of cactus jumping, like double and triple cancelling. Anyway, for anyone who doesn't know, cactus jumping is a technique that gives you extra flexibility in single-jump cancelling. The normal single jump cancelling technique only lets you perform micro jumps (jumps lower than 1-frame jumps), rather than letting you jump at whatever height you want (like a "3.5-frame jump," which is needed to beat the eponymous I Wanna Beat the Cactus Jump).
First, what Hector found was that the reason I couldn't test cactus jumping before was that you need to be using only one type of shift (i.e., use only left shift keys or right shift keys). Common sense suggests the opposite, but for some reason the computer can only differentiate between your shift presses if use the same type of shift over and over. Weird, right? Anyway, this means if you're using a keyboard, you gotta use a key mapper like SharpKeys to map extra left or right shift keys, and if you're using a controller, make sure all your jump buttons are mapped to the same shift. As far as I'm aware, it's impossible to cactus jump on a straight-up, unmapped keyboard (let me know if any of you have done it).
You might ask if cactus jumping is really "legit" then. Eh, I think it's legit enough to do this. You could avoid using key mapping software by plugging multiple keyboards into your computer, but that seems unfeasible. As long as you manually press all the necessary shift buttons yourself and don't use a macro or an imperfect key emulator (like OkayKeebes, that program that makes jump cancelling easier), then I think there's nothing wrong with this.
Anyway, let's begin with cactus jumping at its most simple, with only one cancel. We'll need two shifts (of the same type) which we will call shift1 and shift2. Now, the point of cactus jumping is to perform two shift_releases (which is what slows the Kid's ascent during a jump) at any time the player wants. Let's say you want to perform shift_releases at frame 4 and 9, like for the original Cactus Jump. You will need to press shift1 and shift2, but not necessarily at the same time. In fact, even if you press them in the same frame, the computer will still detect one being pressed slightly before the other (GameMaker observes key presses in real time). Let's say you press shift1 first. By frame 4, both shifts will be pressed. At this frame, you will need to let go of shift2. The order needs to be reversed. If you press shift1 first and let go of shift1 first, the computer will recognize this as you ending the single jump and thus register your input as a 1-frame double jump. This also caused a lot of confusion during testing. So, if you press shift1 and then shift2, you must let go of shift2 in frame 4 and shift1 in frame 9 to achieve the desired jump. To maximize success rate in this, it's probably best to purposefully decide which shift is to be pressed second so that you can sort of nest the shift2 input inside the shift1 input. This will be more relevant for double and triple cancelling, which I'll go into in a moment.
First, a quick point on the application of cactus jumping for single-cancelled jumps. While regular, 1-shift_release jumps have a lot of flexibility (obviously), there are situations where you need a very specific jump height, like the original Cactus jump, squished gates/diamonds at certain heights, or jumps that need you to be on a very specific vertical pixel string (like ). For such jumps, a single well-timed cactus cancel should virtually always suffice because it gives you subpixel precision flexibility. Any more precision for your needed jump height can be achieved through starting at the correct V-align. Going onto a second cactus cancel will almost never be needed, except for the purpose of doing an incredibly low micro jump. While cactus jumping is not needed for regular micro jumps, extremely low jumps require 2 or more cancels, so this is where cactus jumping is needed.
Double cactus cancelling is mainly known because it lets you do a 4-pixel single jump (as opposed to the usual minimum of 5 pixels for single-canceled micro jumps). This lets you jump under spikes that are 25 pixels above the ground. There are two ways to pull this off. First, you can map 3 shifts (of the same type) which we will call shift1, shift2, and shift3. Since it's a micro jump, you will have to press them all in one frame. The tricky part is keeping track of the order. Again, you will need to release the shifts in reverse order, so you will have to either get lucky, or purposely press shift1, shift2, and shift3 in extremely quick succession. Then, release shift3 in the first frame, release shift2 in the second frame, and release shift1 in the third frame. Alternatively, you can use the typical jump canceling method, which will only require having two shifts of the same type. Begin by already having some shift pressed (it can be any shift of either type). At the first frame, let go of this shift, and then press shift1 and shift2. Release shift2 on the second frame and shift1 on the third frame. This is probably the easier method, but you might want to practice the first method if you intend to learn triple canceling.
And of course, as I've mentioned before, it is possible to perform a third cancel. Triple cactus cancelling is incredibly subtle, because it does not make you jump any lower. It just makes you fall 0.05 pixels per frame faster (the reason for this is because of how the calculations work out. If you're curious, check out Wolsk's video on the Mega Jump 3). The main benefit of this is that if you're on a low enough V-align (like .4, the spawning V-align) and you perform a triple-canceled jump, you can return to the ground in just 7 frames, instead of 8 frames for a double-canceled jump. This is why Mega Jump 3 requires a triple-canceled jump. Triple-cancelling could also be used for incredibly precise and contrived TAS jumps. Anyway, triple canceling is what you'd expect given what I told you about double canceling. Either method works, though I'll assume no one will try the method with 4 shifts of the same type (which would require you to press 4 shifts in a certain order all in the timespan of 1 frame). So, what you'll need to is have some shift already pressed. On the first frame, let go of that shift, and then press shift1, shift2, and shift3 in quick succession. On the second frame, let go of shift3, on the third frame, let go of shift2, and on the fourth frame, let go of shift1. So essentially, after letting go of the shift you were holding from before, you perform a double canceled jump with the first method. This obviously requires a decent amount of luck and lot of skill, but I believe it's humanly possible (at least in total isolation, like for the Mega Jump 3. I can't imagine anyone successfully pulling off a triple cancel in practice in an actual super-precise jump).
Anyway, if you have been, thank you for reading this far. Now go out and use this knowledge (or get a life)! Just kidding I love you all
If you're wondering why I hacked Lemon's Brute mod, it's to make use of the V-align caption (and I'm too lazy to figure out how to do that myself).
I included a mini F jump in the select stage so you can test whether a given computer has the alternate physics. You should probably do the tests on one computer with regular physics (like your own) as a control, and when you get a chance try them out on a computer with alternate physics.
Test 0: You start in a room with 3 warps, each sending you to a different test. Before you do any of those tests, jump around and try to see if your V-align behaves unusually. Does doing a full jump lower your V-align by precisely .1? Is your V-align always between .1 and .5?
Now for the 3 main tests, all trying to determine jump height. Note that apples represent spikes whose heights are multiples of 5 pixels, as references. Each spike is 1 pixel below the previous. Determine the lowest spike you can jump under (for the given type of jump). Note that once you reach this spike, you will only be able to jump under it if you are on an adequately high V-align (remember, higher V-align = lower to the ground), so I'll also need to find out what the transitional V-align is to determine the exact jump height. However, the transitional V-align may be an unstable V-align, in which case don't worry about it for now. Just do tests at V-aligns of both ~.1 and ~.5, and if you only die at ~.1, do more tests to narrow the number down to the specific transition. If you survive at both ~.1 and ~.5, but die to the next spike even at ~.5, that means the transition is at an unstable V-align.
Test 1: This tests 1 frame jump heights. The leftmost spike is at a height of 40 pixels, the rightmost 30. In regular physics, the last you should be able to jump under is the one at 37 pixels, at any stable V-align. The transition is an unstable V-align.
Test 2: This tests jump cancels. The leftmost spike is at a height of 32 pixels, the rightmost 22. If you're able to pull off ideal jump cancels (I'm assuming you're not double canceling), then in regular physics, the last you should be able to jump under is 26 pixels. The transition is at .15. Below that, you will die even at 26.
Test 3: This tests full single jump height. The leftmost spike is at a height of 110 pixels, the rightmost 100. In regular physics, the last you should be able to jump under is the one at 107 pixels, at any stable V-align. The transition is an unstable V-align.
Well, that's that. Hopefully that didn't confuse you.
So one day I downloaded a fangame to a school computer. I manage to do a 1 frame mini-F, and I thought it was just the keyboard. I tried it again on a different computer in school, but couldn't do it. Then I realized that the 1 frame I did the first time was on a computer with XP. The one I couldn't do it on was with 7. There is a difference in physics.
Yes, my conspiracy theory has been validated!
Hector I can show you how I do my tests in stream. My methods are pretty rigorous.
Inferno, I don't know how much access you have to those computers, but would it be possible to do some tests on them? With a little data, I can determine how those computers interpret the game code differently and create an "XP patch" for fangames that are broken in Win7 physics.
I've never heard anyone say double block diamond, only ceiling double diamond or roof double diamond.
Ya, you're right, ceiling double diamond is what people usually call it.
Quote
A double invert is pointless because it's just literally 2 inverts in a row.
I see what you're saying, but I think the jump is common enough that its existence should be acknowledged. Like, there isn't one universally accepted form for a double corner, but the double invert is a specific and rather common position (well, in needle games, probably not in the games most people play, though many people would probably be curious about the hardest common positions anyway) that comes up relatively often in discussion.
Quote
Your triple diamond is meh because there are a lot of possible heights
Ok, then erase the 40 px. The point of that was to emphasize that, in its most common form, the triple diamond is raised a little, though I suppose that's clear from the positioning of the block in the image. The triple diamond isn't necessarily a common position anyway, though since the diamond and double diamond are included, not including the triple diamond suggests that it is impossible or never used.
Quote
(in contrast to the 16px gap you can actually see that it's 4.5 blocks without having any knowledge in pixels).
Good point. I mentioned the 4.5 block jump because 16 px gap was included, but yeah, there's a clear difference in how much prior knowledge is required to understand the terminology.
Well, I wasn't talking to qoo and Kamilia about this, but yeah, language barrier is a problem. In any case, since it seems no one really cares, it's safe to say that anything that seems impossible for us (Brute, Butterfly Nova, SAT, etc.) is fair ground for modding.
I don't think this is the case at all. or at-least anymore. if it was then it makes no sense really why we haven't encountered any impossible jumps in newer games that have been confirmed beaten. I have said this before but I feel its likely that qoo used tool while playing these parts as his description said TAS. There is no solid evidence to back up the alternate physics claim and the jumps that have been said to be impossible for us are also impossible for other communities, otherwise the triple diamond in danger 10 makers probably wouldn't have been squished.
I still don't understand how it would make sense that qoo was livestreaming a TAS. Unless it was a pre-recorded TAS that he was showing off on stream, but then why was he constantly dying? Not to mention that it's still not clear why Hourglass causes these subtle changes in physics (though of course, we don't have clear reason why the physics would change in any circumstance).
The fact that this problem doesn't seem to come up anymore would also suggest that the cause might be older computers (OS Windows XP, which of course is also the OS of Hourglass users). Leehe claims to have beaten his jump, but no longer uses it. Perhaps he is aware that people with newer computers can't beat it?
In any case, do you believe Kamilia's claim that Brute of a Man has been beaten by two people? It would be helpful to at least talk to those two guys (or qoo), though I have never had any success in contacting Asians (qoo, the creator of Brute, and Kamilia have ignored my messages in the past). Is there someone in the community who is relatively easy to get into contact with?
EDIT- also i dont think we should change the engine physics -
So is it known that it's Windows XP that causes this? That would explain why TAS'ers often encounter this. Not because of the TAS'ing itself, but because Hourglass only runs on Windows XP.
In any case, when a game is designed to be played with the other physics, wouldn't it be preferable to play it with its original physics? Obviously, it would be annoying to readjust ourselves to the other physics, so I'd only do it if necessary, like in the case of Brute of a Man. It seems that a mod that gives the game its intended physics back is preferable to a mod that tries to adapt it for our physics.
There is evidence that fangames behave differently on certain computers, in particular those of a few Asian players. For those that haven't talked to me about this before, what I'm referring to is a phenomenon of certain Asian players doing challenges that seem to be physically impossible for Westerners. The most famous example of this is multiple Japanese and Korean players purporting to beat "the Leehe Trap jump" (a triple diamond raised one tile), which is impossible on Western computers (Shadow says he did it once, but since he hasn't been able to reproduce this or get in on video, I don't know what to make of that).
The clearest evidence is this video showing Qoo beating a jump in Brute of a Man that has been thoroughly tested and shown to be impossible, at least on our computers: His double jump seems to defy the physics that the Kid follows in Western computers. Also, him getting this far in the game would have required him to beat another impossible jump earlier in the level. Furthermore, Kamilia claims that two players have beaten the entirety of the game (and I highly doubt he was referring to the mod).
I really want us to find out more about this discrepancy in physics, but it's difficult due to the language barrier. I would really appreciate it if anyone here who has contact with a good part of the Japanese or especially Korean community to help me get information on this. I would like to have the players who are able to pull of these impossible tasks to perform certain tests for me that would allow me to figure out exactly what's going on. Also, we shouldn't discount the possibility that there are Western computers with these alternate physics. If anyone wants to test the jump in the video above, just put into Geezer's program and see if it's beatable on your computer. It's a rather simple jump, so if it is possible on your computer, it shouldn't be too hard to clear.
Ideally, we should be able to modify the Yuuutu engine so that it behaves under the rules of these other physics even on Western computers, so that challenges like Brute can be available to us (without having to artificially mod it to be beatable under our physics).
Wow this thread is getting intense. I am so proud I caused this.
In the interest of civility, although we'll rank games that are unreleased, if the creator opposes its early inclusion we'll wait till the official release. Since Paragus (who's the "designer" of Cultured I think?) wants us to play it fully before being ranked, I say we take it off for now. Once released, its ranking will be determined by community consensus.
Alright it will be removed for now, although i highly doubt it will effect the level but ok. - Swordslinger
Lv 53 : I Wanna Get Cultured [Clear] [No Difficult] Lv 53 : I Wanna Get The Yellow Star [Medium]
What? Where did you come up with this? Considering our game isn't out and only 3 people in total have played our game through to the end I'm wondering why our game is on the list and where this rating came from. Yellow Star is a 15-20 minute long face roll game, and out of our 7 Stages of increasing difficulty, we put it into Stage 1. Maybe you should take this off the list until you actually play it.
Ratings for mostly completed games that aren't yet released are often included if it's a game that is expected to be popular and that people will be curious about. Due to the difficulty at the end of the game, I'm inclined to say the ranking should be raised to lower or mid 60's.
That being said, it doesn't seem unreasonable to include rankings for certain well-acknowledged accomplishments, even if they're not winning the game (like they did with Kamilia 3 ranking each stage). Maybe on W-level you can include some challenges that are worthy in their own right. We obviously aren't bound to the Koreans' structure.
Show Posts
Just kidding I love you all 
