werster might be cheating at Sonic 3 & Knuckles

Nov 7th, 2013
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
  1. Recently I noticed an odd trend in werster’s runs of Sonic 3 & Knuckles. He seems to have an unusual consistency for things that require a great deal of luck and/or precise execution, to the point where he practically never fails them on stream. Furthermore, his last several PBs have all been on the 1st or 2nd attempt, making for very short stream sessions of this game. For a game as difficult as this, there is no way that any speed runner, no matter how much they’ve practiced or how skilled, can make improvements that big every time they stream after only a couple tries at most (seriously he has less than 100 total attempts).
  3. Below is a list of things that explains why I think werster’s run (http://www.twitch.tv/werster/c/3195473) is so incredibly unlikely/why he might be cheating. Some of these are luck-based run killers, and some are highly precise and unnecessary risks that save minuscule amounts of time. As you read, keep in mind that he’s never failed most of these on stream. Also keep in mind that I have 2000+ attempts at this game (which I have ran exclusively for 2+ years) and have a solid understanding of how likely various strats should be, even after loads of practice. It is not uncommon for my stream sessions to have 40, 50, even 60 attempts. When you’re done with this list, read the conclusion at the bottom for more.
  5. The timestamps are from the 25:13 run, in-game time.
  7. AI1
  9. 0:09 - The bounce off the fire shield and flight has a small window of acceptable timing. It’s very easy to fly too early and grab the zip line, too late and hit the spikes, or somewhat late and run off the platform, requiring a backup strat that involves jumping off the very tip of another platform.
  11. AI2
  13. 0:31 - Late jump here is extremely precise. Too early and you jump off the slope and get a bad trajectory; too late and you run off the twisted edge and get put into a state in which you cannot jump or fly, wasting loads of time.
  15. HC1
  17. 0:06 - There is a trigger here that changes the left side of the loop from solid to not solid. There is a high likelihood of your spin dash getting interrupted, depending on your spin dash speed and subpixels, before the trigger changes it to not solid. Furthermore, the spin dash itself, loop jump, and slope jump have to be nearly perfect in order to get enough height to jump above the screen.
  19. HC2
  21. 0:07 - This slope jump is precise, and timing changes based on spin dash speed and subpixels. Too early and you’ll bump into the wall; too late and you won’t hit the starpost and wall trigger early enough, or you won’t get a jump at all.
  23. MG1
  25. 0:42 - Getting 6 hits on the boss before he goes into the ceiling is one of the most precise boss strats in the game. It requires getting nearly frame perfect flights after bouncing off him so that you don’t lose much height, and the likelihood of this is very low.
  27. The 6-0 alone is not that unbelievable, it’s the fact that it makes the following setup easier, a setup that is already the single hardest trick in the Tails run: getting the ground clip. It’s so hard that I practiced for hours and did not get it a single time, and marzojr. (one of the main TASers for this game) told me it’s very precise even for TAS.
  29. CN1
  31. Getting a 54 on this level involves a few things. First, you need a nearly perfect spin dash right at the start. Secondly, you need to avoid getting caught temporarily on the wall after spin dashing past the retractable spikes. This depends on your subpixels, as your position will be affected by the spikes and your rightward speed is maxed out instantly due to wheel glitch. However, your speed is constantly alternating based on subpixels, so even with wheel glitch you still have to hope you move slow enough so as not to hit the wall and lose time. Lastly, the platform you run around to change your position has a high likelihood of flinging you to the ceiling, also based on subpixels.
  33. LB1
  35. At -26 after Ice Cap, there is no reason to put the run on the line by going for this trick. It is a 1 frame jump, and the risk is death for failing. It saves about 7 seconds for getting it first try. Considering the level of precision and luck needed to even get this far with that much time saved, I can only imagine going for this being a way of making things look more convincing.
  37. LB2
  39. This is hands down one of the hardest levels to play optimally. There are so many factors that can vary your time and so many precise jumps required. Let’s go through the list:
  41. 0:01 - Right off the bat there is something here that can vary your time. You need a good spin dash, followed by an extremely precise jump off the small curve whose timing changes based on spin dash speed. Depending on that jump you may have aerial control from running off the platform or no control. Either way, there are enough variances here to change where you land, requiring a second spin dash or getting stopped by the upcoming slope based on your speed (which, as mentioned, is also affected by subpixels).
  43. 0:04 - This enemy bounce is very precise and the jump timing changes based on speed. There is a high likelihood of getting hit or missing the enemy, and it is only marginally faster then hitting the spring.
  45. 0:09 - When coming from the ceiling like this, and even when braking upon landing, this booster will often times not hit you properly (though it plays its animation). You need to slow down even more to guarantee that it hits you, which is not possible if you don’t jump from the ceiling and fly to purposely lose speed. Furthermore, the ensuing jump is very precise because the ceiling you can bump your head on is much much tinier than the graphic suggests.
  47. 0:10 - This jump is also very precise. Too early and you’ll miss the Flybot but grab the pulley, in which case he’ll hit you off it almost immediately. Too late and you can’t slow your aerial speed down enough to hit him and grab the pulley.
  49. 0:40 - Extremely precise flight strat, it has a very tiny window of input and it is even possible to go straight over the platform and fall on the ring monitor.
  51. MH1
  53. 0:16 - This is one of the most notoriously difficult tricks in the Tails run. It requires a good brake, specific rightward speed, precise jump timing, and subpixels. Even with the aid of pausing to set it up, it has a high likelihood of killing you or ejecting you, both losing a lot of time and requiring that you set it up again after every failure. The level wrap that follows also has a 5% chance of failing if you do it from the correct spot, which is very small, yet still big enough so as to lose runs there occasionally.
  55. 0:33 - This boss strat also has a high likelihood of not getting the 3 hits at the start. Getting the 3rd hit in this manner involves nearly perfect flight timing after the 2nd hit. Too early and you risk flying before the hit (resulting in a very low bounce) or getting the hit and flying into him and dying anyway, too late and you’ll bounce too low and he’ll start moving before you can get the 3rd hit.
  57. MH2
  59. 0:28 - This strat sometimes does not work. In order to go over the yellow spring, you have to land high enough on the expandable platform, which is a small spot to begin with. But even landing on the right spot, sometimes you will get interrupted right above the yellow spring due to how it’s connected with the ground.
  61. FB1
  63. 0:05 - This strat that involves using the platform to clip into the wall is only 1 second faster, but it is much less consistent than going the normal way. Sometimes you will fall straight through the platform instead of clipping, which I imagine is due to vertical speed and subpixels. Time flight too early or slow yourself down with extra flights and you’ll land on the platform instead. Having to wait until the platform is at a certain point in its cycle means you forfeit time on the small chance that you’ll make another platform cycle 1 second earlier. This also means that the flight timing at the shortcut is significantly stricter. Because you have to jump off the platform when it’s at its highest point, without bumping the wall, without going too low, and with as much horizontal momentum as possible. Risking all this for one additional second at this point in the run is not worth it.
  65. FB2
  67. 0:03 - This trick has a high likelihood of killing the run. It requires a good spin dash speed, precise jump, and right/left inputs to help. Unlike most tricks, this one requires both horizontal AND vertical subpixels to be properly aligned. Your subpixels will always vary depending on spin dash speed, jump timing, and input duration.
  69. SO1
  71. 0:03 - Holding left to slow horizontal speed here is risky. In order to jump up the sand and instantly land on the ground next to the rock, you need to enter the sandfall at the right position and then time your jumps correctly. Doing this is significantly harder and saves a tiny amount of time over simply holding right and jumping at the right spots.
  73. SO2
  75. 0:42 - Clearing this jump without accidentally going over the black block is insanely hard. It requires braking just the correct amount so that you have enough speed to go over the pillar, but not enough to go too far, as well as jumping from a precise spot. It also saves very little time over jumping onto the pillar and braking there, yet will waste several seconds if you fail it. Succeeding is not even worth it because there is a pillar cycle ahead that you have to wait for anyway, so there is nothing to gain and seconds to lose.
  77. 1:11 - This flight during the screen wrap to avoid a second jump is both difficult and not worth it. You will gain some time as the screen will scroll at full speed, as opposed to landing and having it slow down. However, there’s a huge risk in failing it and succeeding, again, is pointless as you have to wait for the upcoming pillar cycle anyway.
  79. 1:31 - Another marginally faster and very difficult strat. Landing on the loop while in a position this low involves timing flights just so that you stick to the ceiling for a certain amount of time and then have mostly horizontal movement. Too early and you will start falling down too much to land on it, too late and you’ll gain upward momentum. You cannot land while moving upward, so it will delay your jump.
  81. 1:55 - This boss strat saves maybe 2-3 seconds over the normal one, and yet it is significantly harder. Your flight timing affects how soon you land and can jump, which is important when trying to sneak in the 3rd and 6th hits. Too late and you’ll get hit by the spikes on his hand, too early and you won’t land quick enough to get every hit. Your position after the 3rd hit also depends on where you were vertically, sometimes resulting in a bounce that makes the rest of the strat nearly impossible to do properly as the boss will have time to move and raise his hit boxes.
  83. LR1
  85. There are loads of spots in this level where the ground will completely kill your momentum, based on your subpixels. Such areas include:
  87. 0:04 - Right under the stalactites.
  88. 0:14 - Right near the block in the lava, where you can sometimes even get hit.
  89. 0:15 - Right after that lava pool the ground will stop you at least 30-40% of the time.
  90. 0:33 - Where the ground meets the switch.
  92. These are all important for varying reasons. Getting stopped under the stalactites is dangerous as you can die from one. Spin dashing immediately afterwards can also result in getting interrupted by the same spot, in which case you will certainly die. The block near the lava can clip you into the ground, saving 9 seconds. It can also make you get hit by the lava, wasting time. The section immediately following that has an even higher chance of interrupting you and spin dashing right after tends to result in the same thing. This is critical as it may cause you to lose an entire lava fall cycle to the left of the spiral. Lastly, the switch is important because the spinning mace cycle begins once it is almost on screen. Getting interrupted by the switch means losing time by waiting for it or risking a hit. And then we have stair clip:
  94. 0:55 - This is one of the most critical tricks to get, as it can greatly vary your LR1 time. It requires a minimum of 75% spin dash speed, good vertical subpixels, and a frame perfect jump release. Furthermore, the ground is uneven except for one section which is a flat, sprite-based object. Jumping from anywhere but the left side of that object makes timing the jump release significantly harder, and if you jump too late you will simply bump into the stairs and fail. So to recap you need to make sure your spin dash isn’t too fast or slow, your jump is off the flat object and within a few frames, and your jump release is on the perfect frame, all while having good vertical subpixels. Needless to say a big chunk of time depends on this trick, and it is very late in the run.
  96. LR2
  98. 0:25 - Another very precise trick. To pull it off you need to stand on the leftmost side of the platform that doesn’t crumble, get a good spin dash, and then 2 insanely quick spin dashes immediately after bumping into the wall. The leeway for these 2 spin dashes changes depending on the screen’s horizontal position. It’s extremely easy to fail this by not spin dashing quickly enough, or worse, accidentally flying and losing a ton of time. Every failure also requires carefully setting up on the left side again.
  100. HP
  102. 0:04 - Braking here is slightly faster then hitting the spring and bumping into the wall. It’s an unnecessary risk as the correct timing for this brake is very strict and you risk not landing on the floor above. It also does not change the final level time, as anything under 29 is impossible without TAS-like execution (perfect spin dashes, etc.).
  104. DE1
  106. 0:08 - This jump is incredibly risky. For one thing, the booster object significantly increases your running speed, which reduces the window of time for this jump. Secondly, jumping too early from the slant will screw up your trajectory big time, and jumping too late will cause you to either get hit by the enemy or hit the spring. Braking makes the jump easier, but then you cannot land straight onto the lightning shield because your horizontal speed will be far too slow. That bounce saves only half a second at most, but getting hit or touching the spring will waste 2-3 seconds.
  108. 1:36 - This jump was extraordinarily risky. Due to being that close to the left edge of the screen, a hit here would’ve resulted in being unable to grab the ring and possibly getting snatched by a platform, making it both difficult and risky to get the remaining 2 hits before invincibility was up. Jumping in between max speed rotating platforms is tight timing and being that close to the laser means having to make sure you don’t get too much horizontal speed, as the laser has a higher hit box than it appears. That means a higher chance of getting hit by the spikes instead. The safer strat here would’ve been to walk right into the laser on the ground, then fly into the boss twice.
  110. DE2
  112. 1:32 - This faster, 3 cycle strat carries a huge risk. Depending on your spin dash release, speed, position, and subpixels, it is possible to get hit on the left edge of the screen, as the finger’s hit box will return. This will completely mess up the rest of the strat, as you wouldn’t have enough time to reach the correct position. As it stands, you already barely have enough time to position yourself correctly for the 2nd spin dash. The floor in this fight is not made up of normal ground collision, but sprite-based objects. In order to even crouch, you must stand perfectly still, or else you’ll roll. That means this strat has incredibly strict requirements and it’s a lot to depend on this late in the run.
  114. 1:50 - Another risky strat. This one is not fully understood, not even by marzojr. who probably knows more about this game than anyone else. He could not replicate it under TAS conditions, and I could not do it a single time on console or emulator. The goal is to hit him continuously with a single, charged spin dash and let him bounce you after every 2 hits, but the requirements are so insanely strict and there is a huge possibility of dying. You can either fall too low with a platform, or you can release a spin dash while a platform is falling, which will cause you to break into a run and get hit. This strat requires that the floor’s solidity is in the right position, which it often times will not be, as well as your own position.
  116. ***Conclusion***
  118. Honestly I like to give people the benefit of the doubt instead of assuming malicious intent, but consider what’s happening here:
  120. -Even amazing players will have to reset when luck-based sections are unfavorable. Any one of those luck-based sections, like CN1 or MH1 or FB2 or LR1, etc., happens often enough to force many resets.
  121. -Many of these precise strats have only frames if not 1-2 frame windows (in a 60 fps game) which change based on speed and other variables. The likelihood of getting most of these without fail almost every time you do runs is microscopically small.
  122. -Not only has this happened for several of his PBs, they’ve also been in a row and also only 1-2 attempts on stream.
  123. -There is no game I know of with any hint of difficulty, and no speed runner I know of no matter how skilled, that can make such large improvements on a nearly optimized run that many times in a row after only a couple tries. And S3K has more than a “hint” of difficulty, it’s one of the hardest, most complex platformers around.
  124. -From what I’m told, werster had cheated in the past (Sonic Center ILs) and had admitted it when caught.
  125. -He inexplicably dropped S3K from the AGDQ lineup, even being one of the people who’ve been upset about AGDQ’s lack of Sonic games. This year especially, Sonic R and Sonic 06 are the only two games in the schedule, so the removal of S3K, even after all the recent improvements to the route, is baffling.
  127. Here is a list of his last few PBs. There is also a 25:4x but apparently he doesn’t have a highlight of it.
  129. http://www.twitch.tv/werster/c/2492138 (28:39)
  130. http://www.twitch.tv/werster/c/2946698 (26:11)
  131. http://www.twitch.tv/werster/c/2993385 (25:25)
  132. http://www.twitch.tv/werster/c/3195473 (25:13)
  134. Now you might be thinking, “what about LB1 and LR1? So much time was lost there!” If he didn’t lose time anywhere wouldn’t that be a perfect run, and by extension, a dead giveaway? Some things have to go wrong, and what better way to make something convincing than luckily get the hardest trick Tails has yet failing one comparatively easier?
  136. Then there’s his webcam. In most of these recent runs, he’s always been in a relaxed position with his arms far apart. Now he plays on emulator with a keyboard so naturally his hands won’t be as close as they would be with a controller, but the fact that he looks so relaxed playing such an intense run, and his keys cannot be heard, does seem bad when you consider all of the above. Furthermore, he’s had some oddly delayed reactions, such as when hitting the last spring in MH2 in his 25:13 run.
  138. All in all none of this is conclusive evidence, but it STRONGLY suggests that, at the very least, his last few runs may not be legitimate. At this point I feel like I can no longer trust his runs unless he switches to console and/or streams all his practice and attempts.
RAW Paste Data