Flowerspeedrun S1 12:09 SS run

UPDATE: The guy made the most obvious splice ever in his Labyrinth 1:17 video. https://www.youtube.com/watch?v=foUrF9s_IGg#t=39s. In case (when) he removes it I uploaded a mirror here: https://www.youtube.com/watch?v=ndi8HHhBuyE
UPDATE 2: He admitted to cheating: http://t.co/rsBucHmnpr


Video: https://www.youtube.com/watch?v=DnSL0k5TDR0
Emulator used: Gens (Gens+?)

Notes
GH1
0:17 - Does a roll + jump that doesn't save time outside getting a 24 (if at all?) but is harder than just jumping over. In short, added difficulty for no gain.
0:22 - Walks up the slope instead of jumping up on it. Time loss without increased consistency.
0:28 - Doesn't hold down at the top. Time loss without increased consistency.

GH2
Nothing is wrong here but 19 is a very hard time to get without wrapping.

GH3
1:42 - Everything about the start here is weird because his second jump sets him up for landing on the slope as he does with his 3rd jump which is visibly slower than jumping onto the hill.
1:55 - Doesn't have any qualms about pause buffering but avoids it later in the run

M1
2:47 - This is really weird for a lot of reasons. First off there's three ways to do the zip, 1) You press left and pause at the same frame, then hold left 2 frames as you unpause 2) You press right for 1 frame and then hold left for 2 frames 3) you press right for 1 frame then immediately the frame after press left for 1 and only 1 frame which will leave you standing in the wall. The run uses the third (and arguably the hardest) of the methods. Now while standing there you need to get a good speed which to be completely honest I don't know the window for but if it's less than 4 frames of holding left you won't have any time to jump at all, unless there's a small subpixel window where 3 frames does work. After you get the zip you have a 1 (possibly sometimes 2 depending on subpixels) frame window to jump while being offscreen and without any good cues. After all this, there's a roughly 75% chance that subpixels will screw you over judging from that your typical speed is around 20-21 pixels/frame and you need to get past a 16 pixel wide wall. The exact number depends on the number of combinations that work compared to the number of frames that will let you jump but judging from personal experience the guesstimate seems reasonable. All in all it can be estimated that the chance of getting this zip in an IL should be roughly 1 in 20. Now consider the route for the whole run; getting to Marble 1 might be about 1 in 5 runs, getting Marble 1 is 1 in 20 and the rest of the game roughly 1/100. With these numbers the average amount of attempts you need to do to simply get all glitches in a run is around 10'000. Not impossible, but now you have to optimise this 1 in 10'000 run and the numbers soar beyond what any human reasonably could do. Even if you estimate the chances of getting the tricks as low the chance of getting them all first or second try should lie in the thousands. Of course anyone could get lucky, but this was the third run with this trick he posted over an 8 day period. Unless he knows something I don't, it sounds unlikely at best.

M2
3:56 - Uses a harder strat for no reason.

SY1
5:19 - Uses a frame perfect strat (with success) when alternatives that aren't frame perfect but equally fast exist.
5:29 - Hard to tell but looks like he's holding left while rolling for some reason. (Might not be accurate)

SY2
6:17 - Does a frame perfect jump again (with success) when alternatives exist that are marginally slower (less than a second) but that has practically 100% consistency.
6:22 - Pretty sure this (best possible) trajectory is frame perfect as well but might be a 2-frame window.

SY3
7:05 - Here he does two 2-frame zips and one 1-frame zip. Using a keyboard or even a d-pad, it is very hard to press a key for 1/60th of a second. To pause is generally considered much easier because it is more easily timed. In other parts of the run he has no qualms about abusing pause so to not pause neither here or in Marble 1 is weird.

L1
8:38 - This level was done more or less perfectly up to this point. There's a very intuitive and simple way to not get hit here: hold left for a little bit and roll. It wasn't a mistake either considering he did the exact same thing in his other runs. He probably does this as a sort of consistent setup, but it's slower and the alternative is consistent.

L2
9:23 - Gets a bump on the head that cancels his vertical speed which is very hard.
9:30 - Gets a hard setup for the zip and then activates the speedcap within a window of a few frames (about 4-5 I think?). Very optimal.

L3
10:07 - Beating the cycle of this spiked ball requires quite optimal play.
10:13 - Largely inexcusable bad roll.
10:46 - A lot of very optimal and cleanly executed jumps here.

SL1
11:44 - Well planned movement and safe execution. You need to be on the "left side" of the last standable pixel which it even seem like he adjusts to. IIRC this wrap has a less than 50% chance of working (subpixels) but gets it first go.

SL2
12:19 - Running into this spring messes with the subpixels for the wrap, ends up missing it.

SL3
14:00 - Creative route, great and insightful bossfight.

SB1
14:47 - Reacts to the platform, doesn't anticipate it.

SB2
16:08 - Has good spatial awareness of the position of the spiked balls. Or just dumb luck.

SB3
16:50 - Didn't even know this was possible. Doesn't seem to go for the bounce on the badnik by the end.

Final
17:10 - Does the trick that abuses pause to potentially save a second rather than do a 100% consistent method. Ends up loosing 7 seconds.


Comments: Most of these notes might seem nitpicky but there's a point I'm trying to make. In some regards, the run seems to be done by a beginner that doesn't know all that much about how the physics engine works or how to handle different situations. He doesn't seem to have tried to test or optimise his routes and mostly just copies whatever strats he sees others do without questioning. A lot of glitches he simple "goes" for, instead of mindfully setting them up.

In other regards, he's a skillfull and intelligent player, using routes I've never seen before (SL3) and beating bosses decisively. He get's multiple very hard strategies that only save fractions of seconds if anything.

So who is this guy? It's a person, likely young, that speaks Russian and bad English that no'one seems to have heard about before. It's a person that claims to have beaten the record decisively 3 times in a single week after a year of inactivity. It's a person that never streams, never talks in forums or to anyone knowledgeable in the game and either don't search for information enough regarding the run or doesn't have good enough English to understand it. It's a person that did multiple extremely hard tricks in a single run even going so far as to doing them harder than necessary or other multiple unnecessary/unmotivable frame perfect tricks. It's a person that uses external programs to record fucking Gens fucking + in fucking 30fps on a fucking PC that can record fucking CS:GO in 720p.

I feel like everything here is contradictory and therefore I wouldn't accept this run as the record without further proof.