These videos - are they real or fake/bot?

[tjenyao]

Was arguing or rather discussing with a RL friend about this, and we decided to settle this matter on HD. So what do you guys think? Plus explanations/reasoning please. Thanks heaps

[spoiler]fake or real + explanation pl0x
tjenyao5:03peopleeeeeeee
tjenyao5:03http://www.youtube.com/watch?v=dZjDIkMapAc&feature=plcp
Agamemnon5:13looks like a good ST stacking run
Agamemnon5:13zt rather
tjenyao5:13Yep
tjenyao5:13Do you think it's real?
Agamemnon5:13probably
tjenyao5:13Look at his other videos
tjenyao5:13I dunno.. but seems..... fake.
tjenyao5:14http://www.youtube.com/watch?v=h8xDLDRfgV0&feature=plcp
tjenyao5:14What about this?
Agamemnon5:15I'm just not sure what the maximal settings for TB are
tjenyao5:16He'd probably be at least mentioned in HD if he could do stuff like that o_O
tjenyao5:16Not as good as nullpo.
tjenyao5:16This one seems soooo fake. http://www.youtube.com/watch?v=pMf-3mKe7Ig&feature=plcp
Agamemnon5:17because nobody on HD ever talks about TB
Agamemnon5:17because it's.. a pretty bad game
tjenyao5:17Yea..
Agamemnon5:17oh, that's cute
Agamemnon5:18he's using massi4hs infinite tst setup
tjenyao5:18Yeah.
Agamemnon5:18thing is
Agamemnon5:18he looks fast and good at spins
Agamemnon5:18but that says nothing about his skills as a player in general
tjenyao5:20True but
tjenyao5:20Argument or rather discussion
tjenyao5:20with a friend, debating the originality of the videos
tjenyao5:20his videos*
Agamemnon5:20it's possibly real, possibly bot
Agamemnon5:20but probably not bot
tjenyao5:20O_o
tjenyao5:21So you think it's real. Fine 1-0
Agamemnon5:21basically, if you repeat those endless patterns, like zt and inf tst
Agamemnon5:21you place the other pieces in specific spots/ways
Agamemnon5:21which means you can stack it really fast compared to freestyling it
tjenyao5:21True
tjenyao5:22But you've casted your vote
tjenyao5:22Thanks a lot.
Agamemnon5:22np
tjenyao5:22Now I'll have to wait for others to come on
tjenyao5:22http://www.youtube.com/watch?v=dZjDIkMapAc&feature=plcp
Agamemnon5:22post it on the forum instead?
tjenyao5:22Guys do you think this is real or bot/fake? And explanations thanks a lot.
Agamemnon5:22it'll probably turn into a bunch of spam
Agamemnon5:22but people might add their view and reasoning
tjenyao5:23Lol HD [/spoiler]

[DAS44]

To my eye it's not legitimate play. It's far too smooth, off the top of my head I saw no hesitations. Plus it's nigh impossible to go that fast on a TB game to my knowledge (heck he's going faster than my null 40L record of 3.46 pps).

I'd defer to somebody more technically informed like Kitaru, but for now I'm calling bullsh*t

EDIT- Check out around the 23 second mark on the second video (kick it at 20 if you're that impatient, it moves by fast), he spins it in while it is facing the wrong way... Or the play is so far sped up that we can't visuall register the correcting spins.

[Barneey]

I'd say they're hacking illegitimate based on the 210 apm~ in his Infinite TST video, he's also going extremely fast, especially considering this is Tetris Battle, meaning there is no instant DAS (correct me if I'm wrong).
On top of that, he claims he invented the Infinite TST setup, I do not approve.

edit: errors just woke up > : |

[DAS44]

I posted my thoughts on the second video. Upon further review, I'd sooner ice skate in hell than believe this was legitimate play.

[arf]

i downvoted the videos because of their fakeness.

[MarioThePhenom]

To fake,wierd things were done like the kick DAS mentioned,and also in the first bid,he went to slide in the I piece into a MD,he rotated it before he started to soft drop...which doesn't make sense cause it's faster to let it soft drop before rotating it

[benmullen]

I have given this guy a "written warning" haha.  Yeah, even an old NES player like me can see they are fake.  Just lacks normal human hesitation that you see even in the microblizz record and lipsi, and lacks any of the logic in battle you see from blink and hebo and the like.

Not only fake, but laughably so

[caffeine]

Watched some of it in slow mo, and it does indeed look a little cooky at times. He possibly played through at a low FPS. Then again, perhaps he possesses superhuman Tetris abilities that we haven't seen yet.

[Integration]

Well I think, such a video is hard to edit because of the sendline effects, so I would say, it's a bot exploiting a fault of Tetris Battle's multiplayer system: You don't have to respond to the lines sent by your opponents. I don't think, the bot can handle all situations.

The channel has also some videos, which look more human to me:

[myndzi]

^^ The replays are from different people. The Youtuber who posted these probably just found them impressive and doesn't know better with the bot-vids.

What gives it away for me is the instantaneous transitions and the regularity of play. For example, he slides the I in the first video pretty much the frame it gets into position, pieces move instantly out the gate 100% of the time, and there are some 1 or 2 frame finesse placements. Even the fastest real players have around 2-4 frames delay before manipulating a piece out the gate, and slides/kicks require reacting to the screen (where playing with finesse technically doesn't) - so there should be around 6ish frames on that kind of input for an extremely fast player.

Couple that with the fact that the setups here are very formulaic, which is much easier to program into a bot, and it's pretty much a guarantee.

[caffeine]

Couple that with the fact that the setups here are very formulaic, which is much easier to program into a bot, and it's pretty much a guarantee.

I would like to go ahead and say that I don't think it's likely that this is a result of a bot. There are situations where he changes his mind, starts to move and then holds, and slight hesitations here and there. It seems more likely that he slowed down the game and played like that.

[myndzi]

I didn't really study them that close. I didn't notice any hesitations in particular, can you reference one of the changes-mind situations?

I did notice at one point it seemed a piece spawned and instantly dropped on the right which may support the speed hack idea.

[Sponch]

Fake until proven real

[caffeine]

can you reference one of the changes-mind situations?

In the messi4h TST vid, this happens at about 10 seconds:
[fumen]110@odA3nbB3hbB3jbA3ibC3gbD3hbG3ibG3gbI3hbH3gb?B3pbXPwUAFLDmClcJSAVDEHBEooRBToAVB7eG3mbPsbvrbP?rbvrbvMBAAA[/fumen]

At around 28 seconds: (Yes, it actually looks like a sonic drop in the video.)
[fumen]110@9bA3obA3obA3gbA3mbA3gbA3mbE3kbF3jbF3jbF3jb?F3jbF3jbH3hbG3ibG3gbI3hbH3gbI3gbI3hbH3gbB3pbWPw?UAFLDmClcJSAVDEHBEooRBKoAVB7eD2nb+nb+7be7b[/fumen]

At around 24 seconds, there's this very curious move:
[fumen]110@NdF3hbA3gbF3hbH3ibI3gbH3hbI3gbI3gbH3hbI3gb?H3gbB3pbt9wUAFLDmClcJSAVDEHBEooRBUoAVB7eF1Vc9Vc?9ac9fcdvcdvc[/fumen]

A couple other things I noticed was that he tends to DAS to the wall and then move twice to the other side (Tnet style finesse!). It's possible that that may've been the case for the first Fumen yet the video missed the frames that showed it. It also appeared as though he was sometimes skill-stopping at what looks like 60hz autorepeat (!!). Definitely seems like slow-mo type play to me. This warrants a hands+screen video if this guy expects me to believe it's real (in which case I'll humbly apologize for raising doubts).

[myndzi]

In the messi4h TST vid, this happens at about 10 seconds:
[fumen]110@odA3nbB3hbB3jbA3ibC3gbD3hbG3ibG3gbI3hbH3gb?B3pbXPwUAFLDmClcJSAVDEHBEooRBToAVB7eG3mbPsbvrbP?rbvrbvMBAAA[/fumen]

That one is wrong; it shifts before rotate, the reason it looks like it jumps to the left is because of the two-state nature of the s/z pieces.

Now that I'm looking closely, it appears to be possibly hypertapping pieces into place (instant changes in "DAS").

At around 28 seconds: (Yes, it actually looks like a sonic drop in the video.)
[fumen]110@9bA3obA3obA3gbA3mbA3gbA3mbE3kbF3jbF3jbF3jb?F3jbF3jbH3hbG3ibG3gbI3hbH3gbI3gbI3hbH3gbB3pbWPw?UAFLDmClcJSAVDEHBEooRBKoAVB7eD2nb+nb+7be7b[/fumen]
More solid, but could be a lag glitch or fast tapping of the down key. Youtube cuts frames, but I think it would have had to show one inbetween frame at least.

At around 24 seconds, there's this very curious move:
[fumen]110@NdF3hbA3gbF3hbH3ibI3gbH3hbI3gbI3gbH3hbI3gb?H3gbB3pbt9wUAFLDmClcJSAVDEHBEooRBUoAVB7eF1Vc9Vc?9ac9fcdvcdvc[/fumen]

This is a replay glitch, I believe. (In the TB replay recorder/player, that is) - you can see that the actions taken, if the piece was shifted one to the left, are the correct sequence to generate that clear. Having dealt with no end of replay headaches on KoS I can support this way of doing things - even if the sequence screws up, the result is correct.

You know what, I'm going to download this video and do a frame-by-frame

Edit: It does look like he uses firm drop. I suspect that maybe it's drop-move/rotate-lock and he's using hard drop. At 0:04 you can see a l33t J placement that most humans wouldn't use.

Edit2: lol at frame 114 the ghost piece is desynched from the actual piece.

The reason the pieces appear to jump around by two units is the Youtube reduction to 30fps. Even a speed hacker won't be frame-perfect, let's see...

Frame 122: movement the frame of/a frame after holding a piece
Frame 130: another move-after-drop scenario. Maybe the replay is wonkier than I thought; it could be that many of these placements are due to the replay recording the wrong input sequence. For example, in this frame it is DAS to left wall then tap back 2 (lolwut); the piece is shown at the floor and then it is shown locked to the right one column and down one row. Another possibility is that the bot(?) is exploiting the game's input handling and movements happen before lock after pressing hard drop. You can see that when this happens it always seems to be one move away. The J-slide is evidence towards this...

Frame 197-199: a clear delay in the placement of the O piece, but it's such an obvious placement I can hardly attribute this to player hesitation. Still, possible evidence of speed hacking vs bot.
Frame 206: A clear DAS skillstop. Doubtful that a bot would use both DAS stops and finesse, particularly the O double-tapback since a double tap is faster than DAS double tap.
Frame 324: Another drop-and-rotate. Anybody have the software to try speed hacking this? I'm quite curious now, I suspect this is the engine's work, probably stumbled into on accident WHILE speed hacking; it's the equivalent of a synchro in TGM.
Frame 330: The S in your first fumen. It does indeed come back, I assume this is a 2-tapback with the missing frame being the wall frame. Your tnet observation seems spot-on, and coincides with my comment that he'd have to be a fairly skilled stacker.
Frame 353: places an O with a double-tap instead of a double-tapback
Frame 356: rotates then holds a piece. Since the T needed to be rotated in that direction, I'm guessing he's using IRS and was hoping it'd work for unholding but it didn't.
Frame 396: This is a particularly interesting placement. It appears to use the harddrop+rotate trick with a kick; the J is dropped on the far right, rotated left which kicks it into the correct column(?), and lock-dropped on the right properly.
Frame 402: Looks like instant DAS; piece moves farther than one skipped frame can account for. I suspect it's a recording hiccup rather than anything game-related.
Frame 413: Another fast placement with too many actions; someone's CPU was running in this section...
Frame 521: A slight hesitation for no apparent reason.
Frame 657: The L piece spawns and is moved one to the left, then held. This cannot be accounted for by DAS preservation/leftover since the previous piece went to the right and the O is dropped straight down.
I think I've seen enough to conclude speed hack, but if so it is ridiculously slow or the player is pretty skilled already.
Frame 914: Rotates an I piece the incorrect direction, forcing an extra tap.
Frame 995: Soft drops the T to the stack before shifting right, possibly twice in a row. Shows intent, because even a speed hacker would be able to hold one key and tap another -- but he perhaps didn't know what would happen if he pressed both in the same frame so he waits a moment.
Frame ~1075: Soft drop/slides a J into place, rather than using the synchro trick.



Notes:
Our boy appears to always rotate right, so he isn't familiar with finesse but he is familiar with the way the pieces behave.
He does know how to stack though, and is practiced with this setup.
It's interesting to see the theory about the infinite TST loop's maintainability in action.
WTF, B2B TST sends TEN?!
If my hypothesis about the synchro trick is correct, repeated TSDs would be extremely fast! It would take the soft drop delay out of this.