0:00 Hello everyone 1:02 1:25 Completely new game 2:42 Sneaky tactic 2:46 Give you a couple of seconds 2:57 Enjoy the show 3:42 Suffer a terrible fate 4:40 A lot of madness 4:55 Triple attack 5:12 Nasty discoveries 5:22 Lastly but not leastly 6:57 Double up on the C file 7:55 13:10 13:27 Captures captures 8:25 Slicey bishops 10:17 Lol 18:38 Bad arrows 19:03 Leela gif 19:57 It was in this position 20:30 Completely lost 21:13 Morphy saga
Maybe. I mean, it's a pretty rare endgame, and as leela is not taught but figures stuff out by playing, she would have had to figure it out within 50 moves before a draw. I mean, she probably could have, but who knows man, neural networks are weird sometimes. Can do amazing things but fail at simple tasks.
@@kekkles4001 That's just Kasparov's opinion. How could you ever prove such a thing? At what point in this game, for example, would you ask a human to choose the move? How would you know if you had changed the result or not? How would you know if you had improved the position from what the computer would have done? You'd probably use engines to analyze it, which begs the question. These engines are nearly 3500 rating. These engines playing Carlsen or Fischer or Kasparov would be like these champions playing a 2200. Believing that humans introduce improvement is just our ego. We don't want to admit we have no place at the cutting edge of chess anymore, but we insist we are still needed because we are biased and self-absorbed, as usual.
It's interesting that so many of Leelas and Alpha zeros wins over stockfish involves exactly this pattern: giving up material to develop a dominant white bishop pair, closing the position with pawns and black being up material but with useless pieces.
Honestly I think this says more about Stockfish than about its opponents. Stockfish is built on the fundamentally flawed assumption that a good move will lead to material gain or checkmate within the foreseeable future, but these games inarguably prove that this is not the case. Leela and Alpha beat Stockfish by playing positionally, which is something Stockfish can never do
Leela points out that it’s not wrong for HER. Humans are incapable of keeping that many balls in the air without losing track of something-it’s just way too complicated to calculate. If you’re a GM you’re playing with fire. If you’re lower rated blundering is inevitable.
The game doesn't show us what moves Leela would have done as black, but the fact that she played the (non-book) moves she did as white means she believed all responses for black were losing.
2:30 4 minutes explanation about why you can't capture the pawn, but should I ever encounter this position, my brain would instantly think *oooh! a free pawn!*
4:25 I don’t know if you’ll read this comment, but I really appreciated how you went into much further detail than usual on a not completely obvious line. Often you tend to leave them when a certain side is winning but not up material, and as a newer player I really appreciate you going into a bit more depth 4 or 5 moves ahead to why the position is so winning. Hopefully you replicate this in upcoming videos :)
@@Radjehuty THAT win, with black, was amazing. With white I think there's at least 10 other similar games where Leela beat Stockfish in similar fashion, against the French and other openings where white gets a space advantage that I've seen so far.
One sugestion from me: You should show both games for TCEC finales with predefined opening moves , that way we can see both engines how perform on same opening.
One don't need to be a supercomputer to get in bad positions like this while playing black! I get positions like this 80% of the time while playing black and trying to close the position.. x.x
@@freedomofspeech2700 No, I'm pointing out the absurdity of your comment, and you've revealed a lack of awareness. (Google "Magritte" + "The Treachery of Images" to learn what I was referencing... "This is not a fun channel" is a dumb comment to make on a fun channel.) Your comment was dumb, and I didn't want to get into it, but you appear to be a chess player. So let's play: By any measure, "this is not a fun channel" is a nonsensical comment. If you meant that with regard to "this channel isn't fun," then why are you watching? If you meant that "chess is serious, and you shouldn't make jokes here," then recall my comment was a reply to Agadmator making a joke. If you meant "this channel isn't about entertainment," then you're mistaken-Agadmator has explicitly said that this channel is in part for entertainment, and that if some GM's watch his channel it's "for entertainment," and not the analysis. If you meant anything else you're not thinking clearly, and playing a silly move. ...And it was in this position that Shivam Shrivastava resigned the game, as there is nothing more to be done here.
01:00 Black plays b6. White can play c3 so then Ba6 is not possible due to Bxa6 Nxa6 Qa4+ winning the B on a6. But what do I know? Leela goes h4. White plays h4, then castles kingside? What is happening. 01:53 White sacrifices a central pawn. Wow. Did not expect that. White gives up d4 pawn to control d4 square with a Knight. Black is up a passed d-pawn, but it cannot move. Very instructive display by Leela.
I'm on Ilos right now during my current playthrough, no need to resign the game in order to wait for Legendary Edition. In fact, with agadmator there is already a legendary Edition chess channel on youtube.
Ahmad Sabbagh if i never pause the video and always find the move, and you always pause the video and never find the move; then, there exists someone who’s an agadmator subscriber and always pauses the video and always finds the move. The demorganian in me had to do this. Also, I am about to list all the possible solutions but I am sparing youtube that.
Thanks for your video. At 20:00, what is the white move to prevent a draw because insufficient materials? If white captures knight, then black king takes white pawn and if white does not take black knight, then black knight will move to c4 or c6 to attach white pawn.
Perhaps Leela gives back material because in her self learning process the number of moves until a win was less when simplifying the position first. Keeping unnecessary material for a clear win risks stalemate or blocks certain faster combinations so giving it back must have shown faster winning chances. Alternatively, since this appears not to be the case, when every path is a win, Leela chooses one randomly and cannot decide between wins since a win is a win. So whether it takes 100 or 5 moves the engine cannot differentiate since it's only goal is an end result win. Certainly the win paths should be trimmed for the shortest one. The goal should be winning quickly not just winning... maybe more draw possibilities that are spurious are also throwing it off. Less pieces would be less draws. Regardless it feels like the tuning around this is not ideal. It cannot be an optional parameter either as this is fundamental to the training that already has occurred. The engine is not flexible if any slight variation requires expensive retraining. Anyway what a great game
At 10:30, instead of 45. ... d4, why not (45. ... Rc8. Then, if [46. Ba7 Kc6.] Or, if [46. Rxc8 Bxc8] ) or other variations? Therefore, the Knight can escape and through trading down, the advantages of white are a bit nullified...
Please provide game 95 as well where Leala beats Stockfish as black with the same opening. I would really like to hear how Leala beats Stockfish with a french defence
How long typically do 1. the chess engines take to make a move and 2. these chess engine vs chess engine/AI games typically last compared to e.g. a GM vs GM game?
How long they take to move depends on the time control obviously. The Season 17 superfinal in this case (90 min + 5 sec increments) had an average game time of 3 hours and except the first couple of moves, stockfish tipically moved every 50 sec or so in the middle game
14:18 white could have played Nf6 to force a Knight trade and white will have a passed F pawn. The king will be stock preventing the passed F-pawn and the Bishop will be forced to capture the passed A-pawn in the future
Did you keep watching? The knight would immediately take F4, which Leela wanted to play even with sacrifice this pawn (You can see it later in the game).
