Leela Chess Zero traps Torch
HTML-код
- Опубликовано: 16 окт 2024
- Featured is game 94 between Leela Chess Zero and Torch from Chess.com’s Computer Chess Championship 21. The opening is a Modern Defense, in which LCZero gains space on the queenside and secures the bishop pair. A pivotal strategic moment unfolds on the queenside when Torch plays 28…a5. Leela’s response keeps both sides of the board in contention. What follows is a rook trap remarkably similar to the one in a previous game where Leela trapped Stockfish’s rook in the same event. The formula to convert remains the same: identifying a convenient setup to maintain the trap and then open the opposite wing.
I'm a self-taught National Master in chess out of Pennsylvania, USA who was introduced to the game by my father in 1988 at the age of 8. The purpose of this channel is to share my knowledge of chess to help others improve their game. I enjoy continuing to improve my understanding of this great game, albeit slowly. Consider subscribing here on RUclips for frequent content, and/or connecting via any or all of the below social medias. Your support is greatly appreciated. Take care, bye. :)
★ CHESS.COM www.chess.com/... (affiliate link)
★ LICHESS.ORG lichess.org/@/...
★ TWITCH / chessnetwork
★ TWITTER / chessnetwork
★ FACEBOOK / chessnetwork
★ PATREON / chessnetwork
★ DONATE www.paypal.com...
your computer videos are my absolute favorites
Jerry’s commentary somehow gives a human touch to these computer games!
Loving all the Leela coverage. Thanks Jerry. 😊
👍
I love how you show the Lichess analysis graph, So basically, StockFish is sitting there, analyzing their game and go "Ah these idiots, they had a few inaccuracies, I would never have done that!"
12:12 I love how you put an actual clown where the rook was, LOL
lc0 games are some of my favorite for you to commentate!
What a hoot how these engines keep getting their rooks stuck. Thank you for the chess lesson, Jerry. I really values these videos.
Just played a classical game and incorporated this idea of trapping the rook between two pawns. Thank you for the inspiration and educational content!
👍😎
I enjoy watching Leela's games. They are so instructive. They're also quite baffling. Thanks for the analysis.
A really fine idea from Lela trapping the rook once more. Thanks Jerry.
My lichess rating has gone up 50-100 points in blitz, rapid and correspondence since you've been back to posting regularly. I'm climbing out of the scrub ranks largely thanks to you
Very instructive analysis. Thanks a lot!
Thank you for watching 👍
Thanks for the excellent commentary
👍
Rook: sees enemy is LC Zero
Rook: *sweats nervously*
The control of blacks rook was tyrannical, never giving so much as an inch of possibility to free it without sacrificing the rest of blacks position, just brutal.
Wtf are u talking abouth you understand computer moves cmon bro 😂😂😂😂
Wtf are u talking abouth you understand computer moves cmon bro 😂😂😂😂
@@MislavIvkovic-sx8vdwe have engines to study lines and moves being played for us is a lot easier to understand than finding those moves and confirming they're good. Even if we don't understand the full breadth of the moves played a general statement like this isn't untrue.
Some computer moves are basically nonsensical to us. But this game isn't really one of those games where we can't understand the underlying principles.
Engine games are art.
Loving these, so interesting!
👍
My favourite part of this game is actually the king walk. Just casually walk over to a2 to hold a3, then casually walk back to e3 to support the f4 break. Chess played at the highest level really can be quite ridiculous.
These Leela videos are excellent
JERRY IS UNSTOPPABLE THIS WEEK
I'd like to see Leela's evaluation graphed. Cool game
The king to e3 supporting f4 break with all majors in and a knight is the most sadistic plan I've seen. Come and get me!
That was a wild game of fun, entertaining wild game
Rather than having time restrictions. Engine games should have move restrictions. Like finish the game in unter 50 or 100 moves. Than we would probably see more humanlike play.
Fuck the humanlike play let the engines play at their full strength
What a fascinating game! Thanks Jerry.
Thank you for sharing.
And thank you for watching. 👍
Such a beautiful pawn sacrifice from Leela ❤
Wow! I always like your high level of educational videos!
Amazing how these bots convert material into a positional advantage. That EG was brutal 😂
very interesting game
More Leela, mooorrreeee
Leela's king's placement on f4 was very satisfying to watch. It's not often we see queens and rooks on the board being this powerless.
Most people would definitely blunder into a draw in that chaotic endgame though.
Maybe titled players could figure it out in classical time and still come out victorious.
I wonder if you could win that endgame the way Leela did, given enough time Jerry :D
that was just rude leela, taking that knight at the end
Hi Jerry!
👋
Jerry Jerry Jerry Jerry.
Anihao great game Jerry
Thx Jerry 😊
JERRY FOREVER
This rook trap pattern needs a name. Pincer?
It's kinda like a rook sandwich.
I like how at the end there Leela doesn't bother with blockading the passed pawns and making the king run around to defend or escape potential checks. White just sacs its queen for a pawn to expedite the removal of material from the board so it can simplify the position and proceed to an even quicker winning endgame.
Hi Jerry.
I have gotten my rook trapped like that before in online games. I feel just as stupid every time. Crazy that top engines don't understand that it isn't positionally sound. Any 1800 could tell you that and be right about it. If you can't calculate all the way until the rook is freed, it's just a blunder. That's the only sensible approach to having your rook in a pawn sandwich.
The engine's response to your "logic":
Beep boop
@@alksjdlkasjdlkajsld9189 I mean, Leela gets it. Gives a huge negative score when the rook is sandwiched in the end position of the line that's evaluated. Stockfish and Torch don't.
I'm a hobby chess programmer. I know how these engine's work more or less. My engines would never be able to recognize and penalize something like this either. It's an edge case as far as I'm concerned, not all that relevant to bother with. For top engines, you'd think they managed something so "basic".
Leela's eval function takes a lot more computing power which is why it's more positionally sound. At the cost of raw calculation skill ofc.
Whoa that's crazy and pretty cool. So Leela basically made the correct decision when other engines, even Stockfish, got it wrong. That's pretty cool
@@alksjdlkasjdlkajsld9189 Yes, take the position after 37. Ka2 at 10:20
Stockfish 14 thinks black is slightly better, but probably a draw after 37. Nxb5 and letting it's rook get trapped. It thinks it's the only way to get an advantage.
Stockfish 16 thinks white has a slight edge, but it also plays 37. Nxb5 at 38 depth which is deeper than this format allows I think. Didn't bother to run it longer. It thinks it's likely a draw.
LCZero 0.30 thinks white is significantly better, plus over minus. Obviously it refuses to take the knight. If you play 37. Nxb5, Leela thinks it's winning for white.
inspirational... neutralising oppotents
Ups - i thought its the old video watched already but it says 23 hours ago so it must be some other lela :) - LiCk0_zero never get older :)
Like a French, often closed. Engines usually play poorly in closed positions.
Definitely a computer game. It was all really long term calculations.
maybe leela always tries to win in 115 moves. more often than not it doesn't work however. it's like climbing a mountain with extra weight.
Wonder why at 15:01 black doesn't sacrifice a pawn to free the rook?
Qe5 forces a queen trade and sacs the pawn on d6, which should set the rook free.
i am very weak but i think after the queen trade, king captures pawn and the trapped rook still has no free square to go, since all squares on the row are covered by pawns, bishop and the king.
in other words after black sacrifices the pawn to the white king, black is 1 move away from losing his rook
Ra8 Rb7 Qe5+ Qxe5 dxe5+ Kxe5 Re8+ Kf4 Rd6 a5 and even tho the rook is free those passed pawns on the queenside are lethal as the rooks are behind them and the pawns are advanced.
In that position, white is also up a pawn, the black knight is useless and the rooks can't activate due to the excellent placement of the white bishop.
@@agar0285 a8 Rb7 Qe5+ Qxe5 dxe5+ Kxe5
After that I was thinking: f6+ and then either the king retreats or g5xf6 and Rb6xf6 freeing the rook and being down a pawn.
But I am sure I am overlooking something else - the engines definitly see it all. So I wonder what they see that is so much worse than losing a whole rook down the line.
@@testthewest123 your line makes a lot of sense but I assume it was easier for the engines to calculate to the end of the game after something like rxf6 d6 and the pawns on the queen side and center race down the board quite quickly
Very interesting game with the trap.
I'm also starting to notice that computers rarely have sexy mates, they kinda play on like there's 500 theoretical moves left and then they run out of pieces. It's like their rating drops to 700 and they don't know how to mate.
what year is it?
does "21" mean 2021?
Season 21 of the CCC
@@AlanTwoRings -_-
WHAT YEAR WAS 21ST SEASON OF CCC IN???
@@NoNameAtAll2at 0:18 on the left is says “Played on October 4th, 2023”
@@Sohrac thanks!
i wish i could use my king to attack...
Torch beat stockfish by trapping the rook the same way. Weird that it was missed in this game
Torch did it too? I thought it was lila only who trapped the rook against stockfish
❤ hello
💖💖💖🙏🙏🙏
tough timing! I was gonna comment "first smiley face 🙂" but moesheri beat me to it.
🙂🙃🙂
Torch noob omg
is this guy ai?
Honestly, Jerry, there is a huge SUBTEXT to all these AI, Engine, Computer games.....
What happens when you let ALPHA- ZERO self-play/learn
for ANOTHER 6 hours???
.....
or another 6 WEEKS????
theres something called overfitting in machine learning.. which means if you train too much you become worse at predicting new data.
not sure if it applies to this type of ML.
If it doesnt, then there is a law of diminishing returns, so it wouldnt get much better
Alpha zero war not special, there is no reason to believe it would out compete any modern engines. It was just the first time a neural engine was used to let machines learn about what is good for chess positions """instinctively""". It was the first time engines tried to win by using and restrictitng the enemies space and piece mobility in ways that exceeds human capability. Stockfish already was beating A0 the last times it competed and tthere is not reason to believe it would get close to todays engine.
I thought there was a lot of shifting of pieces in the middle game, I usually see exchanging pieces to open up maneuverability more often than not.
the only traps i love better are Jerry's~! 520 views. 51 likes. 10%? geez come on~!👍
What a fascinating game! Thanks Jerry.
What a fascinating game! Thanks Jerry.
What a fascinating game! Thanks Jerry.