hiya jerry! I have not found a better channel than yours when it comes to analyzing chess games. you explain in great depth strategic concepts, and care also to explain why some moves were NOT played, which is just as equally interesting and important. I think i have seen practically every single analysis videos you have uploaded over the past 10 years in the space of around 1 year! great stuff. couple of questions though - right now we have an incredible tournament going on, with the best players in the world participating, why on earth are you analyzing games between an engine and a GM? obviously one can learn a lot from the videos, but its kind of like during the FIFA world cup final, for one of the best sport commentators in the world deciding to commentate on a local football game in some town. I think most of us would much rather you cover the top games going on at the moment! second of all, when it comes to chess engines, we are talking about hundreds of thousands, or even millions of NPS - nodes per second, i.e positions the engine evaluates per second. when it comes to time, there is no meaning for an engine between a human classical time control or a human blitz game, they are relatively the same. if time is an interesting factor in your opinion, then the time one should be giving the human is indefinite, and the machine - a classical \ blitz game from THEIR standpoint, i.e fractions of seconds per move... this whole classical time control thing is a bit silly, is what im getting at. no need to mention it because it has no meaning. notice how leela ends up with much more time at the end of the game of course. spent a total of a few seconds for an entire game probably.
Leela usually gets better positions in the opening giving g1 knight than giving b1 knight (this game was an exception. But actual results are about 96 to 97% either way for Leela. In this match Leela broke even giving g1, was minus one giving b1. I'd say that is typical with strong opponents.
While the proposed 960 match with Magnus would be great, it is unlikely to happen since Magnus never plays engines on record. I think the next match should feature rook odds, which is not as much greater than knight odds as the nominal piece values of 5 to 3 would suggest. Actual Leela results show it to be more like 4 to 3. So far strong IMs and "par" GMs, including Joel Benjamin, have done poorly in Rapid with Rook odds, so I think it would be a tough challenge for Jerry at his preferred 15'5" TC! Incidentally, today World number 2 Caruana played nine blitz games with it at g1 knight odds (3'2"), and only managed to score half a point!!
You know what I would like to see? Remember the WorstFish project? Where Stockfish is programmed to play the worst mathematical move on both sides? However, much like Stockfish for finding the best move, WorstFish is a brute force calculator in finding the worst mathematical move available in the position. However, because of this, if both sides use WorstFish, there won't be any constructive plan; blunders are made every single move with no sort of coherency. What I would like to see is a version of Leela trained to come up with constructive plans to force a loss, instead of incoherent blundering like WorstFish. In other words, a WorstLeela bot. WorstFish vs. WorstLeela would be an awe-inspiring spectacle to witness; a brute-force loser versus a methodical, humanlike loser.
@@sonicast5631 I feel like it would be a lot more like AlphaZero vs. Stockfish, but for losing instead of winning. Stockfish 14 NNUE (the version used for WorstFish) is a lot stronger than Stockfish 8 (the version used to fight AlphaZero), but it still has much of its brute force calculation. So while WorstFish makes incoherent blunders (similar to how Stockfish makes incoherent materialistic choices), WorstLeela would find a constructive plan to eventually force Stockfish to grab her material.
@ Fair point that Leela is definitely a lot stronger than A0, but you need to re-read my previous comment. The current WorstFish is a fork of Stockfish 14, which is not the strongest version of Stockfish. A version of Leela would need to be trained to lose, and be able to do so against Stockfish 14 at its worst.
With normal chess knight odds, we estimate that Magnus would pull ahead between Rapid and blitz, say 5' plus 5". Chess960 aka Fischerandom aka FreeStyle is clearly more favorable for the engine, so the break-even time would be significantly more, around 15'10" I would say. So I would call such a match a tossup. We actually had one such match in 2020, Komodo vs GM Alex Lenderman, it was pretty even. Leela knight odds is much stronger at this than Komodo was then, but Magnus is much stronger than Alex, so maybe also a close match.
Leela Chess Zero (LCZero) chess games
th-cam.com/play/PLQsLDm9Rq9bFV5lw1FZa3bThJpKXJyk9O.html
Congratulations Mr Benjamin! Outstanding game, imho, and an impressive win! Thank you Jerry for this one!
Shots fired at Magnus at the end there!
no shots. simply props (to leela knight odds)
Great analysis, Thanks Jerry!
I'm a huge fan of this channel 😎
Thanks 👍😎
@@ChessNetwork 🤍🙏
Thank you for adding the letters and numbers on the chessboard for me. 👍
The honor of humankind has been preserved. For now.
I would love to see Magnus vs Leela
hiya jerry!
I have not found a better channel than yours when it comes to analyzing chess games. you explain in great depth strategic concepts, and care also to explain why some moves were NOT played, which is just as equally interesting and important. I think i have seen practically every single analysis videos you have uploaded over the past 10 years in the space of around 1 year! great stuff.
couple of questions though - right now we have an incredible tournament going on, with the best players in the world participating, why on earth are you analyzing games between an engine and a GM? obviously one can learn a lot from the videos, but its kind of like during the FIFA world cup final, for one of the best sport commentators in the world deciding to commentate on a local football game in some town. I think most of us would much rather you cover the top games going on at the moment!
second of all, when it comes to chess engines, we are talking about hundreds of thousands, or even millions of NPS - nodes per second, i.e positions the engine evaluates per second. when it comes to time, there is no meaning for an engine between a human classical time control or a human blitz game, they are relatively the same. if time is an interesting factor in your opinion, then the time one should be giving the human is indefinite, and the machine - a classical \ blitz game from THEIR standpoint, i.e fractions of seconds per move...
this whole classical time control thing is a bit silly, is what im getting at. no need to mention it because it has no meaning. notice how leela ends up with much more time at the end of the game of course. spent a total of a few seconds for an entire game probably.
Joel Benjamin was one of the people who helped program Deep Blue to beat Kasparov. I guess he knows computers!!
Benjamin played a 72 mover against a 3000+ opponent with 0 mistakes, 0 inaccuracies and 0 blunders. He's superhuman.
Nice match. About the end "freestyle chess" is a good term for when you get to set up the pieces however you like. Otherwise it's a misnomer.
Great game by Joel Benjamin 👏
I wanna see you playing Leela.
He has a few months ago in rapid.
Incredible game by the GM, held the advantage throughout 😎
Would love to see your suggestion!
Thanks, how much stronger do you think Leela would be with the King knight as opposed to the Queen knight as is the case here?
That’s really tough to know. It feels like Leela has an answer to everything.
Leela usually gets better positions in the opening giving g1 knight than giving b1 knight (this game was an exception. But actual results are about 96 to 97% either way for Leela. In this match Leela broke even giving g1, was minus one giving b1. I'd say that is typical with strong opponents.
hi jerry! thanks!
While the proposed 960 match with Magnus would be great, it is unlikely to happen since Magnus never plays engines on record. I think the next match should feature rook odds, which is not as much greater than knight odds as the nominal piece values of 5 to 3 would suggest. Actual Leela results show it to be more like 4 to 3. So far strong IMs and "par" GMs, including Joel Benjamin, have done poorly in Rapid with Rook odds, so I think it would be a tough challenge for Jerry at his preferred 15'5" TC! Incidentally, today World number 2 Caruana played nine blitz games with it at g1 knight odds (3'2"), and only managed to score half a point!!
You know what I would like to see?
Remember the WorstFish project? Where Stockfish is programmed to play the worst mathematical move on both sides?
However, much like Stockfish for finding the best move, WorstFish is a brute force calculator in finding the worst mathematical move available in the position. However, because of this, if both sides use WorstFish, there won't be any constructive plan; blunders are made every single move with no sort of coherency.
What I would like to see is a version of Leela trained to come up with constructive plans to force a loss, instead of incoherent blundering like WorstFish. In other words, a WorstLeela bot.
WorstFish vs. WorstLeela would be an awe-inspiring spectacle to witness; a brute-force loser versus a methodical, humanlike loser.
In that case I assume that Worstfish would be better, as it calculates forcing lines better
@@sonicast5631 I feel like it would be a lot more like AlphaZero vs. Stockfish, but for losing instead of winning. Stockfish 14 NNUE (the version used for WorstFish) is a lot stronger than Stockfish 8 (the version used to fight AlphaZero), but it still has much of its brute force calculation. So while WorstFish makes incoherent blunders (similar to how Stockfish makes incoherent materialistic choices), WorstLeela would find a constructive plan to eventually force Stockfish to grab her material.
@ Except that currently Leela is way stronger than Alpha Zero and if you look at recent TCEC results, Stockfish is the top engine
@ Fair point that Leela is definitely a lot stronger than A0, but you need to re-read my previous comment. The current WorstFish is a fork of Stockfish 14, which is not the strongest version of Stockfish. A version of Leela would need to be trained to lose, and be able to do so against Stockfish 14 at its worst.
impressive. He kinda played like Leela.
Magnus will wipe the floor imo
With normal chess knight odds, we estimate that Magnus would pull ahead between Rapid and blitz, say 5' plus 5". Chess960 aka Fischerandom aka FreeStyle is clearly more favorable for the engine, so the break-even time would be significantly more, around 15'10" I would say. So I would call such a match a tossup. We actually had one such match in 2020, Komodo vs GM Alex Lenderman, it was pretty even. Leela knight odds is much stronger at this than Komodo was then, but Magnus is much stronger than Alex, so maybe also a close match.
Hi Jerry, this is everyone.
thats all 3 types of games yeah
ggg