@@KevinJDildonik Great synopsis! I kinda feel this thinking is probably relevant in SO MANY OTHER human domains that we all take for granted from past humans. For example, our infrastructure of roads, buildings, water, electricity, etc. is usually set up in EXTREMELY ridiculous ways that were the results of additions on additions from prior human generations who knew nothing of engineering or the technology that would be in the future. It's almost useful if we could wipe society clean of certain ideas and give a blank slate to all humans now, so that we could think clearly and freely and SEE what's actually true instead of relying on tradition and dogma. Another reason why A.I. is going to be the fucking BOMB.
It's like a return to the romantic era of chess - but with ruthless precision and sound play. Sacking pawns left and right, rampant piece activity, flashy, razor sharp games. I love it.
AND! What's more important - he can do that with on a such a high level play. Usually, it's pretty hard to do something "romantic" at the GM's level. All theory has been learned and all positions, that people can think of, was anylyzed. You do something wrong - you lose. But this thing right here just don't give a fuck! He can throw all your theory straight to the window and show how to play chess! It's just boggles my mind how people are limited. Hundreds years of playing chess, thousands people and great minds was trying to find the best moves and tactics - all that can beat a machine, that played chess for 4 hours. People can't match machines at anything. If they will have such minds at every aspect of our life, we will not need them anymore. But yeah, that chess game was insane!
@@z483gk It brings humanity back into the game. An engine is only as good as its valuation algorithm. A0 says: screw your valuation, I've learned how this game is played. GMs in turn can learn from A0 and bring the dash and romance back to the game. .
yeah if you are a machine playing with a bazillion cores. these midgames are getting way to complicated, everything is attacking everything. humans would look at it for days and not find the best move. still a bliss to see stockfish get his ass handed. that fuckin engine is unbeatable if you aren't a chess ai or top engine.
This series is my favourite analysis. Both the gameplay and commentary are stellar. Years later i still come back to these videos on the alphazero games.
I've watched a lot of them. It seems that we dont value piece activity enough. Alpha zero really loves open files and diagonals. Itll sac pawns all day long to get an open file in line with the king
I’m so curious to see what Stockfish’s evaluation of this game was at every move. Surely it must have thought it was doing well after up a few pawns, even with AZ’s strong pieces....makes you wonder if the top humans have been relying on the engines too much for “objective” analysis of their games.
Reminds me a lot of Morphy, Anderssen and the romantic gang. No problem in sacrificing to open up lines, bringing every single piece into the attack and creating so many problems due to simply active piece play that the opponent is forced to give back huge amounts of material in order to simply not ouright lose. Sure, Stockfish might refute Morphy, but it definitely has trouble refuting AlphaZero playing like Morphy.
Christoffer the point of the comment is that Carlsen and Caruana played more like boring computers than Alpha Zero. In fact if you followed Stockfish’s evaluation of the games, the two players were nearly perfect, which unsurprisingly resulted in 12 draws. Amazing that a computer program can make chess exciting again! Those types of sacrifices with no immediate justification probably haven’t been seen at the top level since maybe Kasparov? Or maybe even Tal?
@@AlexWyattDrums alpha zero is nothing. Kasparov was nobody. Tal was nobody. You should see my games. I can sacrifice my queen and my rooks with no immediate justification just for the sake of long term strategy. My games are exciting all the way. Sometimes I even sacrifice my king. My opponents never know what to expect. I am ruthless.
@@draeioskronos3721 Considering that checkers differ from chess mainly in the lack of variety in pieces, I say go definitely looks more like checkers than chess (and yes, I am aware that go has more possible moves involved, since the board is simply larger, etc., but still - when making analogies, at least keep them sane imo.)
@@Rocket-qg5jw There's a certain point and a certain level of play where not resigning is just a waste of everyone's time. For beginners, a lot of people might play out K+R vs. K or K+Q vs. K endgames to the end because they make mistakes, but if a grandmaster played on in an endgame that was clearly losing it's just bad sportsmanship.
Alpha Zero plays moves that are just sooooo good! Bishop B1 made me laugh out loud once again as its absolutely beautiful and crushing all at the same time. Alpha Zero is like Kasparov, but also 5 times stronger and simultaneously high on caffeine.
I love games like this.. It just shows that chess for humans is not dead, and in essence will be alive for a long time to come. Players have so much more to learn, and we have barely scratched the surface.
I somewhat question if we have the capabilities to learn these things. We have no way of playing 400k games and memorizing every last one as if it was the last game we played. AZ knows, we can at best guess whether a certain long term pawn sacrifice is worth it and our human brains will probably never have the computional power to ever know, forever being reduced to guessing.
@@testthewest123 We might never see a two or three pawn long term sacrifice confidently played by humans, but this game did give the foundation of chess a good wake up kick. If the last 30 years taught me something then it's that such a kick is enough for people to eventually find a way to break common sense. I truly believe that with time, and not even that much, a new strategy will come into being that allows even humans to decide straight up sacking a pawn will be worth it in the end. Some of the current moves in chess already are based on chess strategies and principles rather than true understanding, but that doesn't mean it's guessing.
@Nico K That's certainly true. Seeing an engine play like this successfully proves that it is possible to be successful by valueing activity and space higher than just material.
It's like Napoleon surveying the battlefield, deploying and redeploying his forces to expose and exploit the enemy's weaknesses. Probing with infantry, holding back reserves, using an artillery barrage to break up defensive squares and set up a cavalry charge, then sending in the reserves to take the main objective.
i get the concepts here but could never generate these ideas on my own! as Saddler said , chess from another planet !! thank u M saddler for this great tutorial/ documentary !!
I think by the time of the next world championship there will be a completely different type of chess being played. This is an excellent video series and brilliantly presented.
Sadly, since GMs can't afford to risk playing such lines they are incapable of fully calculating, it's unlikely there will be any deviation from the boring but solid standard quo.
If you watch all of the games thus far played between AZ and Stockfish you will notice that at some point in the game Stockfish basically becomes "confused". There is a point in space in time where it really doesn't matter how deep Stockfish scans, that Alpha Zero has played a move that is not in "any" book or database. This is the beauty. This is why its an AI.
Thanks so much for your brilliant commentary, Matthew. You are a genius at making the complexities understandable with your wonderfully simple explanations! Chess needs great teachers like you!!
Sacrificing pawns, manoeuvring all over the place, while simultaneously tying up your opponent, and then using one of your last lonely remaining pawns hiding in the brushes all along to waltz in an finish the opponent off. Love it.
"Alien" describes it best, and it is true when you watch the games of Go, Chess and Starcraft. Until now, we have created machines that could lift heavy loads, or go very fast. But Life was still the Queen of the Intelligence realm. Deepmind has gone beyond the "sound barrier" here with something both scary and fascinating.
Thank you, Mr.Sadler, it's a pleasure watching your explanations. Other channels need twice the time to explain the same ideas. This game in particular is the most stunning I have seen from A0 so far.
Alpha zero has saved chess for me, the staid positional games that get played nowadays with book moves into the thirty forty move territory ,down to the use of training with programs might just be weakening its grip.
What I never heard anyone talk about AlphaZero was its ability (when it's a computer and can examine many many more positions than a human in a shorter time) seems to attempt to simplify the position. Now, you might think that I'm crazy for saying this because AlphaZero makes so many moves that seem to complicate rather than simplify. However look at the trend in the games. Its a common theme that AlphaZero just keeps restricting the options of its opponent, sometimes even causing zugzwang type positions in the midgame. I guess this shouldn't really be a surprise, its obvious that you should minimize your options while maximizing your options. . . but that's my point. I think, since AlphaZero is not a traditional chess engine, it can factor in the simplified position of a game as an aspect of the game that a traditional chess engine cannot or might not have the ability to "understand."
What's interesting and scary is that AlphaZero was never told the "chess golden principles" we are taught as beginners, rooks are worth more than bishops in the end game, leave no pieces hanging, don't blunder your pawns, etc. It was simply given the rule book and 4 hours to study the game, so it was free from our human theory about chess, and investigated pure approaches. Stockfish however is the result of our chess culture, it simply computes X moves ahead and compares the resulting positions based on heuristics defined by human theory, like "controlling center squares is better than controlling unoccupied squares" or "a pawn mid-game is worth 198 points and a pawn end-game is worth 258 points".
Awesome commentary, and wtf was that game?! I've never seen anything like that. I've been avoiding watching engine games because I thought they would be dry and boring. This clearly was not and thanks for really taking us along for the ride with the colorful commentary! Definitely going to check out your other engine videos!
If a robot boxer knocks out Anthony Joshua or Wladimir Klitschko in 30 seconds flat and in 15 in the rematch, you don't need to wonder how Tyson, Frazier or Ali would have fared. The whole Stockfish version is a moot point at this level.
I think this is my new favourite alpha zero game. the way it dismantled the centre with pawn sack after pawn sack to get the bishops looking straight into blacks king is just unreal. totally out of this world
10:25 *Yeah* - Anatoli's cold iron hand getting a hold of the back of your neck, threatening to crush it pleasurably. I like that. 13:50 Exactly the moment Bruce Willis would have said: " *Yippee-kai-yeah, motherfucker!* "
This is a very pleasant video, and our host is someone I could listen to for hours. Clearly a towering intellect. But - like many analysese by masters - I find the comments very odd. Yes, I'm keenly aware that this is a strong sign that I'm a chess idiot. But, gimme a moment. Our commentator says "So ALphaZero keeps his king nice and safe in the corner". Except - black's king is also "nice and safe in the corner" with more pieces around it, and a square to retreat to if things get awful. Another one - when white occupies the G anf H files with rooks, it's a clever move, but when it abandones those files, it's also a clever move and somehow doesn't make the previous strategy a waste of time. It seems like so many of the comments are not about the intrinsic moves or positions, but are informed largely by hindsight. Gulp. (ducks)
Despite its terrific record I think that AZ way of playing is still within what humans somwhat already theorized. AZ, however, takes it to the extreme. While this game in particular is extraordinary (perhaps the one where AZ showed the most dominance) it is still based on the concepts of space advantage and piece activity. It "just" happens that AZ can confidently prove its point about these "intangibles" over the mere material aspect of the game and do that with the accuracy typical of machines. By further analyzing this game (and other AZ games as well) one can really appreciate how your own pawns are often more of a liability than an asset and how this assesment only varies based on your tempo advantage/disavantage. Basically, what I get from this kind of play is that the so called "pawn sacrifices" in this game are in fact moves that strictly and consistently improve white position, without a single doubt. Also another recurrent theme in all AZ games I've seen so far is the use of zugzwang based tactics to prove its point about intangibles being more important. Again, a very easy concept to grasp for humans and yet not used enough due to the lack of calculating abilities (to foresee such a risky position where the simple order of moves counts) but also the lack of confidence to adopt such approach, as if it was only a secondary aspect in the game of chess. Apparently, the order of moves is actually very relevant for AZ to win games. With this I especially mean that he doesn't just dominate but also ends up finding ways to score the point in otherwise favorable but drawable positions.
Did he just say 7 pawn sacrifices? That is how u start a chess video 12:10 If alpha zero plays c6, I'm gonna lose it. 13:09 Matthew "I was disappointed with alpha-0, as it took a pawn back. I thought this was materialistic!"
I wish you would slow down in making moves and verbalize the move played. Like "alpha plays knight to B6" and offer some insight as to why. that would add more to your video. Making it longer yes, but worth the effort imo
I don't understand, black plays 4 moves with his g8N just to exchange it. It aganst the most basic principle of Nimzovich. Is it really theory? Is it posible that black was denied from his opening book?
I think the only rule alphazero does not have is a point score on pieces. It likes mobility and trapping S.fish behind its own pawns as it is willing to give away its own pawns for more mobility. Also a seamless middle and end game it plays.
What i don't understand about the Grand Master vs Deep Blue matches was if the computer was also allowed to think about moves during the player's turn. Surely this would have been unfair if it was the case. Does anybody know? Or is 1/2 the time just equal to 1 bit for a computer like Alpha Zero? Sez it would have been unfair because essentially for real players some element of the oponent's move must essentially be "waiting time" whereas for a computer this would not be the case.
About 60 seconds into this commentary, we're told it's a very flexible opening because "white has the two bishops and a lot of possibilities about how to deploy its pieces". Well maybe, but neither of the bishops can move!
Matthew calls this “one of my favourite games” and counts 7 pawn sacrifices from AlphaZero in total in “an absolutely fantastic attack”. He quotes DeepMind co-founded Demis Hassabis as describing this game as, “like chess from another planet”.
So I noticed that black basically went straight for a castle (king side). It's first 4 moves were all in that vein. Does this mean this was definitely a mistake? When you castle that early, isn't it supposed to be easy to attack? Now I know AlphaZero didn't calculate the whole attack, but you don't need to. If you know where the enemy king is going to be the whole game, you know in general what you want to do. I saw the potential attack after black traded the bishop for knight (easy battery with queen and black squared bishop at this point). Perhaps the super early castle was simply a mistake that we can learn from. I wonder how many of AlphaZero's wins where when the opponent early castled?
I don't think so. I don't imagine black castles long in this opening after some of white's early moves are expanding on the queenside. The kingside was pretty safe for black until black decided to win a pawn and open the g file. That could have led to a winning endgame if the white attack didn't work out.
2:45 Qh4 is -0.18 eval and my Stockfish wants instead g6 (-0.68 eval). So this move by the Stockfish they are using is half a pawn blunder compared to a not broken stockfish. Depth 33, I would love to know what depth was used for that move and compare leaving my stockfish to reach that depth and see the eval of both moves. Similar wrong moves compared to my own version was found also in previous game (I posted there too).
This AI conscious collective will make moves that are ones to counter the expected best move. It's like a false blunder. It may look like it is wrong but if the other player plays everything right by theory then that's when alpha's plan will shine. It counts on good play to be made so it can excel in doing these moves that no one will see coming until 25 moves later if it goes according to its predictions. All while calculating different variations in real time making adjustments if the opponent is not following a winning line. It's very complicated to understand but I see that alpha zero is literally just countering everything you do before you even do it.
I wonder if Alpha Zero just plays with the position, or if it seeks for exploitable flaws in the opponents play. You´ll know how to trick your opponent in balanced or stuck positions if you´ve figured out his pattern of play.
What you're describing is the difference between a player aiming to play optimally and a player aiming to play as a nemesis. These two different play styles are often seen in poker, for example, but they exist in every competitive game. However, nobody aiming to win consistently plays primarily as a nemesis in any game, especially chess. Our opponents will always adapt and become better, it's not favorable to handicap ourselves to assumptions on our opponents' mistakes. We can be confident AlphaZero plays from the first perspective and not the second as it trains against itself. AlphaZero chose to make these sacrifices likely due to the long-term benefits as shown in an analysis by Matthew Sadler here.
AZ has learned chess by being given the rules of the game and then by playing against itself and only itself for millions of times. This means that it never had the opportunity to identify and exploit Stockfish's potential intrinsic weaknesses. More importantly, Alpha Zero is called zero because it has ZERO knowledge of anything but the rules of the game. It has ZERO knowledge of whom it is playing against, ZERO knowledge wether it is playing against itself, another chess engine or a human. Because of this, AZ will not modifiy its playstyle with different opponents. Had AZ learned chess by playing only against Stockfish, then most certainly it would have found and used exploitable flaws that are specific to stockfish. However, if presented against another opponent, AZ wouldn't adapt, it would still play trying to exploit flaws inherent to stockfish.
AlphaZero basically said to play like Morphy. This isn't really groundbreaking other than it proves that Stockfish has a bad horizon effect. Humans will learn from this, because unlike auto racing versus cycling, where you can't put an 850-horseplayer engine into your stomach, we can copy the moves of the best chess computers. AlphaZero has given the green light to the romantic style of chess from the 19th Century.
Pawns are the essence of the chess..
AlphaZero: Wrong. They are just little a**holes blocking my rooks.
I know, right? Chess may never be the same.
Haha, so true.
It's pawns are the soul of chess
@@KevinJDildonik Great synopsis! I kinda feel this thinking is probably relevant in SO MANY OTHER human domains that we all take for granted from past humans. For example, our infrastructure of roads, buildings, water, electricity, etc. is usually set up in EXTREMELY ridiculous ways that were the results of additions on additions from prior human generations who knew nothing of engineering or the technology that would be in the future.
It's almost useful if we could wipe society clean of certain ideas and give a blank slate to all humans now, so that we could think clearly and freely and SEE what's actually true instead of relying on tradition and dogma.
Another reason why A.I. is going to be the fucking BOMB.
Yes, like solving poverty, healthcare, and many other things. Wonder if it can diagnose stuff.
It's like a return to the romantic era of chess - but with ruthless precision and sound play. Sacking pawns left and right, rampant piece activity, flashy, razor sharp games. I love it.
AND! What's more important - he can do that with on a such a high level play. Usually, it's pretty hard to do something "romantic" at the GM's level. All theory has been learned and all positions, that people can think of, was anylyzed. You do something wrong - you lose.
But this thing right here just don't give a fuck! He can throw all your theory straight to the window and show how to play chess! It's just boggles my mind how people are limited. Hundreds years of playing chess, thousands people and great minds was trying to find the best moves and tactics - all that can beat a machine, that played chess for 4 hours.
People can't match machines at anything. If they will have such minds at every aspect of our life, we will not need them anymore.
But yeah, that chess game was insane!
@@z483gk It brings humanity back into the game. An engine is only as good as its valuation algorithm. A0 says: screw your valuation, I've learned how this game is played. GMs in turn can learn from A0 and bring the dash and romance back to the game. .
Are these really typical pawn sacrifices? It seems that there is a far higher positional logic behind all of them.
yeah if you are a machine playing with a bazillion cores. these midgames are getting way to complicated, everything is attacking everything. humans would look at it for days and not find the best move.
still a bliss to see stockfish get his ass handed. that fuckin engine is unbeatable if you aren't a chess ai or top engine.
Absolutely CORRECT and true.
This series is my favourite analysis. Both the gameplay and commentary are stellar. Years later i still come back to these videos on the alphazero games.
It protec
It attac
But most importantly
Its going to bring Classical Chess bac
Why?
Why?
@@maciejnowak5251 it will bring an unpredictable style of play, outside the box, a style of play that goes beyond the theory that we know today
@@Alex-xi8ep well...
I have seen this in all alphazero games, it takes all movements from the other player, the other player just can´t move, it´s just amazing
I've watched a lot of them. It seems that we dont value piece activity enough. Alpha zero really loves open files and diagonals. Itll sac pawns all day long to get an open file in line with the king
It's like the perfect combination of karpov and Kasparov
AlphaZuqzwang
I’m so curious to see what Stockfish’s evaluation of this game was at every move. Surely it must have thought it was doing well after up a few pawns, even with AZ’s strong pieces....makes you wonder if the top humans have been relying on the engines too much for “objective” analysis of their games.
Just gotta get AZ to give us objective analysis.
"Pawns are the soul of chess." ~Philador
"Hold my beer."~Alphazero
It is nice to have been alive during the beginning of the end.
stupid humans with their stupid openings made me lose 6 games. Only one way to prevent this from ever happening again! *terminator soundtrack plays*
oh what *never knew you could type in bold*
_not just bold_
@@nicok8203 Wow the terminator soundtrack is sooo gooodddddd
*_terminator is bad_*
Reminds me a lot of Morphy, Anderssen and the romantic gang. No problem in sacrificing to open up lines, bringing every single piece into the attack and creating so many problems due to simply active piece play that the opponent is forced to give back huge amounts of material in order to simply not ouright lose.
Sure, Stockfish might refute Morphy, but it definitely has trouble refuting AlphaZero playing like Morphy.
Well said and AZ has proven that PM was correct all along that 'positioning' rules supreme.
this game feels a bit different from those in the recent Carlsen-Caruana World Championship match.
Engines aren't afraid to lose. Humans must try not to lose first.
Christoffer the point of the comment is that Carlsen and Caruana played more like boring computers than Alpha Zero. In fact if you followed Stockfish’s evaluation of the games, the two players were nearly perfect, which unsurprisingly resulted in 12 draws. Amazing that a computer program can make chess exciting again! Those types of sacrifices with no immediate justification probably haven’t been seen at the top level since maybe Kasparov? Or maybe even Tal?
You mean drawsen and drawuana
Alex Wyatt Not to forget chuky!!
@@AlexWyattDrums alpha zero is nothing. Kasparov was nobody. Tal was nobody. You should see my games. I can sacrifice my queen and my rooks with no immediate justification just for the sake of long term strategy. My games are exciting all the way. Sometimes I even sacrifice my king. My opponents never know what to expect. I am ruthless.
Wow, I am never resigning after blundering pawns no more
Mohamed Mahmoud Why would resign in any position anyways? The point of chess is not to give up until the end and keep on coming up with plans
You should try Baduk out too. Worlds oldest board game and it makes chess look like checkers.
@@draeioskronos3721 Considering that checkers differ from chess mainly in the lack of variety in pieces, I say go definitely looks more like checkers than chess (and yes, I am aware that go has more possible moves involved, since the board is simply larger, etc., but still - when making analogies, at least keep them sane imo.)
@@Rocket-qg5jw There's a certain point and a certain level of play where not resigning is just a waste of everyone's time. For beginners, a lot of people might play out K+R vs. K or K+Q vs. K endgames to the end because they make mistakes, but if a grandmaster played on in an endgame that was clearly losing it's just bad sportsmanship.
Jed Zhu Why is it bad sportsmanship? Maybe the player actually thinks that he still has a chance and why would he waste that small chance
Alpha Zero plays moves that are just sooooo good! Bishop B1 made me laugh out loud once again as its absolutely beautiful and crushing all at the same time. Alpha Zero is like Kasparov, but also 5 times stronger and simultaneously high on caffeine.
A0 is probably over 1000 elo stronger than Gary on steroids :P
@@nikpredom7095 Garry on amphetamines maybe. Garry on steroids would smash the table like Hulk! :oP
When did AZ ever play Bishop in B1 ?
I love games like this.. It just shows that chess for humans is not dead, and in essence will be alive for a long time to come. Players have so much more to learn, and we have barely scratched the surface.
I somewhat question if we have the capabilities to learn these things. We have no way of playing 400k games and memorizing every last one as if it was the last game we played. AZ knows, we can at best guess whether a certain long term pawn sacrifice is worth it and our human brains will probably never have the computional power to ever know, forever being reduced to guessing.
@@testthewest123 We might never see a two or three pawn long term sacrifice confidently played by humans, but this game did give the foundation of chess a good wake up kick. If the last 30 years taught me something then it's that such a kick is enough for people to eventually find a way to break common sense. I truly believe that with time, and not even that much, a new strategy will come into being that allows even humans to decide straight up sacking a pawn will be worth it in the end. Some of the current moves in chess already are based on chess strategies and principles rather than true understanding, but that doesn't mean it's guessing.
@Nico K That's certainly true. Seeing an engine play like this successfully proves that it is possible to be successful by valueing activity and space higher than just material.
@@testthewest123 yeah, especially on faster chess.
EXACTLY
I'd hate to be a pawn in AlphaZero's army. AZ may win but it's pawns have the life expectancy of a Star Trek redshirt.
and those which survive have the life expectancy of queen Elizabeth
It's like Napoleon surveying the battlefield, deploying and redeploying his forces to expose and exploit the enemy's weaknesses. Probing with infantry, holding back reserves, using an artillery barrage to break up defensive squares and set up a cavalry charge, then sending in the reserves to take the main objective.
i get the concepts here but could never generate these ideas on my own! as Saddler said , chess from another planet !! thank u M saddler for this great tutorial/ documentary !!
I think by the time of the next world championship there will be a completely different type of chess being played. This is an excellent video series and brilliantly presented.
Sadly, since GMs can't afford to risk playing such lines they are incapable of fully calculating, it's unlikely there will be any deviation from the boring but solid standard quo.
I disagree. they cant afford not to
Rote memory plagues classical chess. (The game's drawish nature is also a scourge.)
It looks more like thematic play than rote memory. New players will emerge that will use this principle and take over.
No way it's going to change that fast
I wonder what alpha zero would do if they switched places mid game
Same thought bro!
Feed pawns to stockfish while fortifying the positions of its central players. Somehow I doubt it would waste much time catching up.
Its remarkable how Alpha uses his his Bishops in a way that they are worth a Rook every game
And opponents white bishop won't get developed until it's too late...
End of era
...and the beginning of a new one.
Art.
I commented before you.
In art, there actually is no boundary because this piece comes from "another planet" :D
Another great vid! Matthew is so clear and knowledgeable. Definitely the best chess analyst out there imo.
If you watch all of the games thus far played between AZ and Stockfish you will notice that at some point in the game Stockfish basically becomes "confused". There is a point in space in time where it really doesn't matter how deep Stockfish scans, that Alpha Zero has played a move that is not in "any" book or database. This is the beauty. This is why its an AI.
This play is just too scary
Thanks so much for your brilliant commentary, Matthew. You are a genius at making the complexities understandable with your wonderfully simple explanations! Chess needs great teachers like you!!
Sacrificing pawns, manoeuvring all over the place, while simultaneously tying up your opponent, and then using one of your last lonely remaining pawns hiding in the brushes all along to waltz in an finish the opponent off. Love it.
I worked very closely with the Sadler human so be attentive to what he has says to say.
This is one of the most beautiful games in chess history
"Alien" describes it best, and it is true when you watch the games of Go, Chess and Starcraft. Until now, we have created machines that could lift heavy loads, or go very fast. But Life was still the Queen of the Intelligence realm.
Deepmind has gone beyond the "sound barrier" here with something both scary and fascinating.
Thank you, Mr.Sadler, it's a pleasure watching your explanations. Other channels need twice the time to explain the same ideas. This game in particular is the most stunning I have seen from A0 so far.
Alpha zero has saved chess for me, the staid positional games that get played nowadays with book moves into the thirty forty move territory ,down to the use of training with programs might just be weakening its grip.
What I never heard anyone talk about AlphaZero was its ability (when it's a computer and can examine many many more positions than a human in a shorter time) seems to attempt to simplify the position.
Now, you might think that I'm crazy for saying this because AlphaZero makes so many moves that seem to complicate rather than simplify. However look at the trend in the games. Its a common theme that AlphaZero just keeps restricting the options of its opponent, sometimes even causing zugzwang type positions in the midgame.
I guess this shouldn't really be a surprise, its obvious that you should minimize your options while maximizing your options. . . but that's my point. I think, since AlphaZero is not a traditional chess engine, it can factor in the simplified position of a game as an aspect of the game that a traditional chess engine cannot or might not have the ability to "understand."
What's interesting and scary is that AlphaZero was never told the "chess golden principles" we are taught as beginners, rooks are worth more than bishops in the end game, leave no pieces hanging, don't blunder your pawns, etc. It was simply given the rule book and 4 hours to study the game, so it was free from our human theory about chess, and investigated pure approaches.
Stockfish however is the result of our chess culture, it simply computes X moves ahead and compares the resulting positions based on heuristics defined by human theory, like "controlling center squares is better than controlling unoccupied squares" or "a pawn mid-game is worth 198 points and a pawn end-game is worth 258 points".
I love his enthusiasm! It makes me want to see more of this unconventional chess engine at work. It's entirely different way to play.
Awesome commentary, and wtf was that game?! I've never seen anything like that. I've been avoiding watching engine games because I thought they would be dry and boring. This clearly was not and thanks for really taking us along for the ride with the colorful commentary! Definitely going to check out your other engine videos!
This makes me way happier than watching a game between two GM humans. The novelty of play is inspiring. How ironic that it comes from a machine.
Really nice analysis here. Thanks for it, Chess24.
Very instructive and nicely covered by Matthew! THX
Thanks for the video! Looking forward to reading your book on AlphaZero
FANTASTIC analysis, thank you! Very engaging and relevant presentation.
If a robot boxer knocks out Anthony Joshua or Wladimir Klitschko in 30 seconds flat and in 15 in the rematch, you don't need to wonder how Tyson, Frazier or Ali would have fared. The whole Stockfish version is a moot point at this level.
I think this is my new favourite alpha zero game. the way it dismantled the centre with pawn sack after pawn sack to get the bishops looking straight into blacks king is just unreal. totally out of this world
best video of the year.
Like Kasparov play against Deep Blue,let Carlsen play against Alpha Zero to see how he handles a 4,100+ rating artificial intelligence.
4,100?? More like 3000
Yup brilliancy at its finest. Like painting in the canvas.
Wish we had the PGN in the notes. The link at Chess24 wants a membership to download them. I'm sure I can find them somewhere online.
10:25 *Yeah* - Anatoli's cold iron hand getting a hold of the back of your neck, threatening to crush it pleasurably. I like that.
13:50 Exactly the moment Bruce Willis would have said: " *Yippee-kai-yeah, motherfucker!* "
AlphaZero plays chess as it should be played!
Very nice analysis! Well done and thank you.
This is a very pleasant video, and our host is someone I could listen to for hours. Clearly a towering intellect. But - like many analysese by masters - I find the comments very odd. Yes, I'm keenly aware that this is a strong sign that I'm a chess idiot. But, gimme a moment. Our commentator says "So ALphaZero keeps his king nice and safe in the corner". Except - black's king is also "nice and safe in the corner" with more pieces around it, and a square to retreat to if things get awful. Another one - when white occupies the G anf H files with rooks, it's a clever move, but when it abandones those files, it's also a clever move and somehow doesn't make the previous strategy a waste of time. It seems like so many of the comments are not about the intrinsic moves or positions, but are informed largely by hindsight. Gulp. (ducks)
Despite its terrific record I think that AZ way of playing is still within what humans somwhat already theorized. AZ, however, takes it to the extreme. While this game in particular is extraordinary (perhaps the one where AZ showed the most dominance) it is still based on the concepts of space advantage and piece activity. It "just" happens that AZ can confidently prove its point about these "intangibles" over the mere material aspect of the game and do that with the accuracy typical of machines.
By further analyzing this game (and other AZ games as well) one can really appreciate how your own pawns are often more of a liability than an asset and how this assesment only varies based on your tempo advantage/disavantage. Basically, what I get from this kind of play is that the so called "pawn sacrifices" in this game are in fact moves that strictly and consistently improve white position, without a single doubt.
Also another recurrent theme in all AZ games I've seen so far is the use of zugzwang based tactics to prove its point about intangibles being more important. Again, a very easy concept to grasp for humans and yet not used enough due to the lack of calculating abilities (to foresee such a risky position where the simple order of moves counts) but also the lack of confidence to adopt such approach, as if it was only a secondary aspect in the game of chess. Apparently, the order of moves is actually very relevant for AZ to win games. With this I especially mean that he doesn't just dominate but also ends up finding ways to score the point in otherwise favorable but drawable positions.
position development>piece captures. I think most grand masters over value piece count (especially pawns).
Brilliant. Love your enthusiasm. About as good as chess gets.
Excellent analysis, thanks for the video Matthew. Bb1 made me laugh out loud, what a move!
Wonderful, wunderbar, merveilleuse, admirabilis. The languages I love expressing the same concept.
Did he just say 7 pawn sacrifices? That is how u start a chess video
12:10 If alpha zero plays c6, I'm gonna lose it.
13:09 Matthew "I was disappointed with alpha-0, as it took a pawn back. I thought this was materialistic!"
I wish you would slow down in making moves and verbalize the move played. Like "alpha plays knight to B6" and offer some insight as to why. that would add more to your video. Making it longer yes, but worth the effort imo
The way Alpha Zero views our chess grandmasters is perhaps similar to the way alien philosophers view our philosophers... imagine that.
I'm terrified of a future version of AlphaZero that can figure out the best strategies and tactics for any given situation in a combat zone.
I don't understand, black plays 4 moves with his g8N just to exchange it. It aganst the most basic principle of Nimzovich. Is it really theory? Is it posible that black was denied from his opening book?
I think the only rule alphazero does not have is a point score on pieces. It likes mobility and trapping S.fish behind its own pawns as it is willing to give away its own pawns for more mobility. Also a seamless middle and end game it plays.
How remarkable that it is AI that has returned us to the romantic era of chess.
What i don't understand about the Grand Master vs Deep Blue matches was if the computer was also allowed to think about moves during the player's turn. Surely this would have been unfair if it was the case. Does anybody know? Or is 1/2 the time just equal to 1 bit for a computer like Alpha Zero?
Sez it would have been unfair because essentially for real players some element of the oponent's move must essentially be "waiting time" whereas for a computer this would not be the case.
About 60 seconds into this commentary, we're told it's a very flexible opening because "white has the two bishops and a lot of possibilities about how to deploy its pieces". Well maybe, but neither of the bishops can move!
Oh and you can just count Black's offside material to know that the sacrifice is sound. A dozen pawns of value is unemployed for Black.
Please notice Black's initial play of pawn hunting before finishing development. That is Beginners folly!
Bobby Fischer wouldv loved this kind of chess.
I think this may be my favorite game of all time.
The essence of chess talent is flexibility-Alexander Alekhine
Nice commentary!
How can you have pawn sacrifices no. 7 when i still see 3 white pawns on the board?
Matthew calls this “one of my favourite games” and counts 7 pawn sacrifices from AlphaZero in total in “an absolutely fantastic attack”. He quotes DeepMind co-founded Demis Hassabis as describing this game as, “like chess from another planet”.
It is the End of Human Era and beginning of the Machines Intelligence Era
Wow this is a active and attack chess
This game just change me how to think different in chess and come out of the box
So I noticed that black basically went straight for a castle (king side). It's first 4 moves were all in that vein. Does this mean this was definitely a mistake? When you castle that early, isn't it supposed to be easy to attack? Now I know AlphaZero didn't calculate the whole attack, but you don't need to. If you know where the enemy king is going to be the whole game, you know in general what you want to do. I saw the potential attack after black traded the bishop for knight (easy battery with queen and black squared bishop at this point). Perhaps the super early castle was simply a mistake that we can learn from. I wonder how many of AlphaZero's wins where when the opponent early castled?
I don't think so. I don't imagine black castles long in this opening after some of white's early moves are expanding on the queenside. The kingside was pretty safe for black until black decided to win a pawn and open the g file. That could have led to a winning endgame if the white attack didn't work out.
Didn't know Gargamel became a chess analyst
2:45 Qh4 is -0.18 eval and my Stockfish wants instead g6 (-0.68 eval). So this move by the Stockfish they are using is half a pawn blunder compared to a not broken stockfish. Depth 33, I would love to know what depth was used for that move and compare leaving my stockfish to reach that depth and see the eval of both moves. Similar wrong moves compared to my own version was found also in previous game (I posted there too).
15:05 L.E.E.L.A.
I dont understand why d5 was not played earlier. looked like a great move, or am i missing something?
I wonder how long it is going to take before AlphaZero starts to upload its games on youtube and give commentary itself...
set the board up on any move and it'd be interesting to see alphazero play itself?
It's like the best of Karpov, the best of Tal, and the best of Petrosian in one
I would have liked to have seen the end of the match.
I think fide need to add one more category --- M(Master) IM(International Master) GM(Grand Master) GGM(Great Grand Master)
Adventurous and scarily good analysis. Wish GMs learn from AZ
How do I get the PGN for this particular game? I browsing through the database but I can't find it.
I would like to see Magnus carlsen vs Alpha zero without one of the bishops.
This AI conscious collective will make moves that are ones to counter the expected best move. It's like a false blunder. It may look like it is wrong but if the other player plays everything right by theory then that's when alpha's plan will shine. It counts on good play to be made so it can excel in doing these moves that no one will see coming until 25 moves later if it goes according to its predictions. All while calculating different variations in real time making adjustments if the opponent is not following a winning line. It's very complicated to understand but I see that alpha zero is literally just countering everything you do before you even do it.
That's the reti,
It's starts out with Nf3 and then c4,...
I wonder if Alpha Zero just plays with the position, or if it seeks for exploitable flaws in the opponents play. You´ll know how to trick your opponent in balanced or stuck positions if you´ve figured out his pattern of play.
What you're describing is the difference between a player aiming to play optimally and a player aiming to play as a nemesis. These two different play styles are often seen in poker, for example, but they exist in every competitive game. However, nobody aiming to win consistently plays primarily as a nemesis in any game, especially chess. Our opponents will always adapt and become better, it's not favorable to handicap ourselves to assumptions on our opponents' mistakes. We can be confident AlphaZero plays from the first perspective and not the second as it trains against itself. AlphaZero chose to make these sacrifices likely due to the long-term benefits as shown in an analysis by Matthew Sadler here.
AZ has learned chess by being given the rules of the game and then by playing against itself and only itself for millions of times. This means that it never had the opportunity to identify and exploit Stockfish's potential intrinsic weaknesses.
More importantly, Alpha Zero is called zero because it has ZERO knowledge of anything but the rules of the game. It has ZERO knowledge of whom it is playing against, ZERO knowledge wether it is playing against itself, another chess engine or a human. Because of this, AZ will not modifiy its playstyle with different opponents.
Had AZ learned chess by playing only against Stockfish, then most certainly it would have found and used exploitable flaws that are specific to stockfish. However, if presented against another opponent, AZ wouldn't adapt, it would still play trying to exploit flaws inherent to stockfish.
Wow i really enjoyed your commentry. Gd job
Brilliant game although I don't understand why stockfish didn't play Nxe4 to trade off Alpha zero's light squared bishop in the middle game
Well #alphazerosclose to ✌
Nice analysis of A0's moves. Scarily human thinking but with incredible accuracy.
I love these A0 videos with GM Matthew Sadler
It would be better you use mouse and highlight the pic that u r talking about in all conversation.
If Alphazero could speak he would mot totally agree with our explanation of his games.
AlphaZero basically said to play like Morphy. This isn't really groundbreaking other than it proves that Stockfish has a bad horizon effect. Humans will learn from this, because unlike auto racing versus cycling, where you can't put an 850-horseplayer engine into your stomach, we can copy the moves of the best chess computers. AlphaZero has given the green light to the romantic style of chess from the 19th Century.
9:38 - can someone make a gif?
THE NEW ROMANTICISM OF AI CHESS
5/5 stars for the video 👍