Hi Jerry, excellent videos and analyses. There is one game between all those 200 games that is a Semi-Slav, where Stockfish (Black) managed to get a pawn in c4 and g4, and then things get super complicated (atleast for me xD).
I would LOVE to see the thousand-move games, where both computers use continuous draw threat to force moves to be made. ;) Please post one of those videos!
Thanks Jerry. I love your commentary and basically everything you do on these chess games. Are you ever going to participate in the titled arenas that we usually watch Carlson play? I’d really like to see how you do. What is your current bullet or blitz rating?
@@elijahmiles8874 Comp sci = computer science. 65535 (2^16 - 1) is the largest 2-byte (16-bit) integer you could store. In binary, that number is the highest integer you can have with 16 digits.
SpinSpider Just so you know, the “4” in “Stockfish Level 4” refers to a different number than the “8” in “Stockfish 8”. SF8 is the eighth version of Stockfish, which can be configured to play at many different difficulty levels. In this video, SF8 is playing at full strength. When you play SF level 4 on lichess.org (for example), you are playing SF10, configured to be much weaker than its full strength.
It's interesting how Alpha teaches us how we've all been misled by our training. When we, as humans, see that we can get a rook for a bishop, it's almost an automatic move to take it in most cases. Since Alpha taught itself from scratch, it was never miseducated into thinking that any piece has a specific set value. The value of any piece is entirely dependent on what it can actually do on the board, given the position of all the pieces in the game, and has absolutely no inherent value otherwise. I love the games where it's like Stockfish: Ha! It's the mid-game and I'm up three pawns! Suck it, AlphaZero! Alpha: Material means absolutely nothing. Look at the board. You're losing.
Indeed, but if we're trying to reduce inaccuracies in general, it's worth taking note of what AZ seems to show - that thinking of pieces as having set values is itself an inaccurate and unnecessary way to analyse a game. The value of a piece depends on what it's doing now, and what it has some prospect of doing (not in principle, but in the actual game unfolding). To some extent, we've always known this, else such things as sacrifices would never have appeared in human GM games to begin with. AZ is just able to make much better use of this higher-resolution thinking than we can, for the reasons you give (I don't think an non-augmented human is ever going to beat Alpha Zero!).
I always thought the point values are just for beginners to understand exchanges. Even intermediate players understand that pieces value is dependant upon what they are doing. It's engrained into you once you understand the value of connected 7th rank pawns can be more powerful than a queen or the first time you get mated by minor pieces while your up material Looking at high level games, tals games come to mind specifically, they will trade down material if it is a sure advantage. If theres no way to materialize the advantage then a rook is more powerful than a bishop but if there is a strategic reason then bishop may be better. Alpha zero probably just understands deeper than anyone when and how the bishop can be better. Just thinking out loud here though who tf knows right? Certainly be not me
Alpha Zero seems to have this incredible error free, strangling style of play. Whilst not shocking, it’s incredible to see how powerful this approach can be.
*Watches AlphaZero play* Wow, these moves are so obvious. Of course you would play these moves. *Plays against Stockfish 3* Damn, I got checkmated in 4 again...
It's what I like to call the "wheel fallacy". Basically, something logical seems obvious once presented, such as the original man-made wheel, but whose discovery actually takes great insight to realise.
I prefer to think of it more like the Tao of Chess. Alphazero has this because Alphazero doesn't play with intentions in mind. When you have a human player, especially an inexperienced one, they'll play with plenty "more" on their mind. They may be looking to develop a particular strategy (clumsily), or they may be pre-occupied with certain pieces that are plaguing them or serving them well. It's not just that it seems logical in retrospect, it's that Alphazero moves with what you might call grace. Once you take away all the cognitive biases that plague a human player, play seems 'obvious'.
Poisoned pieces, sac the exchange, gambit away a few pawns ... the machines are taking us back to the Romantic Era of chess. All 19th C players would feel right at home here!
The pattern I see from AlphaZero. 1. Put all pieces to ideal position, than only commit to castle. 2. Do not overcommit in the center as it what hold pretty much every position. Plus, there might be chance for the king to stay in the center, keeping it active is playing with an extra piece. 3. As white, most of the time the light square bishop is shut down pretty hard. Restricting the castle at a8 further. 4. Pawn sac and keeps the bishop pair alive so there are less pawn block for greater piece activity. 5. No rush with pawn break. Force stockfisk into zugzwang, then only break open the position. 6. Last but not least, piece maneuver and space acquisition is always on point. Attacking on one side without compromising too much defending on the opposite site. It's pure masterclass. Have to say, every game played is very impressive. Yeah. Last but not least, for anyone who keep saying "The move is clear as daylight." Beat stockfish 8 soundly, with proof 1st and foremost of course, then I will believe you.
Was just watching your blitz tournaments and now im here, love your channel and you do a great job at explaining methods lower ranked players like myself can understand, thanks jerry!
Haven't watched video yet but making prediction: AlphaZero turns Stockfish's light square bishop into a useless piece Also love your vids Jerry :) you inspire me to get better at chess
Your analysis is on another level imo. I really appreciate your videos and hope you have more alphazero games. Break through stuff, amazing to see, and your analysis of the games is awesome.
Thank you, Jerry. I've been watching for years and still love your analyses. Just yesterday I wore my "Back to Tournament" white knight shirt to the gym. It's always good for a few comments. Thanks again.
@@NoahAbrams01 if chess is solved and it is a win for white, either d4 or Nf3 will be the winning opening move. But like you said chess could just be a draw
I have found your analysis of alphazero games to be most useful amongst all the other ones on the internet. You elucidate positional play and strategic ideas brilliantly. I hope you keep them coming, I most certainly will be looking forward to them.
Hello, Jerry! So glad I discovered your channel here ...What I like most here is when Alfazero talks to Stockfish at the end of the game, giving him lessons on how to play. I get mesmerized of that robotic voice rebuking Stockfish...I feel like ET came back to play chess with us, humans.
Oh, yes! AlphaZero returns! I watched the older AlphaZero videos but I love seeing there's a new one! As I've wrote this comment at the beginning of the video, I look forward to what AlphaZero thinks of this match! 23:53 AH, YEAH! Gotta love AlphaZero's messages to Stockfish.
something that's exciting for me about AlphaZero is it's ability to multitask, it can send multiple different threats to all corners of the board while still being connected and without weakening itself
It does seem like the best strategy to translate raw computational power into chess wins is to construct complicated scenarios where you have as many movement options as possible (ideally) while making it hard for the opponent to have an easy strategy of their own. The more time your opponent has to waste to think of a good move, the worse the move they pick under time constraints is going to be. At least with high probability.
It would be nice if you showed evaluations for the newer version of stockfish too, to show how it has changed and maybe some insight on how stockfish has improved.
Jerry at 14:56 I think the reason you want to keep the D rook where it is is so that the e5 advance remains a threat. It's the only explanation I can come up with.
A hint to German terminology: The German "Z" sounds like a "tz" like in "quartz", so "Zugzwang" pronounces like "Tzug-Tzwang". Always makes me happy to see German loanwords in English though :D
@@wirelessbaguette8997 Very well said. You accidentally taught me why so many foreigners pronunciate the "ng" at the end of German and Dutch words as "nk".
12:35 the bishop is heavily restricting black's light-sq bishop. Note how tangled black's pieces are as well; trading would relieve some of that pressure.
15:07 based on what i've learned from your videos, my best guess: he may way to push his e-pawn at some point (candidate pawn for promotion) & if he does, his d1 rook is the only defender of d5... maybe?
How long did Google actually let AlphaZero train itself? And with how much computing power? Thanks in advance! Greetings from Switzerland. ps: Looking forward to many more of those videos! During Carlsen vs Caruana I was actually wondering when they get released. And whether they will let them play again soon.
Why on earth did Stockfish play 12…P-KR3 when it knew it would do nothing to scare the Knight away? I would have thought something like 12…PKN3 would have been better? Thanks Jerry.
15:11 The white pawn is blocking the h-file. The other rook is defending the e pawn in case black's e pawn is forced away by white's d pawn. By moving the h rook to the f file, the bishop is being pinned (indirectly defending the h pawn) and the rook is being put on a better file.
Will you be doing more Leela videos? She is shooting up the ranks in TCEC in this season and will likely make it to premier division, and she has been placing #3 or #4 in CCCC.
@@spy8514 yes he can, once the Queen takes the h5 pawn, the f4 rook can move to f8. The Queen is pinned because the king is behind it. Also the Queen on f2 protects that square. If the queen takes the pawn, essentially the queen is dead. That pawn is called a poison pawn
Is there any site where you can hook up your chess program/robot/AI to play other chessrobots with timelimits? I guess it would need a textfield interface to be feasible and maybe running your robot in another frame sending messages over cross origin frames? If one were to fill in ones own site and the input field name of the "chessiste" was known. It could be rather fun challenge other chess AI's using time constraints of x seconds. The important thing would be that the site was easy to communicate with via input textfields.
So this one is NOT on top 10 games correct? And is funny to explain some combo with not even "best" move from Stockfish!(e.g 13:54 jus BF6 and protect all)
Hi Jerry. I don't play much Chess, but do you think the reason AlphaZero chose the rook on h1 because the pawn in front of it could no longer move diagonal, so it need to "open up" that rook's eyes to further play?
In human chess, a player may make moves to set tactical traps for his human opponent. I assume these AI programs don't do that, and just play the best positional move. Any thoughts?
I don't think Stockfish is programmed to set traps, that would require to know opponent's level and attitude (and a bit of gambling), and well Alphazero is a self learning program which learns chess playing against itself, it's hard to figure how it can learn about traps.
What's the 'thinking time' for AlphaZero between moves? Also, does it think every move through, or does it remember what it 'thought' about in the previous move to cut time when a new move is about to be played?
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.
There is a book coming out in January with a collection of these games and analysis. Daniel King was talking about it on his channel Power Play Chess. Wasn't sure if you would be interested or if you already knew about it. Thanks for the great easy to understand commentary. Really nice for average chess players like myself.
Alphazero seems to not keep pawn close to the king but rather have some air around him? Also don't trade (maybe don't stockfish or GM does either) much and really love open files... right?
I was thinking of starting to play the Queen's Indian as black, should I even bother? The d5 pawn sac followed by Nh4-Nf5 looks devastating. Like holy shit
Here is the AlphaZero vs Stockfish playlist:
th-cam.com/play/PLQsLDm9Rq9bEvlHwbsWlt5sta3M2t0aGe.html
Hi Jerry, excellent videos and analyses. There is one game between all those 200 games that is a Semi-Slav, where Stockfish (Black) managed to get a pawn in c4 and g4, and then things get super complicated (atleast for me xD).
Alpha zero makes the most interesting plays in the history of chess!
I would LOVE to see the thousand-move games, where both computers use continuous draw threat to force moves to be made. ;) Please post one of those videos!
Thanks Jerry. I love your commentary and basically everything you do on these chess games. Are you ever going to participate in the titled arenas that we usually watch Carlson play? I’d really like to see how you do. What is your current bullet or blitz rating?
ChessNetwork 5
"I've only skimmed through over 100 of them."
Jerry, December 2018.
I believe the "only" refers to the fact that he skimmed them, and didn't analyze deeply ;)
He's slacking off a bit
Such laziness. Kappa :)
AlphaZero is busy writing "My 65535 Most Memorable Games"
Ah you also studied comp sci?
Dan Kelly What does it mean?
@@elijahmiles8874 Comp sci = computer science. 65535 (2^16 - 1) is the largest 2-byte (16-bit) integer you could store. In binary, that number is the highest integer you can have with 16 digits.
@@FiresBZ More like "My 2147483647 Most Memorable Games"
18446744073709551615
:P
Me: wow stockfish 8 sucks
Also me: stockfish 4 mated me
SpinSpider Just so you know, the “4” in “Stockfish Level 4” refers to a different number than the “8” in “Stockfish 8”. SF8 is the eighth version of Stockfish, which can be configured to play at many different difficulty levels. In this video, SF8 is playing at full strength. When you play SF level 4 on lichess.org (for example), you are playing SF10, configured to be much weaker than its full strength.
Lool
Hahaha
Lmao
underrated comment
It's interesting how Alpha teaches us how we've all been misled by our training. When we, as humans, see that we can get a rook for a bishop, it's almost an automatic move to take it in most cases. Since Alpha taught itself from scratch, it was never miseducated into thinking that any piece has a specific set value. The value of any piece is entirely dependent on what it can actually do on the board, given the position of all the pieces in the game, and has absolutely no inherent value otherwise.
I love the games where it's like
Stockfish: Ha! It's the mid-game and I'm up three pawns! Suck it, AlphaZero!
Alpha: Material means absolutely nothing. Look at the board. You're losing.
Indeed, but if we're trying to reduce inaccuracies in general, it's worth taking note of what AZ seems to show - that thinking of pieces as having set values is itself an inaccurate and unnecessary way to analyse a game. The value of a piece depends on what it's doing now, and what it has some prospect of doing (not in principle, but in the actual game unfolding). To some extent, we've always known this, else such things as sacrifices would never have appeared in human GM games to begin with. AZ is just able to make much better use of this higher-resolution thinking than we can, for the reasons you give (I don't think an non-augmented human is ever going to beat Alpha Zero!).
Humans thought they had mastered chess but engines and AI proved otherwise.
Except at 12:59 Stockfish 8 recognizes gxh4 (the move played by AlphaZero) as the best move instead of bishop takes rook
@@MrSupernova111 Was it against a chess engine or AI?
I always thought the point values are just for beginners to understand exchanges.
Even intermediate players understand that pieces value is dependant upon what they are doing. It's engrained into you once you understand the value of connected 7th rank pawns can be more powerful than a queen or the first time you get mated by minor pieces while your up material
Looking at high level games, tals games come to mind specifically, they will trade down material if it is a sure advantage.
If theres no way to materialize the advantage then a rook is more powerful than a bishop but if there is a strategic reason then bishop may be better.
Alpha zero probably just understands deeper than anyone when and how the bishop can be better.
Just thinking out loud here though who tf knows right? Certainly be not me
In 2018 we have Humans playing like Engines (Caruana v Carlsen) and Engines playing like Humans (Alpha Zero v Stockfish)
What a time to be alive
In 2020 the boundary between human and engine will be no longer exist
Alpha zero does not play like a human.
Alpha Zero seems to have this incredible error free, strangling style of play. Whilst not shocking, it’s incredible to see how powerful this approach can be.
@@nofanfelani6924 people have been saying that every year for 20 years now
But what is "alive"?
*Watches AlphaZero play*
Wow, these moves are so obvious. Of course you would play these moves.
*Plays against Stockfish 3*
Damn, I got checkmated in 4 again...
Story of my life lol
It's what I like to call the "wheel fallacy". Basically, something logical seems obvious once presented, such as the original man-made wheel, but whose discovery actually takes great insight to realise.
I prefer to think of it more like the Tao of Chess. Alphazero has this because Alphazero doesn't play with intentions in mind. When you have a human player, especially an inexperienced one, they'll play with plenty "more" on their mind. They may be looking to develop a particular strategy (clumsily), or they may be pre-occupied with certain pieces that are plaguing them or serving them well.
It's not just that it seems logical in retrospect, it's that Alphazero moves with what you might call grace. Once you take away all the cognitive biases that plague a human player, play seems 'obvious'.
Nah I take back moves and drag along till move 12 or 13 and then the situation is irredeemable so I close the window
@@kaniza1465 Ah, didn't know it had a term
Poisoned pieces, sac the exchange, gambit away a few pawns ... the machines are taking us back to the Romantic Era of chess. All 19th C players would feel right at home here!
I would, I mean they would.
🎯
Human chess is a joke compared to this. Upload the game where Alpha Zero gives up 4 pawns to dominate with the bishop pair.
yeah that game was awesome
Anna Rudolf uploaded it on her chess channel yesterday
Yea , this one is too much 😪
@@perchix5252 link?
Machines beating up machines. Can I tell my family I'm watching Transformers? Maybe they wouldn't give me more weird looks.
Coming next video:
AlphaZero goes house hunting with Stockholm.
i don't get it :( but seems funny
The pattern I see from AlphaZero.
1. Put all pieces to ideal position, than only commit to castle.
2. Do not overcommit in the center as it what hold pretty much every position. Plus, there might be chance for the king to stay in the center, keeping it active is playing with an extra piece.
3. As white, most of the time the light square bishop is shut down pretty hard. Restricting the castle at a8 further.
4. Pawn sac and keeps the bishop pair alive so there are less pawn block for greater piece activity.
5. No rush with pawn break. Force stockfisk into zugzwang, then only break open the position.
6. Last but not least, piece maneuver and space acquisition is always on point. Attacking on one side without compromising too much defending on the opposite site. It's pure masterclass.
Have to say, every game played is very impressive.
Yeah. Last but not least, for anyone who keep saying "The move is clear as daylight." Beat stockfish 8 soundly, with proof 1st and foremost of course, then I will believe you.
ok
Was just watching your blitz tournaments and now im here, love your channel and you do a great job at explaining methods lower ranked players like myself can understand, thanks jerry!
Titles are getting better and better.
Haha indeed, hood sir.
Edit: good sir.
Haven't watched video yet but making prediction: AlphaZero turns Stockfish's light square bishop into a useless piece
Also love your vids Jerry :) you inspire me to get better at chess
Ur prediction was 100% correct;)
@@scorchedBoard well, is that even a prediction?
Hi Jerry, it's everyone.
More plz! Best commentary available by far. You don’t dumb it down.
At 9:01 I auto assumed that "this is played with" would be followed by "a draw offer" because I've gotten so used to hearing that now.
Drawsen - Drawuana LUL
@@vgamerul4617 more like drawish giri
Your analysis is on another level imo.
I really appreciate your videos and hope you have more alphazero games. Break through stuff, amazing to see, and your analysis of the games is awesome.
Thank you, Jerry. I've been watching for years and still love your analyses. Just yesterday I wore my "Back to Tournament" white knight shirt to the gym. It's always good for a few comments. Thanks again.
Fantastic, unhurried analysis.
Jerry, your commentary is so unique and fun, very instructive and covers so many variations... post as often as you'd like and you can, they're great.
The day when after 1. d4 black resigns the game is coming sooner than we thought
maybe 10 years?? 20? 960 will become the default by then tho
That's called the French defense.
XD
@@NoahAbrams01 if chess is solved and it is a win for white, either d4 or Nf3 will be the winning opening move. But like you said chess could just be a draw
Are you sure black is not the winner? First move is always a mistake and black wins is my brave guess :) So before the game starts white resigns ;)
@@Gasimoe lol maybe at 2832-2835 level
I have found your analysis of alphazero games to be most useful amongst all the other ones on the internet. You elucidate positional play and strategic ideas brilliantly. I hope you keep them coming, I most certainly will be looking forward to them.
Hello, Jerry! So glad I discovered your channel here ...What I like most here is when Alfazero talks to Stockfish at the end of the game, giving him lessons on how to play. I get mesmerized of that robotic voice rebuking Stockfish...I feel like ET came back to play chess with us, humans.
i love love love alphazero on your channel, keep up the great work man !!
Oh, yes! AlphaZero returns! I watched the older AlphaZero videos but I love seeing there's a new one! As I've wrote this comment at the beginning of the video, I look forward to what AlphaZero thinks of this match!
23:53 AH, YEAH! Gotta love AlphaZero's messages to Stockfish.
Omg been waiting for months, thank you a lot for continuing this playlist
210 more games!? I can't wait to watch your reviews!
i wish i lived in the parallel reality where jerry analyzed and commented and all of these :(
something that's exciting for me about AlphaZero is it's ability to multitask, it can send multiple different threats to all corners of the board while still being connected and without weakening itself
It does seem like the best strategy to translate raw computational power into chess wins is to construct complicated scenarios where you have as many movement options as possible (ideally) while making it hard for the opponent to have an easy strategy of their own. The more time your opponent has to waste to think of a good move, the worse the move they pick under time constraints is going to be. At least with high probability.
Thanks Jery, I really appreciate your time and passion for the game. Cheers, Alireza :)
The AlphaZero talking trash is priceless!
It would be nice if you showed evaluations for the newer version of stockfish too, to show how it has changed and maybe some insight on how stockfish has improved.
Jerry at 14:56 I think the reason you want to keep the D rook where it is is so that the e5 advance remains a threat. It's the only explanation I can come up with.
that terminator thing at the end freaked me out LOL
There goes AlphaZero annihilating the poor fish again.
The fish is shit easily defeated jajajaj
+@@christianquinones9347 Play against it yourself and say that again...
Zugzwang! Yay!
PS: thank you so much for covering those games. Love your style, always instructive, especially for me as low-level-layman :)
Thank you Jerry, I am looking forward to over 200 alpha0 games commented by you. 🤠👍
Greetings from Germany, Black Forest. 🌲
Jerry, I know I speak for a lot of people where if you were to commentate every game that was released, I'd watch! Your analysis is great man!
A hint to German terminology: The German "Z" sounds like a "tz" like in "quartz", so "Zugzwang" pronounces like "Tzug-Tzwang". Always makes me happy to see German loanwords in English though :D
T33K3SS3LCH3N a "g" proceeded by a vowel in German is also an English "k" sound, so it would actually be tsook-tsvawng.
@@wirelessbaguette8997 Very well said. You accidentally taught me why so many foreigners pronunciate the "ng" at the end of German and Dutch words as "nk".
Tsuktsuvang is the correct porrunciation 😂
Tsuk tsvang is the most correct one 😂
Everyone's jumping on these... Anna Rudolph has some great analysis too. Material is not as important as activity!
12:35 the bishop is heavily restricting black's light-sq bishop. Note how tangled black's pieces are as well; trading would relieve some of that pressure.
Another fun and educational video from Jerry. Thank you!
"Stockfish resigns" 👀
I'm glad they released more games cause it means more videos from Jerry!
Great series. Appreciate the depth of analysis.
Best chess commentary on this channel. I watched all your videos
It was wonderfull how alpha zero didn’t had any rush in winning the game to rather ending it in a cleaner way like if it cared about it
Hey Jerry, loving the content. Could you please cover the Tal vs Botvinnik match in 1960? Thanks in advance.
14:55 Rdf1 vs Rhf1? If Rdf1 g3 later threatening g2 with fork?
Please make more videos of AlphaZero, I love to watch them.
I think AlphaZero lost a handful of these games right?
Will you review some of those, and perhaps explain what you think went wrong?
From my observations from the released games, A0 lost mainly due to endgame blunders.
Alphazero didn't lose a single game
@@St3lf There are a couple lost games in the big new dump DeepMind's released.
View them here. deepmind.com/research/alphago/alphazero-resources/
@@St3lf AlphaZero lost 6 games, drew 839 games, and won 155. This gives AlphaZero a +52 Advantage over Stockfish 8.
@@Cscuile And how does that translate to elo? How much higher is Alpha's estimated elo?
What a great analysis!! Beautiful!!! Thank you very much.
15:07 based on what i've learned from your videos, my best guess: he may way to push his e-pawn at some point (candidate pawn for promotion) & if he does, his d1 rook is the only defender of d5... maybe?
How long did Google actually let AlphaZero train itself? And with how much computing power?
Thanks in advance!
Greetings from Switzerland.
ps: Looking forward to many more of those videos! During Carlsen vs Caruana I was actually wondering when they get released. And whether they will let them play again soon.
19:19 is Qxh5 too aggressive? Opening lines up to the king? Theres no longer an immediate threat of a pin by the rook.
Why on earth did Stockfish play 12…P-KR3 when it knew it would do nothing to scare the Knight away? I would have thought something like 12…PKN3 would have been better?
Thanks Jerry.
It's crazy how even the material was throughout, but how crushing it was.
15:11 The white pawn is blocking the h-file. The other rook is defending the e pawn in case black's e pawn is forced away by white's d pawn. By moving the h rook to the f file, the bishop is being pinned (indirectly defending the h pawn) and the rook is being put on a better file.
I really enjoyed your video here! Thanks for making it!
Will you be doing more Leela videos? She is shooting up the ranks in TCEC in this season and will likely make it to premier division, and she has been placing #3 or #4 in CCCC.
They played Rhf1 because it keeps the d5 pawn supported, which may be helpful after e5.
Move 25 if you move the D1 rook there could be a Bishop sacrifice on D5 with a now very active black queen
Loving your content Jerry, keep up the good work!
On move 37 @18:31, why didn't Stockfish took the pawn on h5?
Maybe because the black king is hiding behind that pawn
Cos the rook can swing in to fork the queen
@@hanijan480 No it can't. But maybe Aleksandr is right. The pawn helps in blocking for the king.
@@spy8514 yes he can, once the Queen takes the h5 pawn, the f4 rook can move to f8. The Queen is pinned because the king is behind it. Also the Queen on f2 protects that square. If the queen takes the pawn, essentially the queen is dead. That pawn is called a poison pawn
@@spy8514 maybe ur not right
Great explaining man .....those lines helps a lot
At 7:00, why not bxd5? Seems like a clear and free pawn.
You'd lose that bishop for the pawn. Knight can't take the white bishop because it's defending h7 from checkmate.
@@DeanJarratt Ah, yes. Whoops.
I would take the 'tale of the tape' with a pinch of salt as it only analyses each move for a fraction of a second.
Great to see this series starting up again :D
LEELA VS ALPHAZERO?
Or Az on CCC?
@@haromt3 The only way to truly know AZ's strength is if Deepmind Submited AZ to CCCC or TCEC. I would love to see AZ in CCCC/TCEC one day.
Excellent interpretation. Pretty deep analise.
Is there any site where you can hook up your chess program/robot/AI to play other chessrobots with timelimits? I guess it would need a textfield interface to be feasible and maybe running your robot in another frame sending messages over cross origin frames? If one were to fill in ones own site and the input field name of the "chessiste" was known. It could be rather fun challenge other chess AI's using time constraints of x seconds.
The important thing would be that the site was easy to communicate with via input textfields.
love ur channel jerris keep up the good work
19:07 After rook f3 why not Qxh5?
So this one is NOT on top 10 games correct? And is funny to explain some combo with not even "best" move from Stockfish!(e.g 13:54 jus BF6 and protect all)
I was looking for this game around TH-cam. Thanks
Neural networks is an interesting technology to follow. Leela being one of them.
Hi Jerry. I don't play much Chess, but do you think the reason AlphaZero chose the rook on h1 because the pawn in front of it could no longer move diagonal, so it need to "open up" that rook's eyes to further play?
In human chess, a player may make moves to set tactical traps for his human opponent. I assume these AI programs don't do that, and just play the best positional move. Any thoughts?
I don't think Stockfish is programmed to set traps, that would require to know opponent's level and attitude (and a bit of gambling), and well Alphazero is a self learning program which learns chess playing against itself, it's hard to figure how it can learn about traps.
Beautiful games, thanks for sharing.
*Your beginners to chess master videos are great* ....*I have learnt a lot from them*...*thank you very much*....
👍🏼
the major piece-endgame type position we reach at around move 35 reminds me very much of fischer vs spassky game 6 1972
What's the 'thinking time' for AlphaZero between moves?
Also, does it think every move through, or does it remember what it 'thought' about in the previous move to cut time when a new move is about to be played?
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.
What is disturbing is Alphazero is teaching itself . A.I. will rule the world, not just chess
Super video and narration. Easy sub. Thank you.
There is a book coming out in January with a collection of these games and analysis. Daniel King was talking about it on his channel Power Play Chess. Wasn't sure if you would be interested or if you already knew about it. Thanks for the great easy to understand commentary. Really nice for average chess players like myself.
Seems the decision not to play rook on d was to maintain protection for the d pawn that you spoke of running up the board.
Knight or pawn takes d5 at 5:24? That's easy, it looks like you should take with the pawn, so this is obviously not correct! :-)
I live for the Alpha Zero voice scolding stockfish at the end.
I watch these completely for the final voice at the end of AlphaZero.
I learn the most from watching your videos than other yt chess channels. thanks you.
You must have been very excited to watch it go fishing xd
yeah haha i remember the early bulletchess videos - jerry the fishing pole king =D
Alphazero seems to not keep pawn close to the king but rather have some air around him? Also don't trade (maybe don't stockfish or GM does either) much and really love open files... right?
Yes! Thank you for showing us!
11:28 , what does he say here? Beautiful court grip?
Core grip I think.
Looking forward to seeing more of these games!
If you listen real hard you can hear AlphaZero saying "OK Boomer"
I was thinking of starting to play the Queen's Indian as black, should I even bother? The d5 pawn sac followed by Nh4-Nf5 looks devastating. Like holy shit
AlphaZero sees the forest. We are looking only at trees.