[Dr.AD]

Originally Posted by Sutripta

^^^
I think there was an AlphaZero - Stockfish game where it kept on sacrificing pawn after pawn, not for a well calculated combination, but for opening up the game. And won.

There are two sets of AlphaZero - Stockfish games. One 100 game set in 2017, and another 1000 game set in 2019.

https://www.chess.com/news/view/goog...100-game-match

https://www.chess.com/news/view/upda...000-game-match

I have not studied these games, so I am not sure. 100 and 1000 are just way too many games for me to study :-)

However, it is virtually impossible that any modern computer (even the relatively lower rated ones) do not calculate a line. So "not for a well calculated combination" is just not possible, and more with AlphaZero. No engine will sacrifice even one pawn without seeing concrete benefits in return. They are after all driven by some kind of objective function (whether traditional engine or an AI engine like AlphaZero), and they can't speculatively sacrifice even one pawn without seeing the benefits in the objective function.

If you could please point out the game where you think it kept on sacrificing pawns, we could discuss the specifics.

Originally Posted by SmartCat

Some random chess questions:

1) Can we beat a Grandmaster if they play against ordinary folks like us with their Queen missing at the start?
2) Will you sacrifice a Queen for two rooks?
3) What do your prefer generally -> defend & hope that opponent makes a mistake, or attack straight away?
4) Will the above strategy change if the opponent is stronger?
5) At later stages of the game, do you prefer a knight or a bishop? Would you exchange one for another?
6) Typically, how many moves do you think ahead - especially in mid and later stages of the game?

My random thoughts (again, others could have different opinions and we can debate) on these questions:

1. The answer will be purely speculative since I do not know of any queen odds games played by grandmasters. It depends on what we mean by the "ordinary folks". If we mean by someone who is a beginner (say 1300 rating), then most probably, yes, the GM will still beat this player without a queen. Although no queen odds, lot of GMs have played rook odds games (i.e. without a rook) with such ordinary people and defeated them almost every time. Now if we talk about someone rated above 1800, it might get tough for the GM without a queen, but still they might win anyways. Further, beyond 2000, it may not be possible for the GM to win without a queen. Also, it depends on whether queen is taken out right in the opening position and at some random position later. So lot of open questions and this really is way too speculative to say anything.

2. In most positions, yes! Two rooks are usually stronger than a queen, and in most positions, one would take two rooks for a queen. In fact, this is no way a "sacrifice". This is a more than welcome trade! However, there could be some exceptional cases where the queen is stronger than two rooks (say for example, if one of the rooks is stuck behind its own pawn). So again, the answer depends on the position, but 90% of the times, two rooks are stronger than a queen!

3. This is never a per-determined decision. You play the opening lines, and depending on how the opponent responds to your opening, you decide on attack or defense. For example, as Black, I play Sicilian Dragon, which is an extremely attacking line, and if White plays passively, then I get a good attack as Black. However, White in return can play a very attacking response to Dragon, known as Yugoslav attack. If White plays the Yugoslav attack, then Black has to be very careful and do the fine balance between defense and attack. Here in Yugoslav, if Black is careless, he can get crushed easily. On the other hand, if I play French Defense as Black (I used to play that a lot earlier, not now), then I must defend well and establish the right position before thinking of the attack. So it is all part of how the game develops and how the players respond to the openings.

4. Yes, definitely. The whole opening choice depends on whether your opponent is stronger or weaker, or whether he/she likes to play positional games or open games etc, and accordingly, the whole game changes.

5. This question has no one answer. This is one of the most well studied questions and there are chapters written on this. The answer is very complex and totally depends on the positions. And even in a given position, grandmasters would differ on what they prefer. In many openings, there are completely different theories when you decide to exchange the Bishop for Knight or not. To give an analogy, it is like asking in cricket, would you prefer a spinner or a fast bowler. It depends on so many things, and even in a given cricket match, two captains may differ in what they prefer. It is the same thing here.

6. It depends on the position. In a position with lot of forced continuations, a good player can see 10 moves ahead. A grandmaster can see 20 moves ahead. But when the position is closed with no forced line, this reduces. In certain endgames, you can maybe see 20 moves ahead. I am nowhere close to being a grandmaster, but I can see 10 moves ahead in some positions, and sometimes fail to calculate well 3 moves ahead in some very complex positions. In fact, this question is the most asked question to all chess players, and nobody can give any precise answer. Every chess player gives the same answer that "it depends".

For example, in the Problem 3 I shared above, it is actually a mate in 6, and I can calculate that 6 move combination easily in less than a couple of minutes. But there are so many other more complex positions where I may need 1 hour to calculate even a 3 move deep analysis.

[Sutripta]

Originally Posted by Dr.AD

There are two sets of AlphaZero - Stockfish games. One 100 game set in 2017, and another 1000 game set in 2019.

https://www.chess.com/news/view/goog...100-game-match

https://www.chess.com/news/view/upda...000-game-match

I have not studied these games, so I am not sure. 100 and 1000 are just way too many games for me to study :-)

However, it is virtually impossible that any modern computer (even the relatively lower rated ones) do not calculate a line. So "not for a well calculated combination" is just not possible, and more with AlphaZero. No engine will sacrifice even one pawn without seeing concrete benefits in return. They are after all driven by some kind of objective function (whether traditional engine or an AI engine like AlphaZero), and they can't speculatively sacrifice even one pawn without seeing the benefits in the objective function.

If you could please point out the game where you think it kept on sacrificing pawns, we could discuss the specifics.

I thought it would be easy to find, but googling 'alphazero stockfish pawn sacrifice' gives more than 8000 results. Looks like pawn sacrifices are the order of the day. If the lockdown continues, will try to find it.

Conventional wisdom says that any algorithm will have a objective function. And will look ahead as many plys as possible given limitation of time. To increase depth it will do some sort of branch and bound. AlphaZero seems to have something extra.

Regards
Sutripta

[SmartCat]

Originally Posted by Dr.AD

3. This is never a per-determined decision. You play the opening lines, and depending on how the opponent responds to your opening, you decide on attack or defense. For example, as Black, I play Sicilian Dragon, which is an extremely attacking line, and if White plays passively, then I get a good attack as Black. However, White in return can play a very attacking response to Dragon, known as Yugoslav attack. If White plays the Yugoslav attack, then Black has to be very careful and do the fine balance between defense and attack. Here in Yugoslav, if Black is careless, he can get crushed easily. On the other hand, if I play French Defense as Black (I used to play that a lot earlier, not now), then I must defend well and establish the right position before thinking of the attack. So it is all part of how the game develops and how the players respond to the openings.

How important is it for a player to know the various chess openings? I have been playing chess for 30 years now, and I'm unable to understand why chess openings are important. When there are literally a million combinations of chess piece movements, why do players at higher levels follow the chess opening routines?

[Dr.AD]

Originally Posted by blackwasp

Bh6+ Nxh6
Qf6+ Kh7 (... Kg8, Qg5+ Qg6, e8=Q+ ... and so on)
Qxh6+ Kg8
Rg6+ Qxg6
Qxg6+ and white can now promote his e8 pawn

Correct!!

In fact, when you said white can promote, it is actually a checkmate. e8=Q#. Thus, this is in fact a forced mate in 6 problem. The complete solution is:

1. Bh6+ Nxh6
2. Qf6+ Kh7
3. Qxh6+ Kg8
4. Rg6+ Qxg6
5. Qxg6+ Kh8
6. e8=Q#

Originally Posted by Sutripta

Conventional wisdom says that any algorithm will have a objective function. And will look ahead as many plys as possible given limitation of time. To increase depth it will do some sort of branch and bound. AlphaZero seems to have something extra.

AlphaZero definitely has something extra, and that is that it is based on Reinforcement Learning (which is a technique in AI) and that it has learnt on millions of games and has calculated its own objective functions without the programmers having explicitly defined the function. Thus, its thinking process is lot more human-like than computer-like. It plays moves that would mimic human moves (learnt looking at human moves in millions of games it is trained on). However, it has also learnt the results of those games, and thus also knows which moves are the best. So in summary, it would make human-like moves which would also lead to a win.

Look at this article for some reference: https://deepmind.com/blog/article/al...s-shogi-and-go

Originally Posted by SmartCat

How important is it for a player to know the various chess openings? I have been playing chess for 30 years now, and I'm unable to understand why chess openings are important.

Short answer: very important. One need not know the opening theory in detail (grandmaster do need that level of knowledge) but at least one needs to know the opening ideas and themes. Else one is playing blind, and will always come out in a bad position out of opening, and then the middlegame is all about saving that bad position.

When there are literally a million combinations of chess piece movements, why do players at higher levels follow the chess opening routines?

The answer is that out of these millions of moves, we already know which ones are the best in which kind of opening, that coherence of a certain set of moves is essentially that opening. One can decide not to follow it, but then one is asking for trouble because you are choosing sub-optimal moves. If your opponent knows that opening well and if you are deviating, you will get punished. When one studies openings, one also studies how to punish any wrongful deviation.

Another practical reason why players follow these opening routines: To be efficient and save time and energy calculating these on the board. Since you know the starting position of the pieces, you can study lot of initial ideas and prepare your lines and play them out on the board. And then start calculating fresh from the middlegame. This is a far more efficient approach. In fact, just like openings, even end games are well studied and often at grandmaster level endgames are played in technical manner, based on previous preparation.

Of course, like everything else in chess, this also has exceptions. There are some grandmasters who purposely play unorthodox openings to confuse the opponent, and to "bring them out of the book". Of the top of my head, the name of Ivanchuk come up as this kind of player. But even this requires preparation and lot of study. These players actually study and deeply analyze their unorthodox lines before they play it on the board. So even here, you can't just play through the openings without knowing what you are doing.

[Sutripta]

Originally Posted by Dr.AD

There are some grandmasters who purposely play unorthodox openings to confuse the opponent, and to "bring them out of the book". Of the top of my head, the name of Ivanchuk come up as this kind of player.

Would Fischers D4 (against Spassky) be regarded as one?

Regards
Sutripta

[Dr.AD]

Originally Posted by SmartCat

When there are literally a million combinations of chess piece movements, why do players at higher levels follow the chess opening routines?

In addition to the answer I gave earlier, maybe I could give a few analogies. Think of car racing. Why do the race car drivers take certain "racing lines" through the corner when they could take any random line? Or think of cricket. Why do we put the fielders in one of those specific fielding positions when we could put them anywhere on the vast open filed? Why do we start we a few slip fielders when the game starts with our fast bowlers bowling?

The reason to all these questions is that certain strategies are already known from the historical knowledge of the game (i.e. we already know batsmen are vulnerable to edging the ball to the slip), and we use that knowledge instead of learning it from scratch in each game. The same is the case with chess opening.

Originally Posted by Sutripta

Would Fischers D4 (against Spassky) be regarded as one?

No, that is still not unorthodox. 1.e4 or 1.d4 are all extremely well studied openings, and these are very much orthodox moves. That 1.d4 was more of a surprise element to Spassky who expected Fischer to play 1.e4. But that does not mean Spassky would not know the theory of 1.d4. Sometimes we play unexpected openings, to surprise our opponent, but that unexpected opening could still very well be an orthodox opening, like 1.d4. In fact, 1.e4, 1.d4, 1.c4 are all well known choices. And as Black, 1. ...e5, 1. ... d5, 1. ...c5 are all valid and well known choices to various White's first moves, as applicable. None of these are what I called unorthodox!

An example of real crazy unorthodox opening move is what GM Tony Miles played against GM Anatoly Karpov in 1980. That to date remains one of the most shocking and unorthodox opening at top level.

To Karpov's usual 1.e4, Miles played 1. ... a6?! Unbelievable, shocking and unorthodox to the core! Karpov, who was no doubt a better player than Miles, playing White, lost this game!!

Here is the famous game: https://www.chessgames.com/perl/chessgame?gid=1068157

[Sutripta]

Originally Posted by Dr.AD

No, that is still not unorthodox. 1.e4 or 1.d4 are all extremely well studied openings, and these are very much orthodox moves. That 1.d4 was more of a surprise element to Spassky who expected Fischer to play 1.e4. But that does not mean Spassky would not know the theory of 1.d4. Sometimes we play unexpected openings, to surprise our opponent, but that unexpected opening could still very well be an orthodox opening, like 1.d4. In fact, 1.e4, 1.d4, 1.c4 are all well known choices. And as Black, 1. ...e5, 1. ... d5, 1. ...c5 are all valid and well known choices to various White's first moves, as applicable. None of these are what I called unorthodox!

An example of real crazy unorthodox opening move is what GM Tony Miles played against GM Anatoly Karpov in 1980. That to date remains one of the most shocking and unorthodox opening at top level.

To Karpov's usual 1.e4, Miles played 1. ... a6?! Unbelievable, shocking and unorthodox to the core! Karpov, who was no doubt a better player than Miles, playing White, lost this game!!

Here is the famous game: https://www.chessgames.com/perl/chessgame?gid=1068157

Of course D4 is perhaps the most common opening for White. However Fischer had till that time exclusively played E4.

AlphaZero seems to prefer C4 (English opening).

Regards
Sutripta

[McLaren Rulez]

Problem 4: White to play and win

I really encourage you not to Google this or use an engine. This is one of the most beautiful puzzles I've ever seen and has more twists and turns than a Bollywood movie. Enjoy!

[Dr.AD]

Originally Posted by McLaren Rulez

I really encourage you not to Google this or use an engine. This is one of the most beautiful puzzles I've ever seen and has more twists and turns than a Bollywood movie. Enjoy!

This is a well-known position, and very beautiful for sure!

1.c7 Rd6+
2. Kb5 Rd5+

(2. Kb5 is required, else if Kb7, the Rd7 pins the pawn and wins the pawn with a draw)

3. Kb4 Rd4+
4. Kb3 Rd3+
5. Kc2 Rd4!

This is the main move. The idea is that if 6. c8=Q??, then Rc4+!!, and after Qxc4, it is stalemate! Hence the pawn can't be queened now!

However, White has another idea than still wins!

6. c8=R!!

Now there is no stalemate, and White checkmates black with the rook. This can't be avoided now. The immediate threat is Ra8#.

6. ... Ra4
7. Kb3 and White wins!

[SmartCat]

Originally Posted by Dr.AD

Or think of cricket. Why do we put the fielders in one of those specific fielding positions when we could put them anywhere on the vast open filed?

OK, this cricket analogy is very good!

I have come across Queen's Gambit quite often. White's d4 is followed by my d5 and then the perplexing White's c4 move. What exactly is the goal behind Queen's Gambit?



Since I'm always looking for that one pawn advantage (I can then rush into endgame with quick exchanges), I always "accept" the Queen's gambit. And then I spend all my resources on defending that one pawn advantage. Now this works well against equal rated peers, but not so much against higher rated players.

Is there a solid plan to keep the advantage after accepting the Queen's Gambit?

[Dr.AD]

Originally Posted by SmartCat

I have come across Queen's Gambit quite often. White's d4 is followed by my d5 and then the perplexing White's c4 move. What exactly is the goal behind Queen's Gambit?

The idea here is that White trades his "wing" pawn (side pawn) for Black's center pawn. As a result, White has an extra pawn in the center, and can quickly establish center control and gain activity in the center. For example white is free to play 3.e4 directly and get nice pawns on e4 and d4, thanks to Black's d pawn (which would otherwise make e4 impossible) being sidelined, quite literally :-)

Not just in Queen's Gambit, but it many openings it is a common ideas to trade wing pawns for center pawns and then get a strong middlegame in the center. It is a very sound plan, both theoretically as well as practically.

Since I'm always looking for that one pawn advantage (I can then rush into endgame with quick exchanges), I always "accept" the Queen's gambit. And then I spend all my resources on defending that one pawn advantage. Now this works well against equal rated peers, but not so much against higher rated players.

Here is the trick. White's "pawn sacrifice" (which is really not) is a temporary one, and Black can't hold on to that extra pawn.

After 1.d4 d5 2. c4 dxc4, White can play 3.e3 or directly 3.e4, establishing a strong center, getting perfect squares there, and also attacking Black's pawn on c4 with his Bishop. If Black tries to save this pawn with moves such as b5 etc,, then Black weakens his position so much that White will get more material. There are many lines with Qf3 etc for White with threats on the h1-a8 diagonal. As an example, let us look at this line (which will never be played by good players, but used just to show the idea of not being able to hold the extra pawn): 1.d4 d5 2. c4 dxc4 3.e3 b5? (truing to support the c4 pawn) 4. a4 c6 5. axb5 cxb5 6. Qf3 and this wins a piece or rook!

So in summary, Black can't hold this extra pawn.

Thus, you can't really cruise into the endgame with this extra pawn. Definitely not if White is a decent player who knows what he is doing and why he chose this opening.

Is there a solid plan to keep the advantage after accepting the Queen's Gambit?

No, not with correct play by White. If that plan existed, then this opening would have been bad for White and White would not have played it all. This is the beauty of these opening evolved over 200 years! There is no simple plan for any side to just hold on to advantage. There are pros and cons to both sides and if they know their plan, they can get advantage. The good openings are fair to both players!

There were/are many openings where one side can definitely gain advantage with correct play. These are "faulty" openings. Either these are phased out already, or still played by stronger player just to exploit weaker players who do not have opening knowledge. But certainly at professional level, such openings have no place.

[McLaren Rulez]

White to play and find a unique mate in 2

[tortoiseNhare]

Originally Posted by McLaren Rulez

White to play and find a unique mate in 2[/center]

The formation is symmetric if you don't count the a-file. The only possible unique move sequence for white is:
1. na7 ... , 2. nc8 ... that forces mate with 3. nd6

I found the puzzle in the youtube link below very interesting

https://youtu.be/PH_Fw3Jiq-8

[Dr.AD]

Originally Posted by McLaren Rulez

White to play and find a unique mate in 2

This is another well-known problem back from 1914! This is named "Christmas Tree", by T. R. Dawson, first published in Falkirk Herald, 1914.

BTW, you do NOT have to say unique mate in 2. Every correct Mate in 2 problem, by rules of the problem composition, must have a unique answer. Else the problem is said to be "cooked" (a term for faulty problem with more than 1 solutions).

Here, at surface, it sounds there are are two solutions. But deeper analysis reveals that only 1 is valid!

The "deeper analysis" here means you have to go back in the move sequence and figure out the previous move, to know which current move is valid. In Chess, this kind of analysis is called "Retrograde Analysis".

Solution: Black can only have played d7-d5 or f7-f5 as his last move, each allowing an en passant capture and mate next move. But which which of the two did he play? The white pawns must have captured 10 times, which accounts for all the missing black pieces. That means Black's last move cannot have been d7-d5, because in that case the Bc8 was captured on its starting square, and a pawn could not have done that. (This is the retrograde analysis required for this problem). Therefore, Black's last move was f7-f5, and White mates with 1.gxf6 and 2.f7 mate.

In fact, I first saw this problem back in 2003-04 or so, in one of my favorite chess sites "Tim Krabbé's CHESS CURIOSITIES". https://timkr.home.xs4all.nl/chess/chess.html

I highly recommend all chess lovers to read this site. Each article there is a gem!

There, he published this problem along with the solution in December 2001. I saw the above problem and solution in that site sometime in 2003 or 2004.

[Dr.AD]

We will slowly start discussing problems from the games of modern greats such as Fischer, Kasparov, and Anand. Maybe from next week.

For now, given that the weekend is coming, let's warm up with a couple of nice problems, going into the weekend!

Originally Posted by Sutripta

Tal certainly. But not of Fischer? Or will we see his games later on?

Sure, so let's warm up with a position from one of Fischer's games, in which Fischer was White.

Problem 5. White to play and win

The above position is not too hard to calculate, and good as a warm-up.

But the below position will require some precise calculations. This is from one of the all-time greats, the legend, Garry Kasparov!

Problem 6. White to play and win

Note: For Problem 6, you have to calculate all the variations to their logical end, and not just give the first move.