Tag Archives: computer chess

ICCF Rant 2: Procrastination

Leave a Reply

Over the board chess is strictly time-limited and you do have to get on with things. In blitz you have to move so fast there is hardly time to calculate – you have to rely on knowledge and intuition to get you through. Even in the longest time controls at standardplay, you need a sense of purpose. You can’t just sit there wool-gathering.

By contrast, in correspondence chess you have so much time. The standard is 10 moves in 50 days (sometimes 40 or 30 days but it makes no real difference). When you have played 10 moves, another 50 days are added to your clock. If you used all your time then a really long game of 100 moves would take an elapsed period of nearly 18 months out of your life. Who needs that much time to play a game of chess? How long do you need to find the best move and play it?

My own approach is to use up only as much time as it takes and no more. I have the best commercially-available computer with the fastest processor running the strongest chess engines. It takes anything from a few minutes to an hour or two; and then I’m satisfied that waiting any longer is not going to improve the quality of the move, so I make the move. As a result, I always have the maximum time available although I never need it.

Some of my opponents seem to have a similar approach. Our games are played at a fast tempo with several moves being exchanged each day. The record for my completed games is vs the Italian player Mario Palladino where we each made 34 moves in less than one day. In just under 10% of my completed games, each player made at least 10 moves a day. Sometimes I get a message from my opponent at the end of the game thanking me for playing at a fast pace (the unspoken point is that many players slow the game down).

I wouldn’t want people to conclude that I and my opponents in these games think too fast and don’t play the best moves. We take as much time as we need and often that isn’t long at all.

But this post is going to be a rant, so now I’m going to focus on players who take much more time than they need. These seem to fall into two main categories.

First, players who realise they are losing and play as slowly as possible. Either they want to delay their final hours as long as they can, or they simply lose interest and the game becomes their lowest priority in life.

Second, players who take the maximum time available even if they don’t need it. This approach is diametrically-opposed to mine and I am still trying to understand the reasoning. It may be psychological.

Let’s start with an extreme example of the first category. This is my game against Player C. He is from Latvia, he has been playing correspondence chess for over 30 years, and his rating is currently similar to mine, although it was over 2500 at one point. The game is so short that I’m simply going to give all the moves.

Player C v Rodney Barking
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Qc2 0-0 5.a3 Bxc3+ 6.bxc3?!
Very unusual. 6.Qxc3 is almost always played, preserving the purity of the pawn structure. Otherwise why take a move putting the queen on c2 in the first place?
6…c5 7.e4 c5 8.e5 Ne4 9.Bd3 Nc6 10.Ne2 cxd4
All this has been played before. There are 11 games in my database. White normally recaptures with 11.cxd4. But in this game…
11.Kf1??
I do not understand this move. It leaves White a pawn down with misplaced pieces and no compensation. The evaluation goes from about 0.00 to –2.5. Maybe it was a mouse-slip and he meant to play 11.0-0 although even that loses a pawn and White doesn’t have enough for it.
11…dxc3 12.Bxe4 dxe4

Now White can’t take the pawn on e4 because it’s mate next move on d1. Taking the pawn on c3 with the queen leaves Black two pawns up: 13.Qxc3 Qd1+ 14.Qe1 Qxe1+ 15.Kxe1 Nxe5. So does taking on c3 with the knight: 13.Nxc3 Qd4 and the c4 pawn drops. Developing moves like 13.Be3 allow 13…Qd3 with an overwhelming position.

So far my opponent had taken 4 days for his 12 moves, not lighting-fast but still a reasonable pace. But now he stopped. I waited. And I waited. Maybe it was the horror show in front of him but he did not make any more moves at all. 41 calendar days later, the server automatically terminated the game with a loss on time for White. The relevant provision in the 2023 ICCF Rules is section 2.6.2:

If a player does not move or otherwise indicate the player’s intention to continue
during 40 full calendar days (plus 24 hours’ buffer time, not including any leave time) for the same move, the game will be scored as an ETL (exceed time limit) loss.

And there’s a further penalty beyond losing on time. Section 2.4.4 provides:

A player who has exceeded the time allowed shall forfeit the game. If the
player is not under any ICCF suspension when this occurs, the player will be restricted
from registering for any event on the server for a period of 30 days.

So Player C will not be able to sign up for any new events for a whole month. But it does not affect his participation in any existing events. As it happens, Player C is part of the lineup for an event which has already been organised but has not yet started, in which I am also playing. That’s going to be interesting.

I did wonder whether his failure to move in this game might reflect some traumatic event in his life which had taken him away from the chess board and for which he had not taken the period of leave which the server allows. But that is not the case. During the 45 days in which the event has been running, Player C has completed 7 of his allotted 16 games as well as losing this one on time. So it was just a one-off personal tragedy of a game.

That was a case of a player delaying the game through loss of interest. Here’s a more common case where a player realises he is losing but continues to play, although at a much slower rate. I was White against Player D, who is from Poland. At the start of the game he outrated me by about 130 points.

Earlier in the game (by move 18) I had won the exchange through a tactic which my opponent did not see. I guess he was playing without engine assistance at that point. This is the position after White’s move 37. My opponent had already started to slow the game down. His moves 1 to 28 took only 11 days. Moves 29 to 36 took a further 60 days. And now he is clearly losing. My immediate plan is to push the c-pawn to c7, then play Rc4 forcing him to move his knight, then capture his advanced pawn on a2. After this the game is trivial.

My opponent must have worked out this was going to happen because he slowed the game down even further. The game continued 37…Qc6 38.Qxc6 Nxc6 39.Rxa2 after which queens are off the board, I’ve rounded up the a2 pawn, and I have the exchange and two pawns with a completely won position. His moves 37 to 38 took a further 40 days.

The ICCF recognises that players may slow the game down considerably in an attempt to postpone the inevitable. This strategy is known as the “Dead Man’s Defence.” Section 2.15 of the 2023 ICCF Rules concerns the Code of Conduct for players and has this to say:

Extremely slow play in a clearly lost position is not proper behaviour in correspondence chess play, and is subject to a warning from the TD, and will result in disciplinary action if it continues or is repeated in other games. This type of extremely slow play has been nicknamed the dead man’s defense (DMD). The defining characteristics are (a) being in a position that appears clearly lost presuming reasonable play AND (b) dramatically slowing play in that specific game.

The provision is activated when the opponent of the slow player makes a claim. I waited a further 10 days after my move 39 and then contacted the tournament director. He then followed the notification procedure set out in this part of the Rules, requiring my opponent to speed up and start playing at a normal rate again. There are severe penalties for a breach of the Code of Conduct, mainly taking the form of a ban from ICCF events for a certain period. At this point my opponent decided he’d had enough and he promptly resigned the game.

The ICCF does try to hurry players up if they take a long time over their moves. When a player is considering his next move, Rule 2.4.2 provides that real time applies for the first 20 days but after that double time applies (this is known as “duplication”) so that every extra day taken counts as two days on the player’s clock. The rule gives the example of a player taking 23 days over a move, which uses up 26 days on his clock. In other words, 20 days of normal time and 3 days of double time counting as 6 days on the clock.

The penalty for losing on time is based on calendar days (real time) rather than clock days. In the first example I gave, the game against Player C, the penalty was applied after 41 calendar days, but he had actually used up 62 days on his clock. He still had 38 days of clock time left in the game, but his inactivity for 41 calendar days amounted to a loss on time under the rules.

My view – others may not share this view – is that the initial period of 20 days reflection time is too generous. I think the ICCF should reduce this to 15 days or even 10 days. This raises the prospect of players running out of clock time (the clock reaches zero) before the period of 41 calendar days is used up. It would certainly focus the mind. I don’t know whether the ICCF has considered this idea.

But I now realise that this provision is the wrong target and the real problem lies elsewhere in the Rules. This brings me to the case of Player E. I am not going to give any details of his identity or the particular games in question since these are still live. Player E behaves in an extraordinary way. Whatever the standard period of time allowed for a set of 10 moves, he does nothing until almost all the time is used up and then makes a move at the last minute. So for example he might have only one day left for his next 10 moves and then he would finally make a move. You might think he would have to make 10 moves in one day to avoid his clock running out of time. But no. This is what Rule 2.4.2 provides:

Time is counted in days, not in hours, minutes, or seconds. If the first 24 hours has not yet been fully consumed, the reflection time used is counted as zero days. For the next days, a similar method of accounting time consumed will apply. Playing time is accounted for in whole days (24-hour periods). A player will have 24 hours of reflection time to respond to a move before one day of time is charged against the player’s clock by the ICCF webserver.

This means that if you have only 10 minutes remaining on your clock, and you then move, the clock resets itself to 24 hours and that is the time you have remaining for your next move. In effect you can make one move a day indefinitely and you will never lose on time. It’s similar to the incremental time control in OTB chess. You may be down to increment time on the clock (“living on the increment”), but as long as you continue to move within the increment period of, say, 30 seconds, you will never lose on time. The game will continue until it reaches its natural end. This is what happened in the infamous game T.Foster–P.Lalic (Guildford 2023) which ran to 214 moves after 7 hours of play before the final blunder and resignation.

This is being remedied in OTB chess for the next Guildford congress with a stricter time control. Should something similar be done for online correspondence chess? The obvious solution would be to remove the increment so that the clock does not reset itself to 24 hours every time you make a move. This already happens in the alternative timing arrangement known as “Triple Block” used for some ICCF events. Rule 2.4.2 sets out the relevant provision:

Time is counted in days, hours, and minutes. All time while “on-move” used counts and is never given back by rounding to the last full day.

I do not know how long Triple Block has been in use. Maybe this kind of provision is the future. It still leaves me with the problem of how to understand players who deliberately use up almost all of their time rather than only taking the time they need. I don’t think such players are simply disorganised. You have to be organised in order to make one move a day every day until you reach the next time control. I do wonder if the reason is psychological. The player is indulging in mind games by creating an expectation in his opponent’s mind of an imminent win on time, only to dash that hope at the last moment, so as to provoke the opponent into moving too quickly (“blitzing”) and making an error. Who knows? Only Player E can tell us and he’s not saying anything.

I do have one further idea to encourage quicker play. This could be implemented without any changes to the time control arrangements. The server knows how much time has elapsed in every game played by each player in an event. If there is a tie between two or more players and none of the tie-breaks can separate them, for example they have exactly the same number of points, the same number of victories, no advantage in direct encounter, and the same sum of opponents’ scores, then an additional tie-break could be implemented to favour the player who has used the least amount of elapsed time.

ICCF Rant 1: Rating snobbery

Leave a Reply

In what circumstances do you offer or accept a draw in online correspondence chess? I apply a number of tests.

First, I look at the position from the point of view of an experienced OTB player. Is the position simple or complicated? Is it likely to be drawn with correct play? What is my technical knowledge about this particular kind of ending?

Then I check the engine evaluation. If all the lines on all my engines point to 0.00 or close to that, I know the engines don’t think either side is winning. These engines outrate me by over 1000 points. I have to respect that.

Also I take into account my opponent’s play so far. Has he played sensible moves of the kind that a strong human or an engine would play? Does it look as though he is playing with engine support?

If all these boxes are ticked, then I will offer or accept a draw. Notice one test that’s missing – the difference between my rating and my opponent’s rating. I will gain a few points if my opponent is higher-rated and I will lose a few points if my opponent is lower-rated. That’s just how it goes. I am trying to be objective about the position.

In an ideal world, my opponents would take the same approach as me. Many of them do. Unfortunately some of them do not. Let me illustrate that with some examples.

This is from the game Rodney Barking – Player A.

Player A (not his real name) is from England. He has been playing correspondence chess for many years. He is a prominent figure in the world of English correspondence chess but as far as I can tell has never played rated OTB games. At the start of our game, he outrated me by about 150 points.

In this position, I have just played 31.Rbe1. It seemed to me that neither side could make progress. The engine evaluation was 0.00 or slightly better for me across all lines going nearly 50 moves deep. So I offered a draw… which he declined. In correspondence chess you are limited to one draw offer every 10 moves. So I had no choice but to play on.

This is the position after I played 41.Rf4. The position is even more drawish than before. The pawn structure is symmetrical. Neither side has any threats. The engine evaluation is even more firmly 0.00 at over 50 moves ahead. Again I offered a draw, and again my opponent declined.

A few moves later, queens came off leaving a rook and pawn ending that not even Magnus Carlsen could win:

At this point, after his move 49…Rd7, he offered a draw which I accepted.

I have been trying to understand why he turned down the first two draw offers. One possible explanation is that he does not understand the basic principles of chess endings and simply plays whatever move his computer recommends. Maybe he does not know that this type of position is drawn. I think this is unlikely given his high rating and his many years of experience.

A second possible explanation is that he is using a different chess engine that tells him, contrary to the range of engines that I use, that Black is winning these positions and therefore he should play on. Again I think this is unlikely. From what I know of chess engines and how they evaluate positions, none of them would regard this as a winning position. Unless maybe it had become corrupted and had accepted a bribe from Player A in the form of a promise of a processor upgrade.

The only other explanation is that Player A thought his rating advantage of 150 points was sufficient for him to outplay me from a drawn position or that I would make a mistake. In order to think this he would have to set aside the progress of the game so far, in which I (in fact both of us) have played error-free chess. He would also have had to close his mind to the possibility that I might have engine assistance, so that he is not playing a 2200-rated player but a 2200-rated player with the backup of a range of 3500-rated computers.

In short, I think he turned down the draw offers, not because an objective evaluation of the position led him to think he could win, but simply because his rating was higher than mine. That’s rating snobbery and it does him no credit.

Here’s a different example of the same thing.

This is from the game Rodney Barking – Player B (also not his real name). He is an American player who has been playing correspondence chess for less than five years. At the start of this game, he outrated me by over 350 points.

The position is quite different from the other game. It’s unbalanced and tactical in nature. Black has a dangerous advanced b-pawn. I was the exchange up, but have just offered it back by capturing his bishop on e6 (31.Rxe6). He has only one good move (recapturing the rook) – everything else loses – and then my queen and rook combine against his exposed king to force perpetual check resulting in a draw by repetition within a few moves. The engines see this instantly. The lines are forcing, and at all critical points Black’s only choice is between perpetual check or an immediate loss. So I offered a draw… which he declined.

The game continued 31…Rxe6 32.Rc8+ Kf7 (forced since 32…Kh7 33.Qd3+ wins for White) 33.Rc7+ Re7 34.Qd5+ Kg6 (the only other legal move, 34…Kf8, also results in a quick draw after 35.Qd8+ or 35.Rc8+) 35.Qd3+ Kf7 (all other moves allow a forced mate in at most 9 moves) 36.Qd5+ Kg6 and at this point he offered a draw which I accepted.

What made him refuse the initial draw offer? Any engine would have told him that the best he could hope for was a draw by repetition. Was he not using engines and relying on his own natural ability to outplay me in a tactical position?

Or was he simply making the point that chess etiquette requires the draw offer to come from the stronger player? I understand that point, but in online correspondence chess with computer backing, the concept of the “stronger player” no longer means anything. Nor does “higher-rated.” Come on guys, it’s time to join the 21st century.

Visualising the carnage

Leave a Reply

Carnage. Chess engines are now significantly stronger than the best human players. Using an engine makes a real difference to the way you play correspondence chess and the outcome of the game. It’s not in the type of moves you make – engines will recommend the logical rather than the natural move, even if it looks non-standard. Instead it’s in the accuracy of calculation. Human players make tactical errors. Chess engines do not. So if your opponent blunders during the game, you know he’s not using an engine.

This is how Barking’s county captain put it at the start of the season: “An important point to note is that use of computer engines is permitted – nearly all players in the first and second divisions use engines and many in Division 3… [to Barking, the newbie:] I expect you’ll be using an engine; if not please let me know and I’ll put you in the Div 3 team.”

In tournaments at a higher level, where the use of engines is routine, the engine v engine contest is very often a draw. For example in one of my current ICCF tournaments, a 7-player all-play-all with an average rating of 1950, 20 out of the 21 games have finished and 17 of these were drawn with only three decisive games (and one of these was a win on time). The only game outstanding is one of mine, against the tournament’s highest-rated player. I am winning this, and it’s only still going on because my opponent has slowed the game to a crawl.

As you might imagine, a game of engine vs non-engine can be a real mismatch. This type of game is often a miniature (under 25 moves) with strong tactical strokes. Which leads me to…

Visualisation. It is a truth universally acknowledged, that the ability to calculate accurately is one of the most important skills in chess. Training and testing this skill is the basis of one of the most popular types of chess book – the tactical puzzle book. These give you a diagram and ask you to find the continuation, essentially calculating what should happen next, without moving the pieces.

What’s less well understood is how you go about developing the ability to calculate. Enter visualisation – the practice of calculating variations without looking at the board at all. Visualisation without the board is the most effective way of developing clarity and the depth of calculation. This is what the specialists in blindfold chess do all the time, sometimes in a spectular way through simultaneous exhibitions against many players.

This is the subject of chapter 2, ‘Blindfold Chess and Stepping-Stone Diagrams’, in Jonathan Tisdall’s book ‘Improve Your Chess Now’ (Cadogan, 1997). He quotes Dr Siegbert Tarrasch:

The whole game of chess is played as a blindfold game. For instance, every combination of five moves is executed mentally, with the only difference that one has the board before him. The pieces which one is looking at very often hinder the calculations.

Tisdall draws attention to the practice of some GMs (Shirov, Ivanchuk, Svidler, for example) who “often calculate variations by suddenly staring into space instead of at the board. Clearly, they have some built-in belief that they can more clearly focus their visualisation of critical variations by looking away from the board.”

You can try this at home without going out to find an actual opponent and playing an actual game. Take the score of any chess game (I don’t mean the result, but the set of moves as seen on the page), and at first the shorter the better, and try to play through the game in your head. See if you can visualise the position after each move. This is hard work, particularly when you first do it because the relevant chess muscles are weak and undeveloped, and the further you go into the game, the harder it becomes to maintain a clear mental picture of the board. But it is rewarding and it is a very good way to improve your game.

A similar method is to go through a book of chess games with diagrams at intervals in each game. Try to visualise the game without playing the moves on a board, using the diagrams as an aid or stepping-stone (another technique recommended in Tisdall’s book).

For starting out, I recommend ‘Blindfold Opening Visualisation: 100 Chess Puzzles’ by Martin B Justesen (Say Chess Publishing, 2021). This gives you the score of 100 games take from Lichess, all of 10 moves or fewer, and featuring tactical points, where you have to find the continuation.

Which brings us back to the carnage of engine-inspired destruction. Here are four of Barking’s miniature wins from 2022. Play through them with a board if you like, but if you feel up to it, play through them blindfold, and see if you can keep track of what’s going on.

Game 1
Start: 23 May 2022, End: 3 June 2022 (11 days)
White: Don Wade (USA – 1780), Black: Rodney Barking (ENG – 1800)
B99: Sicilian, Najdorf (7…Be7 main line)
1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Bg5 e6 7.f4 Be7 8.Qf3 Qc7 9.O-O-O Nbd7 10.Bd3 b5 11.g4 Bb7 12.Rhg1 h6 13.Bxf6 Bxf6 14.Qe3 Qb6 15.Nce2 g5 16.Kb1 gxf4 17.Qf2 e5 18.c3 O-O-O 19.Rgf1 exd4 0-1

Game 2
Start: 23 May 2022, End: 10 June 2022 (18 days)
White: Rodney Barking (ENG – 1800), Black: Horst Wilshusen (GER – 1044)
C18: French, Winawer, Classical variation.
1.e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5 c5 5.a3 Bxc3+ 6.bxc3 Qc7 7.Qg4 f5 8.Qg3 Nc6 9.Nf3 Nge7 10.Qxg7 Rg8 11.Qxh7 cxd4 12.Bb5 Bd7 13.Bxc6 Bxc6 14.O-O dxc3 15.Bg5 Rf8 16.Nd4 Qd7 17.Qh5+ Kd8 18.Qh6 Re8 19.Nxe6+ Kc8 20.Nd4 Kc7 21.e6 Qd8 22.Nxf5 1-0

Game 3
Start: 1 October 2022, End: 20 October 2022 (19 days)
White: Rodney Barking (ENG – 2016), Black: Alan Ruffle (ENG – 1509)
C41: Philidor, Improved Hanham variation
1.e4 d6 2.d4 Nf6 3.Nc3 Nbd7 4.Nf3 e5 5.Bc4 Nb6 6.dxe5 Nxc4 7.exf6 gxf6 8.Nd4 Rg8 9.Qe2 Ne5 10.f4 Nc6 11.Be3 Bh6 12.Qh5 Nxd4 13.Bxd4 Bxf4 14.O-O Be5 15.Bxe5 dxe5 16.Qxh7 Rg6 17.Qh8+ Ke7 18.Nd5+ Kd7 19.Nxf6+ 1-0

Game 4
Start: 1 October 2022, End: 23 October 2022 (22 days)
White: Francis Watson (ENG – 1800), Black: Rodney Barking (ENG – 2016)
D94: Grünfeld, 5.e3
1.d4 Nf6 2.c4 g6 3.Nc3 d5 4.Nf3 Bg7 5.e3 O-O 6.c5 Nbd7 7.Be2 Re8 8.O-O e5 9.Qb3 c6 10.Rd1 e4 11.Nd2 h5 12.a3 Ng4 13.h3 Nh6 14.Qa2 Qg5 15.Nf1 Nxc5 16.dxc5 Bxh3 17.Ng3 Nf5 18.gxh3 Nxg3 19.fxg3 Qxg3+ 20.Kf1 Re6 0-1

The exchange sacrifice

1 Reply

There are many ways to create an imbalance in chess. For example you can alter the pawn structure, obtain the bishop pair, castle on opposite sides. But the most obvious way is to sacrifice material. Gambits usually involve sacrificing a pawn for the initiative or some positional gain. Sacrificing a whole piece is more spectacular and generally relies on a short-term tactical justification where you might recover the material with interest or even deliver checkmate.

The exchange sacrifice is in a category of its own. You give up a rook for your opponent’s bishop or knight. Sometimes you get a pawn for it, sometimes you don’t. You’re looking for an enduring positional advantage based on superior piece activity. The classic example is in the Sicilian where Black sacrifices his rook for the white knight on c3, wrecking his opponent’s pawn structure and sometimes picking up the e4 pawn as well. Here’s Kasparov showing how to do it, on the black side of a Sicilian Najdorf against Movsesian in a tournament in Sarajevo in 2000.

13…Rxc3! 14.bxc3 Qc7
What has Kasparov got in return for sacricing the exchange? No extra material, but White’s queenside pawns are wrecked and Black’s minor pieces are ready to join the attack.
15.Ne2 Be7 16.g5 0-0 17.h4 Na4 18.Bc1 Ne5 19.h5 d5 20.Qh2 Bd6 21.Qh3
Both sides are attacking on opposite wings, but Black’s pieces are much more active and co-ordinated and his attack is much more dangerous.
21…Nxd3 22.cxd3 b4 23.cxb4 Rc8 24.Ka1?!
This gives Kasparov the opportunity to detonate the centre, when his pieces swim like sharks towards the defenceless white king.
24…dxe4 25.fxe4 Bxe4!
White can’t take the bishop: 26.dxe4 Be5+ and mate follows shortly.
26.g6 Bxh1 27.Qxh1 Bxb4 28.gxf7+ Kf8 29.Qg2 Rb8 30.Bb2 Nxb2 31.Nd4 Nxd1 32.Nxe6+ Kxf7 0–1
White’s knight forks king and queen, but he can’t take the queen because it’s mate in one.

Actually this is just the most famous kind of exchange sacrifice. There are many different ways to do it. Their sheer variety is one of the themes to emerge from a new book by Kotronias entitled ‘The Exchange Sacrifice according to Tigran Petrosian’. Kotronias argues that the former Soviet-Armenian world champion was the greatest exponent of the exchange sacrifice. The book brings together many examples from his games, usually successful but sometimes not, and all annotated in detail. This is from the game Troianescu–Petrosian, Bucharest 1953, in which Petrosian manages to sacrifice the exchange not once, but twice.

25…Rxe4! 26.Bxe4 Bxe4
This time Black gets a pawn for the exchange. His central pawn majority and active pieces provide more than enough compensation.
27.Nc2 d5 28.Nd4 b4 29.cxb4 axb4 30.a4 Qa7 31.Qf2 Rc8 32.b3 Bf8 33.Nb5 Qa6 34.Qe2 Qb6+ 35.Kf1

35…Rc3
Here we go again… but this time Petrosian is grandstanding. He probably couldn’t resist giving up a second rook, but a better plan is 35…h5 intending to advance to h4, open the h-file, and bring a rook to the side with a very dangerous attack.
36.Nxc3?
Taking the material, so that temporarily White has two rooks for two bishops, but it soon becomes clear that the bishops are the stronger pieces.
36…bxc3 37.Rc2 Qxb3 38.Rec1 Bb4 29.g4 Bxc2 30.Rxc2 Qxa4
That’s one exchange back, with a lot of extra pawns. White can do nothing about this. Petrosian gradually advanced his pawns and White resigned on move 57.

The exchange sacrifice is part of the strategic arsenal of every strong player. What interests me is that this form of material disdain has crossed the boundary to computer chess. Traditionally the engines would prefer to take material and hang on to it. But now they know that material gain is not everything. They are prepared to sacrifice for long-term gain in the greater interest. The final example in today’s post is from another correspondence game. Jean-Paul Dellenbach (2079) v Rodney Barking, ICCF 2022.

White has just played 23.Bh6, attacking the rook and no doubt expecting it to move. But, on advice from the computer, I left the rook where it was.
23…dxe4 24.Bxf8 Rxf8 25.dxe4
I didn’t get any pawns back. But the computer has worked out that White is now very weak on the dark squares and will have trouble defending them given than his pieces are passively placed.
25…Bc7 26.Bc2 Qd6 27.g3 Re8 28.Kh2 Re5
Bringing the last piece into the attack. The rook is destined for the h-file where it will bear down directly on the white king. I’m not sure I would have thought of this manoeuvre unaided.
29.Qf2 Qe7 30.Nf4 Nxf4 31.gxf4 Rh5 32.Qe3 Bc8 33.Rf3 g5!
Exploiting the pin on the f4-pawn to open the position further.
34.e5 gxf4 35.Rbxf4 Qxe5
White decides to return the exchange, but this doesn’t stop the attack.
36.Kg2 Bxh3+ 37.Rxh3 Qxe3 38.Rxe3 Bxf4 39.Rh3
The smoke has cleared and White is now two pawns down. His best hope is to get the rooks off the board and reach an opposite bishop ending with drawing chances despite the material deficit. I didn’t need a computer to tell me that rook and opposite bishop endings are anything but drawish, so it was an easy decision to keep the rooks on even though White now gets one of his pawns back.
39…Rc5 40.Bxh7+ Kg7 41.Bd3 Be5 42.Rh7+ Kf8 43.Rh6 f6 44.Rh7 Rxc3 45.Bxb5 Rc2+ 46.Kf3 Rxa2 47.Ra7 a3
Two safe pawns up again, with rooks and bishops still on the board, Black is winning this ending. White could have prolonged the struggle but blundered on move 53 and resigned.