Tuesday, July 28, 2020

Calculation in the Engine Age: A review of Jacob Aagaard's "Excelling at Chess Calculation"

I certainly enjoyed reading this book on calculation (and plan to give Aagaard's book on calculation in his Grandmaster Preparation series a try), but have mixed feelings about it. First the positives and then the negatives.
On the positive side, even now in the golden age of chess books there are very few books on calculation, and Aagaard's is probably as good as any. I like Aagaard's emphasis on the importance of finding candidate moves because, as he says, the most important errors are often right at the beginning of the calculation process when a key possibility is ignored. Finding good candidate moves is, he argues, more important to success than the accuracy of long calculations. Aagaard gives some interesting suggestions about how get better at finding candidates. This part of the book is original and first rate. Otherwise, most of what Aagaard says about calculation techniques and when to calculate is not especially new. Aagaard draws extensively from the legendary trainer Dvoretsky as well as Jonathan Tisdall's discussion of the use of stepping stones in visualization in his book Improve Your Chess Now. Not that this is a problem. Aagaard freely acknowledges his borrowings. His presentation of the key ideas is engagingly written, and most of his illustrative examples are excellent.
My reservations about the book have to do with the question of how to approach calculation in the age of chess engines. Much of Aagaard's analysis is engine reliant (whose isn't nowadays?), and some of the moves and variations he considers would be very difficult for all but the best players to get right the board. In fact, they are so difficult that some of Aagaard's computer aided analysis contains significant errors. One important example of this is his discussion of the 2003 game between Riazantsev and Arbakov on pages 79 to 82. Aagaard criticizes Arbakov for his 21st move and gives an alternative that he claims is better and makes black's defense easier. But the move he gives simply does not work.

(For those who want to explore this, in the above position Arbakov's move was 21..Nxd5, Aagaard's move is 21...Kg7, which does not work because of 22. Qd2, a move that Aagaard skims over, saying that it "will transpose in most lines."(page 80) Stockfish 11 at a depth of 49 gives a 3.33 pawn advantage to white after 21...Kg7; in other words, it considers it losing.) My guess is that in 2004, when Aagaard wrote the book, chess engines simply did not see the refutation of Aagaard's move that they now see in 2020. In actual fact, Arbakov played the best 21st move--that Stockfish 11 can find. This would not be all that important except for the fact that Aagard uses Arbakov's supposed error to issue a more general criticism of him. "Arbakov is not really interested in calculation" (page 81) and basically became a grandmaster because of his natural talent, Aagaard argues. He has not progressed beyond being a normal grandmaster, Aagaard suggests, because he has not put in the hard work needed to go further. Am I wrong to detect a touch of resentment here by Aagaard who had put in hard work and had not yet become a grandmaster when he wrote these lines? Similar insinuations about GM Joel Benjamin (p. 51) indicate that talented players who don't rise to their full potential are a pet peeve of Aagaard.  
Putting the criticism of Arbakov aside, this example points to a larger problem with Aagaard's book. If Aagaard (a conscientious author and international master when the book was published) can't get his home analysis right with the help chess engines, how can an ordinary player hope to get it right over the board? Of course, the best players will get it right most of the time, but how do these examples help improving players (who are, after all, the audience for his book) improve? Some easier examples showing typical errors of less illustrious players playing simpler positions would be helpful for the target audience, described by the publisher as "club and tournament players." After all, as Aagaard himself, wisely says, "one of the most important aspects of training is to build confidence and motivation which is not achieved by trying to solve far to difficult exercises." (page 17) Part of the problem, I suspect, is that Aagaard did not really write this book for "club and tournament players," but rather for himself as part of his eventually successful effort to become a grandmaster. This shortcoming of Aagaard's book is symptomatic of a larger problem with chess books, which rarely give much thought to issues of pedagogy, but that is a topic for another day.
Beyond this, what is missing from Aagaard's book is serious reflection on how incredibly difficult it is to consistently calculate accurately and what this means for practical, over the board chess.  The great world champion Mikhail Tal used this to his advantage by forcing his opponents to calculate and repeatedly won games in which his attacks were objectively flawed, but too difficult for his opponents to refute over the board. Other great players, (I think of Ulf Andersson) although excellent at calculation when they had to do it, preferred to avoid its risks and often won by putting positional squeezes on their opponents that did not require difficult calculations. It may seem strange to want to talk about games won without complicated calculations in a book on calculation, but it is important to know when and how to steer a game toward or away from calculation. More attention to how calculation fits into broader practical, human strategies for winning games would be useful.
Finally, Aagaard said on the Quality Chess blog that "the exercises are not that great [in this book]; I could skip them." I did not try them, so I can't say if he is correct, but I will take his word for it.
For the record, I play over the board tournaments, and my rating is currently 1887 USCF. I know it is a modest rating, but it places me in the 93rd percentile of active USCF tournament players, which should make me part of the intended audience of this book.
I don't imagine that this book would be very helpful for a player rated below about 1800 USCF. Lower rated players might find Andrew Soltis's The Inner Game of Chess: How to Calculate and Win a better book for them. I read it when I was rated around 1600 and found it useful.

Wednesday, July 2, 2014

Beating Inferior Openings: 4.Bc4 in the Four Knights Revisited

In his July 2014 ChessCafe.com review of Or Cohen's A Vigorous Chess Opening Repertoire for Black (2013) Carsten Hansen criticizes this Petroff repertoire book for giving three pages to the Four Knights with 4.Bc4?!, which Hansen considers to be a line that "should have received little or no mention." As I have argued in my first posting on this opening, Hansen is simply wrong when it comes to amateur chess. Amateurs need to learn how to refute such lines, which they will see orders of magnitude more often than the latest theoretical lines played by titled players. More to the point are Hansen's other criticisms of Cohen's book, notably its excessive reliance on long lines of computer analysis and its lack of clarity on its intended audience. Cohen's coverage of the 4.Bc4?! variation of the Four Knights is appropriate for rank amateurs, but its long lines without verbal comment are hardly useful for them. Most amateurs below about 1900 USCF need, I humbly submit, analysis like that offered in this blog's coverage of this line: lots of verbal and visual explanation of the most basic issues in an opening.

Friday, June 20, 2014

Rook Endings 4: Rook Versus Two Isolated Pawns.

We considered the fascinated case of rook versus two connected pawns in the previous post in our rook endgame series. This time the focus is on the rook versus two isolated pawns ending. Let's get right to it.

Second Case: Rook Versus Two Isolated Pawns.

In most rook versus two isolated pawn positions the side with the pawns is defending. The further separated the pawns the more difficult it is for the player with the pawns. This is illustrated clearly by the fact that pawns separated by four or more files can't defeat a rook on the 8th rank.

Monday, June 16, 2014

Rook Endings 3: Rook Versus Two Connected Pawns.

Endings in which a rook faces two pawns are more common and difficult than rook versus single pawn endings. There are 420 games in my mega database from the beginning of 2013 through mid-June 2014 in which this ending is played for four or more moves. 57 or 13.6% of those 420 games did not reach the result expected by the tablebases (meaning that there was a result-changing error). Some of those errors were, as we shall, by very high rated players.

There are basically two types of rook versus two pawn endings:
  1. Where the pawns are connected.
  2. Where they are isolated.
We will cover the case of connected pawns in this post.

Sunday, June 15, 2014

Rook Endings 2: Rook Versus Pawn Continued

In a prior post we examined the key ideas in rook versus lone pawn endings. In this post we will look at some special and complicated cases of this ending. Specifically, we will be looking at the following:
  1. A non-trivial case where the side with the pawn is winning. 
  2. A case in which the rook is between the pawn and its queening square and the attacking king on the other side of the pawn.
  3. The special case of the rook pawn.

Friday, June 13, 2014

Rook Endings 1: Rook versus Pawn

Rook endings are among the most common endings and playing them well is one of the most important and subtle endgame skills. To improve my chess (and hopefully yours as well) I am beginning a series of blog postings on rook endings. We are going to begin with the most basic rook ending: rook versus a lone pawn. These seemingly simple endings can actually be quite tricky and are mishandled more often then you might expect. Since the beginning of 2012 through mid-May 2014, 523 games in Chessbase's mega database reached a rook versus pawn ending that was played out for four or move moves. 47 or 9.2% of those 523 games did not reach the result expected by the tablebases at the moment the rook versus pawn ending began. To be sure, many of the poorly handled endings were by weak players who made silly mistakes, but not all of them. Even grandmasters erred. Clearly, this endgame is not trivial. Its key ideas need to be mastered. They are the following:

Attacking ideas:

1. Cutting the enemy king off at his fourth rank.
2. Gaining or losing tempi with rook moves.
3. Avoiding stalemate.
4. Getting the king in the game.
5. Moving the king to the side of the pawn opposite the enemy king.

Defending ideas:

1. Shouldering away the enemy king.
2. Promoting to a knight.

Let's look at them one by one, and then we will examine a few cases in which more than one idea is used.

Thursday, June 5, 2014

Mate in 549!

Mate in 549 is the latest discovery of the position-crunching supercomputers that calculate tablebases. Tablebases are databases of chess endings that have calculated the optimal moves until a draw or checkmate results. They can be purchased to put on your personal computer. Chessbase, for example, sells a product called "Endgame Turbo" which gives you the solution to all endings with five or fewer pieces and some endings with six pieces on the board. The data for this comes on nine DVDs and occupy 43 gigabytes on your hard drive when fully installed. That may sound like a lot of memory, but the full six piece endgame tablebase requires 1,200 gigabytes. Don't want to pay or clutter your computer with chess moves? You can also plug positions into online tablebase tools to find the chess truth that you seek.

Recently, seven piece tablebases were calculated by a supercomputer at Moscow State University and occupy 130,000 gigabytes. Among the nuggets of chess knowledge uncovered was that the longest forced mate with seven or fewer piece on the board is in 549 moves. Want to try to solve it? The position is as follows:

White to Mate in 549

Can't visualize it to the end? Give up? The solution is here.

After seeing this, I can't help but conclude that chess will never be solved. No computer will every play it perfectly, and the royal game will never be exhausted.