John Watson Book Review (65)
The End [game] Comes Before We Know It
IM John Watson - Monday 31st January 2005
DVD Endgame Turbo 2: Nalimov Tablebases; ChessBase 2003
Endgame Challenge; John Nunn; 256 pages; Gambit 2002
Chess Endgame Training; Bernd Rosen; 176 pages; Gambit 2003
Test Your Endgame Thinking; Glenn Flear; 160 pages; Everyman 2002 [Author of 'Improve Your Endgame Play' (2000) and 'Mastering the Endgame' (2001)]
Fritz-Endgame Trainer Pawn Endings ; Martin Weteschink; CD ROM; ChessBase 2002
Chess Endings Made Simple; Ian Snape; 144 pages; Gambit 2003
Dvoretsky's Endgame Manual; Mark Dvoretsky; 384 pages; Russell Enterprises 2003
Dvoretsky's Endgame Manual; Mark Dvoretsky; CD; ChessBase 2003
Modern Endgame Practice; Alexander Belyavsky & Adrian Mihalchishin; 208 pages; Batsford 2003
Winning Chess Endings; Yasser Seirawan; 239 pages; Gloucester/Everyman 2003
Etcetera! The last column that I wrote about endgame books/products had 16 of them. Here I've listed 10 more (including two CDs but without counting Flear's earlier books). Just looking at my shelf, I see that I could have listed yet another 9 such books and more CDs published since 2000, and that's from fairly mainstream publishers. I've read about but not received quite a few more (including Fine's updated classic Basic Chess Endings). Furthermore, nearly every beginner's book and improvement book contains a section on endings. Of course there are more total opening books in the world, but they deal with more varied and separate subjects like the Queen's Gambit and Sicilian Defence, which themselves branch into dissimilar variations of each opening. In this sense endgame books are easily the most frequent specialized genre books, and perhaps even on a numerical par with improvement books and beginner's books (I have fewer of the latter, so I'm not sure). I will discuss a most of the products above and merely alert the reader to the existence of the another. As you will see, these books and CDs vary wildly in both their approach and with respect to the level of player they are trying to address.
One of the most impressive endgame products that I've seen, and one that I've already made considerable use of in writing, is the Nalimov Tablebases DVD from ChessBase. The 5 discs that make up the product contain the exact solutions to literally every endgame with five pieces or less, the kings counting as pieces. They also include some endings with 6 pieces such as knight and 2 pawns versus rook, rook and 2 pawns versus rook, and the like. By 'solutions' I mean that every possible move from every possible starting position from these endings is written out on the disc. Thus when you use a ChessBase playing program such as Fritz, Junior, Shredder, etc. or the analytical engine within the ChessBase program, the solutions appear instantly (with not the slightest actual analysis required by the machine) in the analysis window. You can use the tablebase 'engine' directly in ChessBase to get not just the 2 best moves, but every possible move with the evaluation and number of moves to mate. Not only is there an instant solution to any 5-piece ending that you put on the board, but the 'Endgame Turbo' allows the program to anticipate the table solutions in analysis and thus know whether to go into a particular endgame. There are a few qualifiers, e.g., the 'anticipatory' function works with the stand-alone programs but not directly with ChessBase. Furthermore, only Fritz8 and HiArcs can use the 6-piece solutions.
You can use the tablebases right off the CD, or you can install them on your hard disk. You won't likely install them all for space reasons. For example, I have the 3- and 4- piece endings and what I consider the most common and useful 5-piece endings on my hard disk. They use up 1.38 GBytes on my hard disk. If I chose to install all of the tablebases (all 5 discs) it would use up nearly 24GBytes! That's a bit much for most of us, and the space-devouring 6-piece endings can be efficiently run off the CD anyway.
Here's a simple example that I took from John Nunn's recent book Endgame Challenge. It's a study, with White to move and win:
The tablebase find 1.Ka3! instantly, announcing mate in 23 moves. It assesses all other moves as draws (otherwise the study would be busted – use of these tablebases has led to surprisingly many such busts and doubles). In a normal ending (not a study) there would likely be multiple moves leading to a win, listed according to how many moves there would be until checkmate. Back to the example: after White plays 1.Ka3, we find a panoply of legal Black replies in the order of their assessment. Thus 1...Kc6 is the mate in 23, all the way down to 1...Nf8 which yields mate in 6. Because it's a study, only one White move will win by force, for example, 1.Ka3 Kc6 has only one solution: the amazing 2.Ka2!!. Perhaps an engine without access to tablebases would eventually find this solution, but only after a long search and many false assessments.
The solution (with best play) goes 1.Ka3 Kc6 2.Ka2 Kc5 3.Nd6 Nf6 4.Ne4+ Nxe4 5.e8Q, and here it's still 18 moves until White mates but you get the idea.
This segues into Nunn's book, which deals about finishing off the game although not so directly as a traditional endgame book. The title, incidentally, varies only slightly from John Hall's Endgame Challenge! (with it's final '!') that I reviewed in an earlier column. Nunn's brilliantly organized and explicated book is a collection of 250 studies with 200 pages of solutions! I'm going to steal the key points of his Introduction since it describes the material in the book better than I could. I looked at only 3 of these studies and couldn't solve any of them, my only excuse being that I lazily jumped to the solutions before giving the positions enough thought. But there's absolutely no way that I could have solved the great majority in any reasonable time, since they tend to contain numerous brilliant ideas rather than a single one. In fact (looking at the solutions) I can see that sometimes a study with variations on a known theme is actually harder to implement than a study without such themes, because the subtle solution of the former involves the avoidance of so many tempting manoeuvres that one has seen before but aren't effective in the exact situation. This is a terrific if somewhat esoteric book for the lover of beauty in chess. It is also good mental exercise for the developing player if he or she can maintain the discipline that is required by such advanced material.
The next books are aimed at teaching the students by examples. The first is Chess Endgame Training by Bernd Rosen. This fairly basic book came out in German in 1995 with a second edition in 2001. A training book tends to be centered around exercises, and this book is no exception. Each chapter has examples (as many as 20) that emphasize particular themes, with the solutions at the end of the chapters. Rosen logically devotes the most time to pawn endings and rook endings, about half of the book. Those endings are not only frequent and fundamental, but they are more easily categorised than others. All of the most common endgame themes are illustrated, normally by master games and studies. Rosen used this material in courses about the endgame. I like this book and think that it's best used by players who are no longer rank beginners but at any other level up to 1800 or so.
Test Your Endgame Thinking by Glenn Flear is similarly organised around exercises, with solutions in separate chapters at the end of the book (which is not very user-friendly). Right off we see that it is aimed at a higher-level audience, beginning with rather complicated tests of strategic thinking. Even the 'Basic Endgame' chapter is well beyond what is usually associated with that term. Flear's goal is to teach one how to think about endgames and develop a sense of strategy and planning in ones play. Unfortunately I don't see much of this beyond the first two excellent chapters, and the 'hints' that accompany the later exercises in Chapters 4 and 5 (wrongly entitled 'hints for Chapters 5 and 6') are primarily one-line tips that have nothing to do with strategy. This is a well-written book, but I would recommend Flear's earlier work Improve Your Endgame Play, which includes more basic theory while examining endgame ideas in a more thematic fashion. That book has a great deal of briefer but clear and specific advice. Flear's Mastering the Endgame is also a good choice but it is a collection of endgames and a more advanced book that bypasses a lot of basic theory.
Alas, I haven't had time to examine Martin Weteschink's Fritz-Endgame Trainer Pawn EndingsCD from ChessBase, but it sounds like an excellent training tool for those who prefer a mouse and software program. According to ChessBase's description on the case, Weteschink offers 100 training positions from elementary to top level. These endings are annotated thoroughly. Then there are 50 studies provided; one can play the positions against Fritz. I wish I could say more, but I do know that Chess Base has a training mode that is provocative and useful. The moves are essentially 'covered up' by the program and one is challenged to find the next one, with or without hints. You can always skip moves if you wish and proceed to the next question. This is a product worth looking into.
Ian Snape's Chess Endings Made Simple is the most basic book in this list (Seirawan's is comparable but lengthier). Even its considerable section of exercises from master games sticks closely to the basic endgames he discusses. That is unusual in an endgame book, where the exercises are usually far more difficult than the theory presented. Snapes begins with pawnless endgames, although he skips the basic mates and attacks the more difficult cases of the knight-and-bishop mating process, queen versus rook, and rook-and-bishop versus rook. Thus even this book is not appropriate for absolute beginners. But every elementary aspect of king-and-pawn endings is explained (opposition, triangulation, the square, etc.), and the most elementary endings with bishops (including those with the same and opposite colours) are covered clearly and simply. Similarly, rook-and-pawn endings are given a very brief but clear treatment, including even the Vancura position.
The only issue with Chess Endings Made Simple is that the material has been covered in so many books over the years, usually with a bit more depth. The range of players addressed seems to me to be from just past beginner to a developing player who is not ready for something like Rosen's work above. This would be a good book for a teacher to use in a first course on endgames, depending upon whether he or she prefers Snape's style over other books aimed at this level.
Dvoretsky's Endgame Manual is quite simply a masterpiece of research and insight. It is a tremendous contribution to endgame literature, certainly the most important one in many years, and destined to be a classic of the literature (if it isn't already one). The famous trainer Mark Dvoretsky has put together a vast number of examples that he has not only collected, but analysed and tested with some of the world's strongest players. This is a particularly important book from the standpoint of clarifying, correcting, and extending the theory of endings. Most of all, Dvoretsky's analysis is staggering in its depth and accuracy.
However, before discussing the specifics of Dvoretsky's Endgame Manual [henceforth 'DEM'], a word of warning is in order. I must emphasize that this is a terribly advanced work that I don't think is a very good way for the average player to study the endgame. The majority of the examples are complex and position-specific, and neither the average student nor even strong masters will follow or play over most of the hundreds of positions that are given extensive analysis, not to mention the subvariations derived from those positions. Even when introducing 'the basics', Dvoretsky's approach is often more complex than is necessary for an average student, and in any case such a thick book will seldom be used for the sake of elementary instruction. The majority of the other material is frankly very difficult. So take note: I don't want to be blamed, in praising this book, for your purchasing something that you find intimidating, relatively dull, or otherwise unsatisfying. That said, if you are up to a real challenge and have a great deal of time to devote to reading and playing over examples you will inevitably derive great value from this work.
DEM is a translation from 'the Russian text' (the year of a Russian edition doesn't seem to be given; the book also previously appeared in German in 2002 as Die Endspieluniversität). The translators are not credited on the cover or back of the title page, but I found them at the end of the Author's Preface: Jim Marfia (who has translated many books) and Valery Murakhveri. This was quite a project: Dvoretsky cites help from Karsten Mueller, Taylor Kingston, Harold van der Heijden, and publisher Hannon Russell, a high-quality set of assistants! In addition, Arthur Yusupov and Jacob Aagard write an Introduction and Preface and are quoted in the back cover copy.
Traditional readers will prefer the physical (softcover) edition . Dvoretsky uses a scheme to highlight ideas in the book: basic parts of endgame knowledge are shown in blue print, less important variations in Black, and important rules, recommendations and names are given in boldface. I found this an attractive way of differentiating material.
But those who are not averse to reading by computer have a particularly attractive option in the CD version of DEM produced by ChessBase. From my point of view, the ability to play over all the examples and notes by clicking with a mouse outweighs the annoyance of reading on a computer screen. One can go back and forth between lines at will and use an analytical engine or tablebases to assist in examining the position. The exercises are also easier to work through, since solutions are hidden on one screen.
Let's return to the content. By 'endgame' Dvoretsky means 'the stage of the game when at least one side has no more than one piece (in addition to the king).' Thus various multi-piece endgames are not covered, but this modest restriction still leaves us with 350+ very densely-packed pages. In general, the guidelines (e.g., what material configuration normally wins or draws, 'Black usually loses if his king is cut off...'), rules (the bishop is a long-ranging piece, the rook should be placed alongside the passed pawns, ...etc.) names, and tips are not much different from those given in other sources. But they are gathered together in one volume and tend to carry more authority than statements in earlier books that were not based upon thorough investigation and computer-checking. This alone makes DEM valuable as a general reference. Mueller and Lamprecht's Fundamental Chess Endings is usually more convenient in this regard, but DEM outdoes all other works that I know of.
For this book, Tablebase collections are used to supplement or create the analysis. Although I can't find them accredited either in the book or on the corresponding CD, some form of them is obviously employed (apparently Ken Thompson's famous ones, since 'BELLE' is cited a few times). The relevant endings come up very often because Dvoretsky's manual is limited to endgames with no more than one piece on each side (excluding pawns). Because we reviewed Tablebases above, it would be interesting to follow two examples and see how Dvoretsky's book was influenced by them. I decided to look at the Q+P versus Q endings because of the confusion that has always existed around them. Here's one of the examples that I found in which Dvoretsky explicitly mentions computers (bracketed notes are given by tablebases, as far as I can tell, with a few of my comments):
L Shamkovich – H Wirthensohn, Biel 1980
'We shall not go deeply into the intricate lines; instead, we merely want to match the computer evaluations of the actual moves with the general considerations that are already known to us from the annotations to the previous example. Because of the rook pawn, Black can hope for a successful defense. And, as a matter of fact, the position was still drawn after ...Qg3 or ...Qc4. [jw: These are in fact the only 2 moves that draw; they both instantly appear on the screen because of the tablebases. Every alternative is given with the exact number of moves until mate]] 79...Qg7+? This move would have made sense if the series of checks could continue. However the white king is superbly placed on the rank adjacent to his adversary's, and even one single check will not be possible after White's reply. Hence Black's move is bad. It allows White to rescue his queen from boredom with tempo. [79...Qg3= ; 79...Qc4=] 80.Qf7!+- Qg3 [80...Qe5+? 81.Qe6+] 81.Qf6+ Kc7 82.Qg5?? L. Shamkovich only worsens the position of his queen, moving it closer to the edge. He should have pushed his pawn in order to obtain a position that can be won in ... 69 moves (!). [jw: Dvoretsky is using the line beginning with 82.h5! Qa3+ 83.Kf7 Qa2+ 84.Kg7 Qg2+ 85.Qg6 Qb2+ 86.Kh7 Qb3 87.h6, etc.] 82...Qa3+ 83.Kf7 Qb3+ 84.Kg7 Qc3+? A drawn position (not a draw as such – Black would have spent a good deal of sweat for it) could be maintained after a check from b2. Detailed analysis is not required to see that the black king will get in the queen's way in some lines now. [84...Qb2+=] 85.Qf6 Qg3+ 86.Kh7?? But this is not merely an error; this is neglect of principles. As we have stated, the king should not seek exile in front of the pawn. Both Qg5 and Kf7 are winning, but Kf8! is the most precise. [Tablebases: 86.Kf8!+-; all these moves appear instantly] 86...Qh3 87.Qg5 Kb6?? This counter-error is also very instructive. As we know, there is a drawing zone near a rook pawn, and this zone is rather spacious (its precise borders depend on the placement of the pieces, and most important on how far advanced the pawn is; we shall not give precise definitions here). The king was already standing in the zone, therefore many queen moves were not losing, but the most logical decision was to go towards the pawn: ...Kd7(d6)!=. [87...Kd7= Also some other moves are possible.] 88.h5 Qd7+ 89.Qg7? White worsens his queen's position. All king moves were winning. 89...Qh3?? The black queen had to guard the central squares. [89...Qd5!= was good enough for a draw.] 90.Qe5!+- The white queen has finally arrived in the center, Black's king is out of the drawing zone – White's position is winning! 90...Qd7+ 91.Kg6 Qd3+ 92.Qf5 Qg3+ 93.Kf7 Qc7+ 94.Kg8 Qb8+ 95.Kg7 Qc7+ 96.Qf7 Qh2 97.h6 Ka5 98.h7 Qe5+ 99.Qf6 Qg3+ 100.Kh6! [Black resigned in view of 100.Kh6 Qh2+ 101.Kg6! and the checks expire due to the correct position of the white king (on a rank adjacent to his black counterpart).] 1–0
Since everything above was given by the Tablebases, and since Dvoreysky has been examining these positions for years, I suspect that only the details of Dvoretsky's analysis were amended with the Tablebase's help, not their essence.
Sometimes the DEM CD varies from the book because there is a better way to describe the Tablebase results that were already used to solve the position. An amazing example is as follows:
M Botvinnik – N Minev, Amsterdam ol 1954
On the CD, Dvoretsky introduces a new take on this position that is not in the book:
'Now we come to those exceptionally complicated cases when the king of the weaker side is placed far away from the pawn. Computers have proved that a win, when it exists, can often be achieved (when both sides play correctly) after more than 50 moves! Practical players should not delve too deeply into this jungle, for these endings occur quite seldom. We shall confine ourselves to basic theoretical statements and the most important practical methods.'
This was obviously added, perhaps in the proof stage, when someone noticed that of the five wins provided by the Tablebases, the fastest was in 54 moves! Continuing on with Dvoretsky's account:
'M. Botvinnik was the first to find the correct method for the stronger side, during an analysis of the following adjourned game: 74.Qf6 [74.Kh6? Qh4+ 75.Kg7 Ka3!= is much weaker, as Botvinnik played in an identical position against G. Ravinsky eight years earlier. One should not place his king in front of his pawn; According to computer analysis 74.Kf5! is more precise; if 74...Qc8+ (this position already occurred on a previous move) then 75.Kf4 ... [jw: in the book this leads to a line with an 'etc.', where Dvoretsky adds that 'this "etc.", by the way, lasts more than 20 moves at least.)', a Table-based comment! But with the Nalimov Tables we see that Black can actually last for 37 more moves. This is all academic, obviously, but an interesting contrast of BELLE and Nalimov.] 74...Qd5+ 75.Qf5 Qd8+ 76.Kh5 The stronger side should place the king on the same file or rank where the defender's king is standing, or an adjacent file or rank (this rule is also valid when more pawns are present on the board).This tactic often enables counter-checks when the queen provides protection from a defender's check by interference. 76...Qe8
77.Qf4+? An error that was left unnoticed by Botvinnik. [The computer analysis shows that the correct winning process is 77.Kg4! Qe2+ 78.Kf4 Qd2+ 79.Ke5 Qb2+ 80.Kd6 Qb8+ 81.Ke7 Qb4+ 82.Kf7 Qb7+ 83.Kf6 Qb6+ 84.Qe6 (this is only an introduction: a lot of precise moves are still required for achieving success).'
So much for the book; here the CD version continues '84...Qf2+ 85.Ke7 Qh4+ 86.Kf8 Qh8+ 87.Kf7 Qh5 88.Qe4+ Ka3 89.Kf6 Qh6 90.Qf3+ Ka2 91.Kf7 Qd2 92.Qa8+ Kb2 93.Qb8+ Ka2 94.Qa7+ Kb1 95.Qb6+ Ka1 96.Qf6+ Ka2 (96...Kb1 97.Qf5+ Kc1 98.g7 Qa2+ 99.Qe6 Qa7+ 100.Kg6 Qg1+ 101.Kh7 Qh2+ 102.Qh6++-) 97.Qe6+ Kb1 98.g7 Qf4+ 99.Kg6 Qg3+ 100.Kh7 Qh2+ 101.Qh6 Qc7 102.Qd2! ...+- BELLE']
'But why is the move actually played wrong? Because, when dealing with a knight or rook pawn, the defender's king is best placed near the corner that is diametrically opposite to the pawn promotion square. In this case, when the stronger side defends his king from checks with a queen interference, a counter-check is less probable. Black could have played 4...Ka3! here and theory says that it is a draw, although it is a long way from a theoretical evaluation to a half-point in the tournament table, because these positions are very difficult to defend.'
This statement that "theory says that it is a draw" is a bit odd. Surely the only definitive answer could be given by computer, accessible to Dvoretsky, and thus it is not surprising that Botvinnik didn't "notice" the error. In fact, the king on a3 still loses in some positions if White's queen is centralized when it returns there, whether or not it heads for the corner or stays on a3. Black would have to know a wide variety of queen positions to avoid going to a3 at the wrong time, so this would be a very shaky rule to depend upon in a real game! It's slightly depressing that after all the advice we have no reliable guide to even this specialized case. Moving on:
'We should add that the indicated drawing zone does not exist in case of a bishop or central pawn. One can only expect that the opponent's play will not be precise (although defender's errors are more probable in these situations) or... that the king manages to reach the area in front of the pawn like in the Botvinnik - Tal endgame. By the way, the drawing zone, near the pawn, is considerably larger in case of a rook pawn, compared with other pawns, because the defender can go for a queen exchange much more often. Having arrived at general considerations about various pawn cases, I add two more remarks:1) The further the pawn is advanced, the less the defender's chance for a draw; 2) The closer the pawn is to the edge of the board; the greater the drawing chances. With central and bishop pawns, practically all positions with a remote king are lost. With a knight pawn, winning positions occur very often. With a rook pawn, a draw can be reached in a majority of positions, although the defense is not simple. 77...Ka5? The wrong way! But this choice was not made purely by chance. The above-mentioned game Botvinnik - Ravinsky was thoroughly annotated by P. Keres, and the Estonian grandmaster erroneously suggested keeping the king on a5 and a4. [77...Ka3!=] 78.Qd2+ Ka4 79.Qd4+ Ka5 80.Kg5 Take notice of White's last moves. The queen is placed best on the central squares (this is usually valid for the defender's queen as well). The closer the queen is to the edge of the board, the winning process is more difficult and the probability of a perpetual check is higher. By the way, now we can easily explain why Botvinnik's 1 Qf6 was less accurate than 1 Kf5!. His queen should not leave the center unless it's an emergency. 80...Qe7+ 81.Kf5! Qf8+ 82.Ke4 Qh6 83.Qe5+ Ka4 84.g7 Finally the pawn succeeds in moving forward, and the climax is near. The finish is also very instructive: White approaches the black king with his monarch in order to create a situation when every check can be met with a counter-check. This method (king-to-king) is characteristic for queen-and-pawn endings. 84...Qh1+ 85.Kd4 Qd1+ 86.Kc5 Qc1+ 87.Kd6 [87.Kd5?! Qc8] 87...Qd2+ [87...Qh6+ 88.Kd5]] 88.Ke6 Qa2+ 89.Qd5 Qe2+ 90.Kd6 Qh2+ 91.Kc5! 1–0
You can see the incredible detail that Dvoretsky goes into and his continual search for truth that leads to new or amended conclusions. This is an ultra-high-quality book. But is most of this particularly valuable for players up to 2200, or even 2400? Sure, a thorough study of each example would definitely improve one's understanding, but that's true for any complex position from any book. Perhaps the most accurate conclusion is that the stronger you are the more you will get from DEM. In my own teaching to average players I am still using Mueller and Lamprecht's Fundamental Chess Endings, which has a wonderful balance between Encyclopaedic coverage (I can find almost anything), examples that can be shortened at most points, and clear explanations that bring together endings of the same sort. To me it provides a simpler method for giving students both information and a sense of why they are proceeding as they are. In either case, it seems to me, a teacher's guidance is preferable, but lacking it I would like to see students learning the fundamentals in a systematic and comprehensible way.
I have included the other two books on my list because they have their own admirable qualities. Yasser Seirawan's Winning Chess Endings is an excellent example of giving the absolute standard basic endings in a systematic fashion. It is well-written and chatty, the key idea being to present 'nothing but the facts'. Seirawan does cover some unique positions (e.g., queen-versus-queen with several pawns) and incorporates a discussion of the effect of computers. Nevertheless, the point of the book is to make a friendly presentation of everything the average player might need and thus provide a good reference book at the same time. An unpretentious and successful idea, and perfect for an instructional text.
Alexander Belyavsky and Adrian Mihalchishin have written a number of books that I haven't liked, but their Modern Endgame Practice is an exception. With a host of great examples from master practice over the years, the authors show endings from the usual categories but also discuss some issues such as the realization of advantages, typical mistakes in various endgames, and mistakes that young players make. In general the book isn't tightly organized and doesn't deliver systematic theoretical instruction. What I really like is their extensive collection of master-level positions that were misplayed and the lessons to be learned from them. These are spread throughout every chapter of the book, with one chapter specifically entitled 'Grandmaster Mistakes in the Endgame' and a short one called 'Shameful Mistakes'! I enjoyed this book, which is made for a browser's pleasure and edification.