In 2013 I wrote The Kaufman Repertoire for Black and White. The Black half of this book is an update of the Black half of that one, mostly updated in 2018 with some further updates in 2019, with the addition of the Marshall Attack being the biggest change. The White half however is completely new; it had to be, because in KRBW I recommended 1.d4, whereas here I recommend 1.e4. Many of the games, and most of the analysis, for the White book are from 2019.
The main theme of the book, especially the White portion, is that you can obtain good positions, meaning slightly favorable ones as White and only slightly worse ones as Black, without having to play the most complex, theory-heavy lines in most cases. I was pleasantly surprised to see how little White gives up by avoiding the most critical lines. Apparently the elite GMs agree with me as they have been playing many of my White sidelines against each other in 2019. White can usually maintain a plus even in these sidelines well into the endgame. It won’t be enough to win many correspondence games when your opponent is using an engine, but for over-the-board play, you will generally emerge from the opening as White with good winning chances if you are the stronger player, and with black should at least obtain positions where you won’t just be fighting a one-sided battle for a draw. Although this is a repertoire book, I have made a much greater effort than in my previous books to give alternatives for the chosen side, especially in the White book, as I really don’t want the book to become obsolete just because one or two variations prove to be dead draws or otherwise dubious. The price for this is less coverage of rare moves by the opposing side. Usually inferior moves by Black can be rather easily refuted with any modern engine. I can’t cover everything!
Although I am the oldest active GM in the U.S. and no longer play near GM level, I do have some real advantages for writing a book like this. Computer chess is having a revolution now, based on Monte-Carlo Tree Search and Neural Networks, inspired by the success of AlphaZero. This is not a book on computers, so you’ll have to look elsewhere for more information about these terms, but suffice it to say that I am very much involved with these developments as a partner in KomodoChess, which has a very strong Monte-Carlo version already, and so I know what engines to use, what hardware to buy, and how to use them effectively. In March of 2019 I purchased a computer with a very powerful GPU (RTX 2080 for the tech-minded) and 8 very fast CPU cores. My method for working on this book is to run each position on the latest Lc0 (which is a neural network designed to roughly replicate Alpha Zero, rather successfully I would say) on my GPU and 2 CPU cores while running Komodo 13 MCTS on the other 6 CPU cores. These two engines complement each other quite well. Lc0 is in general stronger due to the extremely powerful GPU (which has almost 3000 cores!!). But it has no chess knowledge except what it taught itself by playing games, whereas Komodo MCTS has ten years of refinement of its chess knowledge behind it. Also Lc0 is relatively weaker in the endgame, and rather blind to perpetual checks in many positions. Note that both of these engines can be used in ‘MultiPV’ mode to display the top 5 (or more) moves without any loss of quality, which is not at all true of normal (non Monte-Carlo) engines, nor do the two engines have to share resources.
The result is a quality of analysis that vastly exceeds what most people will get using normal engines on normal pcs with shared resources and MultiPV displays. Aside from using these two engines, I also keep an eye on analysis done by others using Stockfish, Houdini, and normal Komodo, as well as database statistics using two databases. One is the Hiarcs Powerbook (mostly engine vs engine games I believe), and the other is a combination of the ChessBase MegaBase and a database of correspondence games. Of course I also consult books and magazines (especially New In Chess Yearbook), but due to the amazing developments mentioned above I consider anything older than 2018 to be unreliable so looking at older books was not a priority. One book I did consult on several lines for White was Keep It Simple by Christof Sielecki, both because it is new enough (2018) and because we chose some of the same lines, since simplicity was also one of my goals in this book. But I wanted my book to be suitable even for grandmasters, so in general my choices are not as simple as his; I’m really trying to prove an edge for White, not just interesting lines with surprise value. Although the variations chosen are aimed at reasonably strong players, my explanations are at a more elementary level, so even if some of the lines are a bit difficult, I hope that less advanced players will learn how to evaluate positions from my comments.
My role in choosing which moves to give is primarily as a referee. When the two engines (plus other analysis and database stats when applicable) agree, I will very rarely argue. These engines play somewhere in the 3400 to 3600 Elo range, and only in special circumstances would I ignore them. But when they disagree, which is pretty often, I have to decide which one is right, and here my chess understanding and knowledge of chess engines both play a role. The default assumption is that Lc0 is right, but if Komodo MCTS strongly prefers a move that is only slightly below the best according to Lc0, or if Lc0 seems to be blind to some feature of the position or to a perpetual check, I’ll probably go with Komodo’s choice. I also consider whether the move is easy or difficult to understand; it is common that Komodo will pick the same move that I would pick, while Lc0 prefers one that just doesn’t seem as good. Lc0 may be correct, but if I can’t figure out why, probably the reader will also have difficulty, so I do consider this factor.
I generally quote the evaluation shown by Komodo (example: (+0.26) – between brackets) because until recently the Lc0 evaluations +0.27 were unrealistic, and I try to put into words the factors that justify the assessment shown. I tend to use symbols showing advantages a bit more aggressively than is customary, because if both engines show around +0.15 (for example), the position is almost surely favorable for White, if only slightly, and calling it equal just seems wrong.
The book is full of novelties, which I mark with an N, although it often happens that someone plays one of these moves shortly after I wrote up the game, so don’t be surprised if you see games with my ‘novelty’. When I refer to material advantages, I use the scale that I have promoted (see the chapter called ‘Material values’, namely pawn = 1, knight or bishop = 3½ (with a slight preference for the bishop in general), rook = 5¼, queen = 10, and the bishop pair earns a half point bonus. Checkmate may be the nominal object of the game, but nowadays it seems as if the rules have been changed to say that being the sole possessor of the bishop pair wins! It’s an exaggeration, but if you don’t believe that winning the bishop pair for nothing is generally a serious advantage, a lot of the White book won’t make much sense. All modern computers and top GMs accept this.
Writing this book has made me feel like a time traveler. It is full of the latest games and novelties, many in 2019, and features some games by players born in the 21st century. Yet I also have ties to the distant past. My first chess teacher, Harold Phillips, was Greater New York champion in 1895, and played twice against the first World Champion, Wilhelm Steinitz, in 1894!! I met Edward Lasker, whose most famous game was played in 1912, and had some instruction from Norman Whitaker, a top player around 1920. I played against Sam Reshevsky and Al Horowitz in the U.S. Championship, and won a ten game rapid match from Arnold Denker, three of the four top American players during World War II. My first big success was winning the American Open championship in 1966, but I didn’t earn the Grandmaster title until I won the World Senior Championship in 2008. I was part of the team that created MacHack, the first chess computer to earn a rating in human tournaments, in 1967, and 52 years later I’m still working on chess computers and playing in tournaments! In short, I have had a very long chess career!
I would like to thank Daniel Clancy for the correspondence database, Hiarcs for their database, Mark Lefler and the late Don Dailey for their roles in KomodoChess, the late Steve Brandwein for teaching me a lot about chess so long ago, Christopher Gallardo for encouraging me to write this book, and New In Chess for publishing it.
Big changes are happening in the chess world, in an effort to combat excessive draws and to minimize the role of preparation for specific games. One top event introduced ‘Armageddon’ playoff games after every draw, and FIDE is organizing a serious World Championship of ‘Chess960’ aka ‘Fischerandom’ with most of the elite players. I don’t know where these changes will take us, but I hope to be involved in these new developments. Despite my age, I’m receptive to new ideas.