The ambitious “purpose of this book is to give [the reader] enough extra understanding so that [he or she] will eventually be able to defeat [his or her] chess computer no matter on which level it is set.” Keene and Levy do offer some insights into how chess programs work, from which they derive some “computer-hostile” playing strategies successfully employed by each in actual play. Levy in particular has run up a long string of well-publicized and profitable victories against computer opponents. Few people possess the necessary talent and commitment to achieve the purpose of the book, however. Those who do are probably already aware of the suggested strategies.
The general approach is to become aware of the principles underlying the computer’s play, then treat those principles as rigidities to be exploited. For example, chess programs use a “book” to select moves in the opening. If the human opponent uses unorthodox moves in the opening, the computer will be taken out of its book and must rely on its standard evaluation function. Since this function is typically geared to middlegame play, it will often select strategically poor opening moves.
The evaluation function itself can be treated as a rigidity. The trick is to create situations where the function performs poorly. One of its rules may encourage exchanges if they will give the computer’s opponent a doubled pawn, but in some situations a doubled pawn is not a disadvantage. So inviting the computer to give up a powerful, fianchettoed bishop for a knight in such situations is one way to gain an advantage.
Within its search range, the computer makes no tactical errors. So one should avoid tactical complexity unless one is sure the point of one’s combination is beyond the computer’s range. A better middlegame approach is to make quiet, positional moves whose long-term effects are not apparent to the computer since they are not easily quantified, and thus are not part of its evaluation function. In the endgame, the superior human ability to discern the long-term consequences of given moves becomes even more important.
Many game scores are included. Some illustrate the authors’ computer-hostile strategies. Sometimes the machine wins, particularly when the human plays into the computer’s strength. To be appreciated, these games must be closely followed and analyzed, move by move.
This volume is slightly dated, so some specific suggestions will not work against the best current programs. But they will probably work against the cheaper standalone modules and microcomputer programs if the human player is skillful enough to use them effectively. Furthermore, the overarching principle remains sound: however intelligent the program’s evaluation function and however powerful the hardware, the system can be treated as a rigidity that is at least theoretically surmountable by human adaptability and imagination.