Chess Reimagined

Dylan Wu

In the digital age, the advent of strong chess engines and databases containing millions of computer games and volumes of opening theory is making top-level chess more achievable; however, this rise in playing level is not without a decrease in creative play and decisive games.

Perhaps the most recent example of this phenomenon is the 2018 Carlsen-Caruana World Chess Championship Match that contained near-perfect play—according to computer standards—but contained almost all uneventful draws and no decisive matches.

This more methodical approach by even the most advanced players is quite understandable, as playing it safe according to well-known lines usually leads to more positive outcomes than risky “creative” variations. In spite of this, such sterile lines of play from both sides inevitably leads to drawish outcomes, and many chess amateurs and masters agree that the game must somehow be reinvigorated.

As such, Vladimir Kramnik, Russian Grandmaster (the most prestigious chess title) and former World Champion, met with Dennis Hassabis, founder and CEO of artificial intelligence lab Deepmind’s Alpha-Zero, perhaps the strongest chess engine that is unique because of its machine-learning capabilities, to discuss the issue. Specifically, Mr. Kramnik wanted to explore and experiment with a variant he believes will restore chess’s creativity from the days before computers.

One variant, Fisher Random or Chess960, has become quite a popular solution for many looking for more exciting games. In Fisher Random, the pieces on the outermost ranks are rearranged at random for every game while the pawns remain in the 2nd and 7th ranks; white’s position is still symmetric to black’s position. The idea behind Chess960 is that opening theory is completely eliminated, forcing players to think of more original moves.

However, Mr. Kramnik does not view Chess960 as the best alternative, as the format is “less aesthetically pleasing than traditional chess and the unfamiliar positions arising from the different setup are less enjoyable to world-class and amateur players, according to Mr. Kramnik. Instead, he proposed a no-castling variant that has a traditional setup but does not allow either side to castle.

To test the game balance of no-castling chess, Alphazero mastered the variant through trial and error—like it did with traditional chess—and played millions of games against itself. The win-loss percentages for black and white were similar to that of traditional chess (~50%) so both sides had an equal opportunity of success.

Mr. Kramnik emphasizes that no-castling chess does not allow the king to reach a safehaven on the board, creating more entertaining and dynamic play encompassing all pieces. He states that there will be more original patterns and less memorized ones, less forced draws, and a more level playing field with less known opening theory (much is based around castling). Mr. Kramnik anticipates an increase in decisive games, unexpected and new ideas. Although there are many details to be deciphered—more complex positions, new opening theory and king safety strategies—he is confident that the no-castling variant will restore the energetic spark to chess.