Is This The Future Of Chess?
GM Emil Sutovsky recently posted some shocking news on his Facebook page: a computer program, Ponanza, beat the best shogi player in the world. Not only that, but the computer program pretty much introduced a novelty on the very first move: no professional shogi player ever played this move before. I assume it is like opening the game of chess by playing 1.Na3 or 1.f3.
Ponanza vs the shogi professional Takayuki Yamasaki. Photo: Shinji Fukamatsu.
Of course, this is not shocking news for chess players. Unfortunately, we passed this stage of computer development a long time ago.
Today, a top engine can play literally any first move and handily beat Magnus Carlsen. The future of the chess competition between humans and computers is pretty clear: with every day passing the gap between us is going to get bigger.
The real question is how the inevitable progress of chess engines will affect human play.
As Yogi Berra correctly pointed out, it is tough to make predictions, especially about the future. Indeed, no one knows how tournament chess will look 10 years from now. I want to discuss only the obvious trends we can witness today.
The defining moment in the chess competition of human vs. computer happened in 1997 when the program Deep Blue beat the world champion Garry Kasparov. Then even the most optimistic proponents of the theory of human superiority realized that in this kind of competition, we are doomed. So, instead of the "humans are smarter than computers" mantra, we comforted each other with "yes, computers play stronger, but they don't understand chess, they just calculate better!"
Indeed, we had numerous games as proof of this statement. Take for example the three moves that shook the chess world in the very first game of the Kasparov-Deep Blue match:
Three random moves not linked by any plan and making Black's position more vulnerable looked like beginner's play. And yet they were played by some entity that managed to beat the world champion!
Fast forward to our day. Computer play has become more human-like (or maybe we started playing more like computers ). These days the idea of humans' superior understanding of chess has quietly disappeared. To tell you the truth, sometimes I feel that it is actually computers who have the superior understanding of the position.
Case in point is a game from the recent super-tournament:
When I watched the game, I expected the players to agree to a draw any minute, especially when a pair of rooks was swapped and the game had reached the following position:
Now look how the game ended:
While many human players would indeed just agree to a draw when the queens were traded, the computer maintains that White had at least a half-pawn advantage throughout the whole endgame! Now tell me, who understands chess better?
If we don't talk about the problem of computer cheaters, then by far the biggest influence of computers on modern chess can be seen in the openings. The abundance of the Berlin Defense in top events is not a coincidence. It is easier to analyze the positions with fewer pieces on the board and therefore, the Berlin endgame became the weapon of choice for most of the top GMs.
Chess players are digging deeper and deeper with the help of their silicon friends, essentially trying to solve the game of chess. Some of them come very close! In his recent interview, GM Boris Avrukh mentioned one episode of his work with the world #2 GM Wesley So. Wesley showed an opening line for Black that in 50 moves would lead by force to a rare endgame two knights vs. rook and bishop. When Avrukh asked if Wesley knew how to draw this endgame, Wesley immediately showed him how to build an impregnable fortress!
While this episode is extremely impressive, it is also very sad.
At some point something will have to be done to return the human element into opening play. I don't think Fischer random chess (Chess960) is the solution, as the starting positions are too far from classical chess. Just imagine: we teach our kids to play chess by showing them beautiful examples of Morphy, Fischer or Kasparov's games and then abruptly say: sorry kids, now you play Fischer random chess and you are on your own in this mess!
Computers can actually help us to solve this problem. If you ask a computer engine to evaluate the initial position, it will give you about a quarter of pawn of an advantage for White. But if you force White to play 1.h3, then the evaluation drops to almost dead even! Something like 1.a3 e6 also leads to an absolutely even position according to an engine. So, why should we switch to Chess960, when we can find a dozen of starting positions in regular chess that the computer evaluates dead even and that have no opening theory at all?
Our current world champion shows how to play classical chess with no opening theory:
Another trend I anticipate is a dramatic change in the way the opening monographs are written. Currently all the opening books base their evaluations on human games. Something like, "this position happened in the game GM X vs. GM Y, London, 2015 where White had a better position."
But think about it: the difference between the human world champion, Magnus Carlsen, and the top engines is about 500 rating points!
To give you perspective, if you are an 1800 player, you wouldn't really be interested to follow the games of 1300 players. Then why do opening monographs practically ignore the games of top chess engines and concentrate mostly on the games of weaker players? In 10 years or so, the opening monographs will completely switch to computer games and analysis, in my opinion. They will probably mention human games only to demonstrate typical mistakes to avoid.
Finally, I am expecting a major change in opening trends. Remember, the modern Berlin craziness started after Vladimir Kramnik beat Garry Kasparov in their world championship match in London, 2000. There the Berlin Wall was Kramnik's surprising weapon that helped him to win the world title. Kramnik deserves the highest praise as he managed to re-evaluate the Berlin.
It is just a matter of time till some top GM listens to his chess engine and discovers that, say 3...Nd4 in the Ruy Lopez is actually as good as 3...Nf6, but leads to more lively positions. So, he follows his engine's recommendation and beats Magnus Carlsen, which will consequently make people crazy about 3...Nd4.
Yes, I completely made up the 3...Nd4 variation being the next chess craze in the Ruy Lopez, it can be any move really!
However, chess engines give White approximately the same advantage after 3...Nd4 as the main lines of the Berlin, so why not...especially since we already know that computers understand chess better than humans!