The Chess Improver: The double-edged nature of computer-based preparation
For The Chess Improver, I wrote about “the double-edge natures of computer-based preparation”.
(The original Chess Improver site is no longer active; the rescued article follows below.)
When I returned to playing chess in 2005, after an absence of twenty years, I was surprised not only by the digital clocks and new time controls (“sudden death”), but also by the availability and strength of computer programs for personal computers. I soon decided to make use of chess engines as part of my return to serious tournament chess, as a way to explore different kinds of positions.
In 2006, I played a game in which I unleashed a computer-generated theoretical novelty in the opening on move six, a Pawn sacrifice. It shocked my opponent. I ended up winning the game, but not without being surprised myself, as I ended up in a position that looked equal to me. I had mixed feelings about the whole experience. Furthermore, five years later, in 2011, I saw an article on Chess Cafe analyzing my very “novelty” as having been presented in Chess Informant 110. I was a little thrilled to see some in-depth analysis of the entire line, but deflated to realize that I had only ever touched the surface of the possibilities in the line.
Tiger’s discovery
At some point, I was looking for ideas for White against the Chigorin Defense, since I noticed that a particular local chess player always played the Chigorin as Black and I needed something against it. I turned on my chess engine Tiger, which was running off my laptop. You have to remember that seven years ago, in 2006, computers were much slower than they are now, and chess engines, although strong, were nowhere as strong as they are now.
While puttering around, I noticed that Tiger recommended in an important variation at move six a Pawn sacrifice by White! (It turns out that Tiger was well known to have an aggressive bias in its evaluation function.) I looked up all the books I could find on the Chigorin Defense and not a single one of them even mentioned this sacrificial continuation. I was intrigued, looked at a sample line, and concluded that it was worth playing: White gets a clear initiative for the sacrificed Pawn.
Unleashing the theoretical novelty but then being surprised myself
At some point, I finally faced the Chigorin Defense player in a tournament game: it was our very first game against each other. I remembered the sacrifice and unleashed it. My opponent was clearly shocked and spent a huge amount of time thinking for each move. To my surprise, he came up with a fantastic defense that I never even looked at during my superficial computer preparation: he counter-sacrificed an exchange in return for attacking my King! It was my turn to go into deep thought. I saw no alternative but to play into the counter-sacrifice. We reached a position in which I feared that there might be a perpetual check.
I kicked myself for not having seen and analyzed this entire possibility. In 2006, chess engines were still slow, and I only used them for interactive exploration, not for overnight analyses or the like, so it was quite possible that Tiger would have seen the exchange sacrifice resource if I had let it run for a while.
Anyway, my opponent made some errors, I maintained an advantage, and then I started letting it slip during a Queenless middlegame, but then he blundered and immediately resigned.
What was the truth of the gambit?
Five years later, seeing the 2011 article covering the gambit, I was pleased to find that in fact, there was a way for White to decline the exchange sacrifice while maintaining a strong initiative. The article had many complex lines demonstrating a White advantage. As a matter of intuition, indeed, one would expect that White has compensation in any case, because of Black’s weak King, but it’s definitely fascinating to see computer-verified concrete justification.
Generation vs. verification
Generation of interesting ideas can come both from humans and from computer engines. My experience with this Chigorin Defense line illustrates this. Tiger generated an idea. I followed up by looking at what it was trying to do. But I should have generated counter-ideas also, and did not. I believe that humans will for a long time have the edge in generating possible ideas, while using computers for verification or falsification of them. For example, even though Tiger found one sacrifice, it did not find the rejection of Black’s counter-sacrifice. That required human exploration of possibilities and exploring long lines to prove White’s attack would come to something fruitful in the long run.
I have not checked with today’s engines and higher compute power whether a modern engine could now completely discover the whole idea without human guidance. It may or may not be the case. Meanwhile, chess innovation still continues. Chess is not dead; it has been transformed, with computers as helpers for humans.