Computers in chess... Good or Evil? Part Two.

  • GM Gserper
  • | Feb 17, 2013

In last week's article, we looked at a very interesting phenomenon of computer chess - the "horizon effect." Basically it is a computer's inability in certain situations to see the future no matter how fast its processor is or how many plys it can calculate. In the following game our "human computer" Hikaru Nakamura takes full advantage of this phenomenon. This brilliant game shows why people are still smarter than computers, no matter how powerful the computers chips become. Let's examine this epic battle step by step.

Part 1: Fortress

The position is an obvious draw since neither side can open the position to his advantage.

Part 2 : Nakamura's First Trick

Nakamura sacrifices (or maybe just blunders, since it is just a 3 minute ICC blitz) an exchange.  But the resulting position is still a 'dead' draw because it is still the same fortress:

Part 3 : The devilish trick!

As the position is still a draw, both sides keep moving aimlessly back and forth and the game comes to the inevitable "50 moves" rule. As most of you know, if in the duration of 50 moves no capture was made and no pawn moved, one of the opponent's can claim a draw. The computer could see that the game is about to be drawn, but despite the positive evaluation of the position it couldn't do much as the only way to avoid the draw was to push and sacrifice a pawn which change the evaluation to about even.  So, Nakamura tries to avoid a draw by sacrificing an exchange one move before the "50 moves rule" takes effect!

And now we have another fortress, but the game continues!!

Part 4 : The Bait!

The opponents' aimlessly moves and even the trade of Queens on move 124 doesn't change anything. Another instance of the '50 moves rule' comes into play and on move 174 the computer must make a tough choice. Either it makes just about any move and allows the game to be drawn or it sacrifices a pawn just to avoid the draw.  Unlike the situation on move 60, computer's evaluation of the position is so positive that even after the pawn sacrifice it thinks it is winning.  So it sacrifices a pawn!


Part 5: Massacre

After White loses a very important pawn (really, the anchor of his position), he loses his whole queenside, pawn by pawn. After that, White's clumsy Rooks are no match for Black's minor pieces and the armada of passed pawns. So, the game reaches the next position:

Part 6 : Public Humiliation

The computer again 'forgets' to resign and just like in the game from the previous article, Nakamura humiliates the computer in front of cheering audience:

It is a pity that it's impossible to make a Hollywood movie based on this game. In my opinion this game is as exciting for a chess player as the classical movie "Sting" with Paul Newman and Robert Redford.

to be continued...


  • 4 years ago


    Awesome article! thank you for sharing it! Great game! :)

  • 4 years ago


    The one thing that computer chess has done to hurt me playing in real life is my "angle" of vision.  You're never playing an OTB game looking directly down at two dimensional pieces.  It's always from a very low angle sitting in your seat, and the pieces are vertical.  It hides open lines, and changes your view.  I feel I can "see" the board better in computer chess, and for a very good reason - I CAN!  This makes it harder to play OTB.

  • 4 years ago


    I cant belive on my ipad i quit a game with the computer once and then each time i try to play with now it always glitches up and closes the app!

  • 4 years ago

    NM coolthing


  • 4 years ago


    Wonderful article. I have enjoyed it a lot!

  • 4 years ago


    Date of the game : 15 march 2008. We are in 2013. Way outdated. How about you show us a game between houdini 3 and nakamura of this year? Maybe the "massacre" and the "public humiliation" would change sides. Of course, Nakamura is forced not to resign, and we turn on the "humiliation" feature of houdini.

  • 4 years ago


    Do Computers get mad when their human oponent abandons a game? :)

  • 4 years ago


    we need more information.if deep blue beat kasparov and fritz beat kramnic,then rybka must have naka for fun especially in a blitz game.i am waiting a challenge man vs the machine beetwin top grandmasters magnus,aronian,anand,kramnic etc against top engines deep rybka 4 and houdini 3 64 pro without ''special'' rules for the engines helping thiw way the grandmasters

  • 4 years ago


    Very interesting moves and games!Nakamura won!

  • 4 years ago


    "his brilliant game shows why people are still smarter than computers, no matter how powerful the computers chips become"

    This statement is just plain illogical. The author of this and his previous article clearly does not understand how these chess computers function. It is entirely possible to refine (and I am sure the programmers have already done so) their computers to handle these types of situations.

    Honestly these articles are the most biased, illogical, and ignorant that I have seen on I understand that everyone wants to believe in Human superiority over computers in Chess, but doing it by claiming falsehoods is just embarrasing.

  • 4 years ago


    Another important thing to remember is that engines, no matter how well-designed, can only play the strongest moves based on their algorithms and the computation power of their computer platform.  Thus, if it determines its position as "lost", it can only play the least-losing move, but if its determination is somehow flawed, then so is the move.  Humans, though the chance is remote, could still play a winning move out of desperation that exploits the hole in the programming, if only by sheer accident.  This will always remain true, until the game of chess is "solved", which is still very far from being a reality.  I'm inclinded to agree wih Dan Heismann's assessment that the "perfect" chess game is a draw, and until the strongest human players can play at that level with consistency (which may never happen), the game will continue to be very much alive, in spite of computer analysis.

  • 4 years ago


    Nakamura's claim to have only trained with computers seems to hold up in these "fortress" positions.  He understood (at that time, anyway) that the computation methods of the engine and computer would force it to play a losing move when up material and faced with a drawn position, because the algorithm simply wouldn't "accept" the draw while assigning itself a positive positional analysis as white.  Therefore, his strategy would only be to reach the draw position down in material (minor pieces vs. rooks, in this case), hence why all of his big computer wins are nearly 200 moves or more and involve him outrepeating the machine.  As others have said, I'm sure those algorithms have been very much refined over the last few years.  I know Rybka's engineers in correspondence with Roman Dzindzichashvili have innovated stronger algorithms in positional aspects of the game that aren't strictly calculation-based.  While there are a couple programs that beat Rybka in "brute force" fashion, I'm pretty sure Rybka 4 would not have fallen victim to the so-called "Horizon Effect" and lost this game against Nakamura.

  • 4 years ago


    It was not epic. Boring.

  • 4 years ago


    Regardless of al the other comments, their are some good insights into the major differences between human and computer "thought" processes...

  • 4 years ago


    Interesting topic, considering the setting. How could we be having this discussion without the aid of computers? It will still be available in five or ten years with minimal effort, most likely; anybody remember how hard it was to research a five-year-old newspaper article back in the day?

    Besides enjoying the Tactics Trainer and Game Explorer features of this site, I've had the privilege of playing correspondence chess with opponents in far-off places. The cost is minimal and the turn-around time phenomenal; also the Analysis Board is very convenient. Back 30 years or so, I played correspondence chess using USPS instead. We normally were assigned 3 opponents for rated games, one game as white and one as black with each opponent. I had a cardboard book with six chess boards and cardboard men that I could plug into slots to keep track of positions. If I wanted to set up an analysis board, I was free to do that... again and again if I wanted to look at several lines. We sent postcards, each with two moves, and the answers came back after a week or so. That meant you could finish a short game in six months, but a year was not uncommon. If you wanted to do opening research, you could devote a yard or so of shelf space to your chess library... but only if you had a lot of money to invest in the books. I once visited an opponent at his cottage, where every horizontal surface was covered in chess boards and note-paper.

    Have computers been good for chess? Computers will likely beat most chess players most of the time, but they won't enjoy the game as we do!

  • 4 years ago


    Quite astonishing how a GM can write such a poor shows that a title doesnt mean you have any idea about the current state of computer chess

  • 4 years ago


    271 moves in 3 minutes doesn't seem likely to me. Also, I think today's Deep HIARCS would have a fair chance against Nakamura

  • 4 years ago


    seems to me, the most of guys commenting here are computers :)

  • 4 years ago


    The computer on this website is tough (~1600 level). It appears to remember how I played and know how to move when playing against me. It plays very aggressive and a zone offense and defense. 

  • 4 years ago


    Cool, 271 moves in a 3 minute game. That means 0,66sec per move. It's just impossible, and end with 5 bishops? Sorry, but it's impossible.

Back to Top

Post your reply: