Overtom's weblog

BRAIN VERSUS BRAWN  (15 october 2004)

"Oh, you are a grandmaster! You must be able to think many moves ahead," a lady admirer exclaimed when she was introduced to chess grandmaster Richard Réti (1889-1929).

"Yes, madam, as far as two moves, but only occasionally," the grandmaster seems to have replied, explaining that the number of possible moves in any given position is so staggeringly large that most often it isn’t worthwhile to calculate further ahead.

   Richard Réti

Réti suggested that calculating moves ahead is not the most important thing in chess. A really good chessplayer is characterized by his (or her) insight into the position.

That may be so, but would this also apply to chess computers? Computers don't really have any insight ... or do they?

An interesting way of testing this question would be to compare two chess programs:

  • one that we allow to think ahead a few moves, but which has not much insight into the position.
  • a program that is is good at evaluating the position, but which we allow to think no more than one move ahead.
A program that gives the impression of playing a positional kind of game is Fritz. Unfortunately, I don't have a more recent copy than Fritz 4, which is probably not the strongest version available.

    

For the 'brute force', let's take a program that contains so little memory that it cannot be expected to have much insight into position, the Tandy 1750L.

We'll give this computer 10 seconds per move, which is quite generous compared to microscopic amount of time that Fritz gets.

Fritz has enough insight into the game to see how valuable the black d-pawn is. Unfortunately it cannot look ahead and consequently fails to see that white can force a draw by threefold repetition

Tandy 1750L (10 sec/move) - Fritz 4 (1 ply)

That was hard luck for Fritz! Let's see if Fritz playing with black can cash in on its superior insight into the position:

As black (Tandy) can look ahead further, it has discovered it can win a rook for a knight. But it won't be enough to win the game. On the contrary: again, the Tandy computer fails to see the importance of passed pawns. And this time, there is no draw available to come to his rescue.

Fritz 4 (1 ply) - Tandy 1750L (10 sec/move)

If you had been present during this game, you could have seen that Fritz hardly took any time; all its moves were produced seemingly without any thinking at all, whereas Tandy took ten seconds for each move, which is hundreds of times as much as Fritz had.

Now we can understand why a novice at chess sold his chess computer at eBay (click ). A chess computer that is good positional player - even at its lowest level - is no match for a beginner .

top of page

statistics by pcintelligence.nl

bottom of page