Stuff Happens: Another Exploration into the Realm of What Computers (Allegedly) Can't Do
The sponsors of the Mercosur Cup in Argentina have decided to allow my favorite ChessBase engine, Shredder, to play in their little shindig, and in three of the four games, the computer has done about what one would expect and scored 2.5/3.
But it's the other game that's noteworthy. Here's the position after White's 19th move (19.Bxb7):
The position is pretty dull, and after the obvious, normal, forced, etc. move 19...Qxb7, we can expect 20.Nf3 Rfd8, when whatever battle there is will revolve around White's attempt to bury a knight on c6 against Black's attempts to utilize the d-file.
Now, humans are capable of all sorts of idiocies over the chessboard: leaving pieces en prise, hanging mate, you name it. Sometimes players are sick, drunk, depressed, sleep deprived, exhausted or suffering in some other way that leaves them incapable of a reliable performance. But not computers: as long as the power's running, they do what they're supposed to do.
I guess.
Now see what Shredder actually did:
19...Rfd8?????
What?! I've run Shredder to depth 21 on my computer, to see if there's some bizarre horizon-effect hallucination it's suffering from, but my version resolutely sticks to the obvious 19...Qxb7 and deems 19...Rfd8 horrible, leaving White a more than 4 point plus.
The only semi-rational explanation I can come up with, other than a programming bug (assuming this is a tweaked version at least a little different from the last commercially available version of Shredder), is that at some search depth or other the program thought it could triple on the d-file, win the Nd2, and penetrate with the major pieces to White's second rank. The problem is that refuting that plan is very simple and doesn't require a deep search ply at all. In the game, Lafuente continued 20.Bc6 Ng4 21.Bf3 Qh4 22.h3 Nh6 23.Nb3 and won with consummate ease.
So I wonder what's going on. Are the programmers trying to pass the Turing test by throwing in a random blunder? If anyone learns the true explanation, please send me a comment or an email.
Meanwhile, lest anyone get the wrong idea here, let me reiterate my fondness for Shredder 9: it's an excellent engine. I've used (and still do, on occasion) Fritz, Junior, Hiarcs and Chess Tiger, but Shredder is my hands-down favorite of the bunch, and I wholeheartedly recommend it to anyone interested in buying a program for analytical work.
But it's the other game that's noteworthy. Here's the position after White's 19th move (19.Bxb7):
The position is pretty dull, and after the obvious, normal, forced, etc. move 19...Qxb7, we can expect 20.Nf3 Rfd8, when whatever battle there is will revolve around White's attempt to bury a knight on c6 against Black's attempts to utilize the d-file.
Now, humans are capable of all sorts of idiocies over the chessboard: leaving pieces en prise, hanging mate, you name it. Sometimes players are sick, drunk, depressed, sleep deprived, exhausted or suffering in some other way that leaves them incapable of a reliable performance. But not computers: as long as the power's running, they do what they're supposed to do.
I guess.
Now see what Shredder actually did:
19...Rfd8?????
What?! I've run Shredder to depth 21 on my computer, to see if there's some bizarre horizon-effect hallucination it's suffering from, but my version resolutely sticks to the obvious 19...Qxb7 and deems 19...Rfd8 horrible, leaving White a more than 4 point plus.
The only semi-rational explanation I can come up with, other than a programming bug (assuming this is a tweaked version at least a little different from the last commercially available version of Shredder), is that at some search depth or other the program thought it could triple on the d-file, win the Nd2, and penetrate with the major pieces to White's second rank. The problem is that refuting that plan is very simple and doesn't require a deep search ply at all. In the game, Lafuente continued 20.Bc6 Ng4 21.Bf3 Qh4 22.h3 Nh6 23.Nb3 and won with consummate ease.
So I wonder what's going on. Are the programmers trying to pass the Turing test by throwing in a random blunder? If anyone learns the true explanation, please send me a comment or an email.
Meanwhile, lest anyone get the wrong idea here, let me reiterate my fondness for Shredder 9: it's an excellent engine. I've used (and still do, on occasion) Fritz, Junior, Hiarcs and Chess Tiger, but Shredder is my hands-down favorite of the bunch, and I wholeheartedly recommend it to anyone interested in buying a program for analytical work.