The Chess Mind

Author: Dennis Monokroussos.
This is a blog for chess fans by a chess fan who is more than a chess fan - other topics do creep in from time to time, per my interest.
All material here is copyrighted, and may not be reproduced without my prior permission.

Wednesday, November 23, 2005

Man vs. Machine: Final Bilbao Update
Three draws in the last round of the second People vs. Computers World Team Match let the humans leave with some dignity; the final score was 8-4 in favor of the computers. Granted, it's not the most flattering result, but it's half a point better than last year's performance, and much better than Michael Adams' dreadful shellacking against Hydra earlier this year. Best of all, I say, it was done playing relatively normal chess - no 1.d3 or Stonewall nonsense. We are the underdogs, but we can still compete.
Events Update: The World Junior Championships and the Man-Machine Matches
Starting with the World Junior Championships, the final result of the Open ("Boys'") section was long ago decided, but only now official: Shakhriyar Mamedyarov of Azerbaijan regained the title he won in 2003. After blasting out with 8.5/9, he coasted home with four draws (three of which were merely pro forma) and a final tally of 10.5/13. Finishing in second was Hungarian Ferenc Berkes (9.5/13), while Evgeny Alekseev of Russia beat out Vugar Gashimov of Azerbaijan on tiebreak to take third (both had 9/13).

In the Girls' section, a late run propelled German IM Elisabeth Paehtz into first place with 10/13, half a point ahead of Gu Xiaobing of China and a full point ahead of Beata Kadziolka (presumably the third-place winner on tiebreak), Turkan Mamedjarova and Dronavalli Harika of Poland, Azerbaijan and India, respectively.

Congratulations to the winners, and congratulations too to the computers - or rather, their programmers - as they reasserted their dominance over carbon chess players in round 3 of the II People vs. Computers World Chess Team Match in Bilbao. Kasimjanov held a draw, with Black, against Fritz 9 (poor Fritz, again!), but Junior slowly ground down Khalifman and Hydra applied a fearful beating to Ponomariov, which you can replay here.

That puts the computers up 6.5-2.5 going into today's last round, and as of this writing, all three games were still going with the humans still kicking.

Related Posts (on one page):

  1. Man vs. Machine: Final Bilbao Update
  2. Events Update: The World Junior Championships and the Man-Machine Matches
  3. Humans, Computers, and the Horizon Effect

Tuesday, November 22, 2005

Humans, Computers, and the Horizon Effect
About five years ago, before the situation had reached its current, near-hopeless state, Kramnik told an interviewer that computers were weak in tactics.

What?!

The explanation had to do with depth and assessment: while computers see everything to a certain depth, and won't miss a move because it looks funny by human standards, humans have advantages too. While computers spend a great deal of time looking "sideways" - that is, looking at all or many legal moves at each ply, humans can pare off the junk, delve deeply, and often do a better job not only of assessing both what ought to be examined but what's really going on at the end of the variations, too.

These advantages are diminishing every year, as more powerful software/hardware allows deeper searches and better assessments. Diminishing, yes, but not gone altogether! Take a look at this position from round 2 of the ex-world champs vs. computers match in Bilbao:



Ponomariov (White) is in trouble, and with good technique Fritz (with Black and on the move) should win. A good way to start is with 39...Qxg3+ 40.Kxg3 f5, but instead Fritz chose the horrible 39...Bc2??. It's a great move against all lines but one, but that one line leads to a win for White. Because the human is down even more material most of the way, and the payoff occurs at a relatively deep search ply, Fritz doesn't work it out. (Nor is Fritz (9) alone; I tried it with Shredder 9, Junior 9 and Hiarcs 9 [reminiscent of a Beatles "song", isn't it?], and they all flopped, too.)

The variation is fairly easy for a strong player to calculate and find, as White forces Black's hand just about every step of the way. For the computer, however, it's an instance of the so-called "horizon effect": it can calculate only so far - to the "horizon" - and then just past that comes the disaster. In a sense, we're all susceptible to it, but sometimes computers will stop calculating in positions humans know require further thought. That may not last for long, but while it does, enterprising humans will be there to pick up the points.

Curious about the variation? See this story on ChessBase for the moves, more of the story, and a photo report on round 2 - won by the humans by a 2-1 score!