Smoke and Mirrors
This is a reprint of an article I wrote for another site
- "Larry Kaufman has made special preparations with the help of Rybka author Vasik Rajlich. They have created an “anti-human” version of Rybka as well as a special handicap opening book. Both handicap book and “anti human” version remain secret... "
-from the Rybka-Ehlvest match page
I feel that computers have finally become strong enough that humans should stop giving them odds.
On March 6 through 8, 2007, GM Jaan Ehlvest played a series of eight odds games against the chess program Rybka. The eight games correspond with the eight pawns. Each game differed as Rybka, playing white, offered a different pawn in each game.
The result was +4-1=3 in favor of the computer.
From the results alone, one would readily conclude that Rybka's odds-giving was justified and that computer chess programs have reach a level beyond that of highly rated grandmasters (Ehlvest was rated at 2610).
It all smoke and mirrors, my friends, and has less to do with chess than it does with hype.
First, let me direct you to Hans Ree's article about this particular match.
While Ree's article, "The Advantage of Giving Odds," gives a cynical and rational perspective on this particular match, it fails to go both far enough or broadly enough.
Human-Computer matches are organized by folks with vested or personal interests in computers/programs, and, believe me, they are after results.
Once upon a time Chess was about ideas and a decent game was a stuggle between those conflicting ideas. Even annotations were expressed in words with lines only given to show perhaps a subtle tactic that supported a move. The ideas conveyed in the games of the old Masters were deeply analyzed which propelled the advancement of theory off the board. Chess, while indeed a search for truth, was still mostly a fight between persons of limitations and flaws. This is what made the game exciting. The elusive ultimate truth could be hunted down later in a more pristine environment.
Today, opening theory has evolved, from the weight of years and, yes, from the use of computers, to a deadening, even stultifying, level. Tedious and impossibly complex endgames have been solved, or nearly so, for anyone with the ability or the time to memorize it all and produce it on demand with perfect recall. Simply put, the textbook of Chess has outgrown human capacity. This surfeit of theory, in itself, has caused some GMs, Fischer for instance, to look elsewhere - and this relates to human vs. human chess.
Now we'll factor in Chess programs. Chess engines employ brute strength. They calculate, pure and simple. I read a statement by Vasik Rajlich, the author of Rybka, in which he claims that he doesn't like brute force and is trying, and succeeding in his opinion, in making Rybka a strong evaluator and an example of artificial intelligence. I'm not sure if he really believes that or if he just wants everyone else to, but Rybka is brute force, highly dependent on fast, multi-processors for optimum results. There's no intelligence, just number crunching and clever algorithms. And, as this is what computers do best, there's nothing wrong with it. But to imply that Rybka is beating humans at their own game is an affront .... smoke and mirrors.
Logically, it would follow that, if brute force is stronger than ideas, then brute force should, by itself, win on the virtue of its abilitly. Yet, this never happens in a human-computer game. As we noted earlier, the weight of theory, especially in openings is too much for humans to bear. But opening theory is a cinch for computers and available in immediate and flawless precision - not because of brute force, but because of stored information. As the opening progresses, only one side has the opportunity to make an inferior move - guess which side!
That's worth Pawn and move in itself.
Guaranteed to go into the middlegame secure, and most likely better, (but not because of brute force - remember?), the computer only needs to make it to the endgame, equal or slightly less than equal, where known techniques gives it the edge since mistakes are far less likely
- Pawn and two.
That leaves the middlegame. This is where brute force come into play. From here, until the endgame, the computer program is entirely on its own. Chess engines are particularly good with combinational play and almost never miss a tactic. Regardless of their legendary tactical skill, computers seldom decide a game by tactics alone, particularly in an even middlegame. Going into the middlegame better counts for more and by virtue of the advantage, the computer may have enough tactical edge at this point to decide a game.
People are human. Into every game they bring their human problems. Seldom does anyone play without some human frailty raising it's ugly head. Whether one is distracted, has a headache or a cold, whether one has some personal problem tugging at the back of the mind or some silly jingle that keeps running through one's head, whether one has a sudden fear or loss of confidence or a misplaced euphoric surge of over-confidence, this all affects one's play, and seldom for the better. Computers aren't human and every game is the same quality as the last one, barring some glitch. The human factor might push it up to Knight odds.
I think the time has come to turn off the opening books and eliminate the engame databases. No more searching for positions from past games.
Computers are now strong enough they don't need Knight odds.
And we don't need the smoke and mirrors.