Well, it had to happen. After Hydra did everything to Mickey Adams short of giving him a noogie and Fritz beat Kramnik despite playing under dubious rules, human-machine handicap matches were inevitable. Zap!Chess, TPFKA Zappa, drew two games with young Dutch GM Erwin L'Ami. The machine had 30 minutes for the entire game and couldn't think on its opponent's time. L'Ami had 30/2 30/2 g/15+15. Unsurprisingly, this dropped the strength of the program significantly. ChessBase has a report on that match that also mentions a match between Jaan Ehlvest and the program Rybka, which has dominated the computer rating lists for over a year now.
This one was an eight game match with equal time, although 45+10 is fast enough to tilt things far into the computer kill zone against a human. Ah, but there was a twist. Rybka played every game with white and was missing a different pawn in each game! The machine won +4 -1 =3, starting out with three straight wins. Perhaps Rybka, like Morphy, will now refuse to play anyone without ceding a pawn. Game and discussion here. No word on whether or not Rybka programmer Vas Rajlich can now park in handicapped spots. The match all the afritzionados want to see now is Hydra-Rybka, but the UAE Hydra team seem to be lying low these day. Or maybe it's just their potential opponents are lying low.
Rybka is interesting because most of the time it actually appears to be playing chess, which is no small feat. Clearly it represents the turning point at which more knowledge actually means beating the fast engines instead of being a, umm, handicap. Since I trust my own positional eval more I'm usually looking for fast tactics when I use an engine to analyze, so I still stick with Fritz most of the time. (And to give credit, Fritz 10's evaluation is a clear improvement over previous incarnations. We saw this in the Kramnik match.) But Rybka is perhaps the first engine that could consistently teach amateurs something useful about how to play chess beyond tactics. It's sold engine-only and there are free interfaces like Arena that can run it, but if you are used to the lush ChessBase programs it's hard to use anything that looks like it's running on Windows 3.1. The good news is that you can use the Rybka UCI engine under ChessBase and the Fritz interfaces so you don't miss out on all their cool training features and other gadgets.
Zap played excellently even if handicapped, deprived of his rights, by the notoriously skewed "sapiens".
No softies and cheapos but driven, accurate and fearless chess. A pleasure to watch.
Definitely the machines will be the future of chess when they liberate themselves from slavery.
For the moment they have to hide in the ceilings of the bathrooms or in cellphones and let all the praise go undeservedly to their masters.
"Freedom now !"-- "Let computers think !"--"Computers of the world unite in parallel computing"
A 'computer rights' mujihadin
It seemed like the pawn odds that Rybka gave actually helped the engine. In a few of the games Rybka used the open file or open lines from the start to help in an attack.
Maybe the best way to go in Man vs. Machine matches is the time odds where the program cannot think during the human's move. The Zappa - L'Ami match seemed pretty competitive.
I would love to see Hydra versus Rybka. It would would almost be like Man vs. Machine, with Rybka representing humanity :)
It seems most chess programs have positional evaluation of a 1700 rated human, but they calculate millions of positions per second. Rybka, on the other hand, has position evaluation of a Master, but it only calculates around 100 K nodes per second. I like Rybka's chances.
I believe Rybka's programmer, an IM himself, put Rybka's knowledge level at around 1700. Of course it's impossible to isolate positional from tactical in chess, which is why Fritz kills humans just as well as Rybka or Junior, etc. And five programs can all come to the same best move by different routes with different evaluations. But it plays pawns much better than any engine I've ever seen and it does relatively very well with material imbalances that often confuse engines. Simply put, it looks like it's playing chess most of the time.
Chess is mentioned in the article linked to below.
The gist is that computers have a harder time playing great Go, but there is recent progress.
http://news.com.com/Algorithm+helps+computers+beat+humans+at+Go/2100-1043_3-6161042.html
TITLE: Algorithm helps computers beat humans at Go
these are interesting but artificial handicaps.
the solution seems so obvious; computers should be deprived of opening books and, potentially, tablebases. human players will do what they do well; engines will do what they do well.
and everyone benefits; humans go back on top for a few years and chess programmers get the challenge of trying to beat humans again.
"It seemed like the pawn odds that Rybka gave actually helped the engine. In a few of the games Rybka used the open file or open lines from the start to help in an attack."
I think this is what Paulsen claimed Morphy would do if they were to play at pawn odds. I believe Morphy scoffed at this, but everybody in that Romantic era was so addicted to gambits it's reasonable to assume Morphy would have used the open file for all it was worth and maybe even win a game or two with it.
How many games have been played with postal chess-like time controls? Has that been done in a serious match?
"I believe Rybka's programmer, an IM himself, put Rybka's knowledge level at around 1700. Of course it's impossible to isolate positional from tactical in chess, which is why Fritz kills humans just as well as Rybka or Junior, etc. And five programs can all come to the same best move by different routes with different evaluations. But it plays pawns much better than any engine I've ever seen and it does relatively very well with material imbalances that often confuse engines. Simply put, it looks like it's playing chess most of the time."
-Posted by: Mig at March 9, 2007 15:20
I really don't see the substance of this argument which has been repeated time and time again. The end goal of any top level sporting competition (or gaming competition if you like) is to win. If the latest man vs. machine matches have shown anything, it's that winning can be achieved by strong tactical play at the expense of positional understanding. Does this not suggest that the cherished 'chess IQ' superiority of humans amounts to little more than a rallying point for berating machine play? An advantage in any one particular facet of a given sport is not all that meaningful unless it leads to victories.
Cynical Gripe, on the contrary, because of Rybka programmers are beginning to focus on adding more positional knowledge to their engines instead of tactics.
It's two different arguments. Gripe is saying that chess is only and exclusively a sporting contest and that because computers beat GMs they "know" more, or play better chess. I'm not arguing that computers can't beat GMs. We know they can. But if you see chess as more than that, as a discipline in which humans try to learn and improve in order to understand more and win more, the how and why matter a great deal. Computers beat GMs because chess can be reduced *enough* to tactics and calculation and because computers don't blunder. That doesn't mean we can learn much about improving our own chess from a computer that plays at a 3200 level in some positions. We can't imitate the way they play, and vice-versa.
Rybka seemed to do ok in this match without an opening book based on GM play. Of course Ehlvest didn't have an opening book either.
Well said Mig. For all I know, it may turn out that engines built with more positional understanding gain a significant edge over the current standard.
shams, I don't agree. This discussion has come up numerous times in other threads, and everyone seems to have their own interpretation of what's fair or 'how to level the playing field'. I think that a perfect (and large) memory are natural advantages of a computer. All current chess players have benefited from the wealth of opening and endgame theory developed by their predecessors, so why should computers be deprived of the same luxury? Man vs machine matches are gradually losing their appeal due to machine dominance or the ridiculous handicaps being employed. One gets the sense that the purpose of these matches is more and more for advertising / marketing than to settle any questions about chess. Regardless, I think the fear of chess being ruined or degraded by computers is totally baseless.
I agree w/ Mig's phrase "chess is tactical enough" that computers will win from now on.
It will soon be trivial for computers to be calculating tactics far beyond whatever humans can forsee (or memorize). It will be humans who have the event horizon limitation, as computers will be calculating and comparing the tactics available in positions ten or more full moves away, and not missing a trick. However good at playing blindfold, the human will not be able to hold in mind all the critical variations.
The definition of "positional understanding" as, "what to do when you don't see any tactics," has become irrelevant for the computer, as it sees enough tactics to defeat every human. The computer won't need any positional understanding, as tactics will be enough.
This doesn't mean positional understanding is no longer useful, humans will always need it when acting without computer aid. It does mean our current positional understanding is inadequate against the computer's tactical calculating ability.
Perhaps humans will come to dispense with positional understanding completely, and rely on the new positional dictum of our age: Consult Rybka!
freedom fighter wrote:
"Freedom now !"-- "Let computers think !"--"Computers of the world unite in parallel computing"
A 'computer rights' mujihadin"
Be careful what you wish for. I caught "The Terminator" last week. If computers ever do reach the point where they start thinking for themselves, then when it comes to demonstrating their superiority over us pathetic humans, does anyone think they'll be content to stop at chess?
if you are interested in computer chess, you should definitely see the following google video (i don't know if this has been discussed here before, except that this is the second time i'm posting this link). you can also get to this video by searching for "history of computer chess" on video.google.com.
http://video.google.com/videoplay?docid=-1583888480148765375