"Adams
can beat Hydra"
This assessment is given by Correspondence Chess Grandmaster Arno Nickel from
Berlin, who recently won a match against Hydra by 2:0 in a correspondence chess
match, after a six month struggle. On the occasion of the forthcoming match
Hydra vs. Adams the correspondence chess server chessfriend.com
asked Arno Nickel about his experience with the so-called "hardware monster".
Chessfriend.com: You won two correspondence
match games against the previous version Hydra Chimera. Would you draw any
conclusions from this match for the man/machine event in London? What is your
bet?
GM Arno Nickel: Well of course, correspondence chess and over-the-board
chess are nowadays two extremely different disciplines, more so than ever, because
correspondence players in contrast to o-t-b players have full access to computer
engines and databases. Nevertheless my test games against Hydra (I think, we
should call them this way) show some aspects of the potential of Hydra's chess
abilities as the successor to Deep Blue. In the two games, where I played White
against the Sicilian and with Black against the Tarrasch-System in the French
Defence, Hydra disregarded principles and unnecessarily got into difficulties,
once in the middlegame and the again in the endgame. If the positions are not
guided by tactics, I wouldn't be surprised, if a super grandmaster like Adams
will give the machine a thrashing on the board. A lot depends if he manages
to neutralize Hydra's tactical power and in reaching human type positions, where
long sighted plans are called for.
This in fact occurred to some of the leading grandmasters, who played
Hydra in Bilbao in 2004 and who failed to follow your strategy. Namely ex-FIDE
World Champion Ponomariov and the youngster Karjakin (both from Ukraine),
while the Bulgarian Top-GM Topalov was the only one to take half a point from
the machine and even had winning chances at one stage. When this became possible...
Nickel: ...exactly, it was in this game, where Hydra also
blundered positionally. You cannot eliminate such deficits overnight, and without
intensive training against very strong players...
However Hydra Scylla is five times faster than the previous version,
which "only" calculated about 40 millions moves per second.
Nickel: As said before, in positional situations such numbers
are completely meaningless. Really important is the direction in which the program
goes and how it evaluates the positions. Once the machine has got the wrong
idea, it will not correct itself by deep calculation in quiet positions. This
was evident in my correspondence games, where Hydra had lots of time and could
calculate much deeper than the new Hydra will be able to in a classical tournament
game with four to six hours for all of moves. The results were not better in
most of the critical positions.
Some deficits of Hydra were also seen in the recent PÂL/CSS-Freestyle
Tournament on playchess.com, where
the sea monster didn't even reach the quarter finals.
Nickel: Well, this is of course a hot potato since Hydra as
in correspondence chess had to play against strong players, amongst them even
some FIDE grandmasters who could use engines as in "advanced chess".
I was also surprised by Hydra's failure, but on consideration I realised that
all programs including Shredder 9, Fritz 8 and Junior 9 which were also running
on strong machines, got into trouble, when playing without human advice. Probably
this wouldn't have happened with shorter time limits, let's say 15 minute games
or even shorter, but in one hour games (one hour per player plus 15 sec. per
move), human chess competence can already show its qualities in tandem with
engines. By the way, I also participated in this very interesting event and
didn't manage to reach the quarter finals either. However in the process I met
Hydra again and played a nice game which was drawn after 102 moves. The sea
monster must have gotten tired of attacking my fortress without success.
Once again what is your bet on the match Hydra-Adams...?
Nickel: Okay, it will very hard for any human being, but as
I always hold on for humanity, and especially as Michael Adams is a gifted positional
player, I think he might achieve a draw, that means 3:3. Maybe he can even beat
Hydra, if he is able to completely control his emotions and avoid any unclear
complications. Should this be so, he could succeed where Kasparov failed when
playing Deep Blue eight years ago. This would mean that humans have learned
since then even though the computers didn't get weaker, but obviously stronger
since that time. Hydra, don't forget, is supposed to be stronger than good old
Deep Blue!
Postscript: the correspondence chess match Hydra vs. Nickel
on the Chessfriend server isn't over yet. The Hydra Team is ready to continue
the match against GM Nickel by playing two further games. This will take place
after the match with Adams.
Man versus Machine: the battle for supremacy
The multi-processor hardware system “Hydra” will make its worldwide
debut in London on 21-27 June, playing a six-game match against the UK’s
top GM Michael Adams. The event will take place at the Wembley Centre in London, with a purse of $150,000 (Ł80,000) up for grabs. The prize fund will
be given to the players on a per-game basis, the winner of each game getting
$25,000, whereas in case of a draw both opponents get $10,000 each. If Adams
loses a game he apparently gets nothing.
Hydra is a hardware-driven chess machine. The basic hardware consists of a
32-processor cluster with the same number of FPGA hardware chess cards. These
do the brute calculations at the ends of the search tree. The Xilinx Field Programmable
Gate Arrays are able to reach much greater computational speeds for specific
tasks than the general-purpose 3.06 GHz Xeon processors in the host machine.
The Xeons only do the initial part of the search. On this hardware configuration
Hydra has a processing power of 100 billion calculations – or 200 million
chess moves – a seconds.
England's top GM Michael Adams, born on November 17, 1971, lives in Taunton,
Devon and London. He first won the British Championships in 1989 at the age
of 17, becoming a grandmaster in the process. Adams was awarded Player of the
Year by the British Chess Federation in 1990, 1993-96 and 1998-2002 –
a record number of titles. He has been an international chess professionally
for nearly two decades, reaching his highest FIDE ranking of third in the world
in 2002. He is currently ranked number one in Britain and number seven worldwide,
with an Elo rating of 2741.
Schedule
Game 1 |
21 June 2005, 3 PM |
Game 2 |
22 June 2005, 3 PM |
Game 3 |
23 June 2005, 3 PM |
Game 4 |
25 June 2005, 3 PM |
Game 5 |
26 June 2005, 3 PM |
Game 6 |
27 June 2005, 3 PM |
The games will be covered live on Playchess.com
in a special broadcast room:
Links