Many times in several newsgroups and forums I've read questions, references
and advice about chess-engines/programs that are supposed to play
'human-like' chess. (Very often Rebel/ProDeo comes up in these cases, btw.).
Chess being a somewhat mathematical game, I've never understood these
remarks. Besides, several humans play different styles and types of chess,
so that could also be the case with programs, which after all are developed
by (different) humans.
I often thought it was because some engines make positional mistakes
(human-like), or the tactical accuracy of engines , making them sometimes
nasty to play (robot-like), but I believe these 2 are both present in
Chess-engines, so that cannot be it.
Am I overlooking something here?

The Turk.



John

--


Von Herzen, moge es wieder zu Herzen gehen. --Beethoven

Ruud wrote in message
...
Many times in several newsgroups and forums I've read questions,
references
and advice about chess-engines/programs that are supposed to play
'human-like' chess. (Very often Rebel/ProDeo comes up in these cases,
btw.).
Chess being a somewhat mathematical game, I've never understood these
remarks. Besides, several humans play different styles and types of chess,
so that could also be the case with programs, which after all are
developed
by (different) humans.
I often thought it was because some engines make positional mistakes
(human-like), or the tactical accuracy of engines , making them sometimes
nasty to play (robot-like), but I believe these 2 are both present in
Chess-engines, so that cannot be it.
Am I overlooking something here?


Computer Chess playing programs initially played with a ruthless attacking
style, where they always attacked, attacked, attacked. Over time these
programs have developed a positional style where the all-out attacking
methods have been restrained some while adding intelligent choices to the
playing style. The "human" quality seems (in my opinion) to be found when
the program is not always playing with severe attacks at all times, but
instead seems to have a developing plan, and where the choices for moves
have variety, and where it plays artistically. I have various Chess
programs, and I can remember when I first got Shredder 6 how much I enjoyed
its playing style because it did not just play extreme attacking moves
immediately at all times, but develops a careful stategic game that would
take me by surprise. I still enjoy losing to it. I have Rebel (various
dos versions and the windows version) and I have really liked this program
from day 1 also, very nice playing style.

I don't equate playing weak to playing with human style necessarily. You can
see a somewhat mechanical choice of moves when playing against a handheld
chess computer more so than many of the new Chess playing programs - at
least I see this somewhat from my $15 Excalibur portable. I think that many
authors of chess programs have gone a long way to introduce some human
qualities to gameplay.

I have Junior 9 which I really like, and feel the authors put some good work
into its playing style, which has human qualities. Deep Sjeng would be
another, having personalities to choose from. ChessMaster 9000 has a whole
lot of personalities with blends of human and borg type play, but also
offers low to GM strength levels so you can dial in your choice (good when
you want a fighting chance to win). I like the playing some GM levels of CM,
then bounce down to lower level players to practice tactics,
opening/midgame/endgame, and to win from time to time. The Lasker GM is a
fun one among many of the others. I like reviewing GM games from the past
masters, then trying to play their representations in CM. Hiarcs 9 has human
qualities, but they seem to be that of Kasparov to me because it is very
strong. I do like the wealth of opening knowledge - very nice to have many
openings choices.

It is true that human play does not fall easily into a specific pattern -
this can be seen by looking at different styles: Morphy, Alekhine, Lasker,
Capablanca, among many others. Each played a great game, each had different
methods, many shared the same methods, but overall they all played the game
differently. All human, alike and yet different.

A computer really has a tough example to try and live up to, especially with
the method most commonly used for programming, which mostly consists of
testing a great many moves to a certain depth, discarding moves that lead to
bad outcomes, choosing the best from a set of good outcomes. The computer
applies a brute force approach where it evaluates as many moves as quickly
as possible to reach a conclusion. The programmers do find ways of improving
the game play and also to enhance the playing style to the degree that the
program can play with more human like style. When they do this it is quite
an accomplishment. I have noticed different playing styles between many
chess programs, and even more so by using settings or opponents available
within each program. It makes different programs fun to play against for the
variety.

A person sees patterns, uses imagination, remembers previous outcomes based
on the same or similiar moves and patterns on the board. A person can have
the edge because of creativity, but it is the computer that has the greater
memory (usually), more knowledge of variations, and it does not tire or
overlook most weaknesses. A person may be more inclined to use the sacrifice
to win, where a computer may not always see the expected result of the
sacrifice. The human player will understand some chess positions more fully
because we have a real brain, where the computer may not understand it in
depth. Obviously the strength of the human player does play a factor into
this. I myself am still working at attaining many of those GM qualities
(maybe someday).

my thoughts, matt