After all the post’s lately regarding programs. I wrote this up . For
anyone worried or interested about playing against a “Program” I
would suggest finding a copy of “Playing Computer Chess” by Gm’s Al
Lawrence and Lev Alburt. I found this for $4 at a Book Warehouse.
This book contains some excellent insights on how chess computers
and programs think., What their weakness’s are and how to exploit
them. 90% of this knowledge will also help you in your game against
human opponents. It will make you also realize a program isn’t
superman and has some real faults.
Some tips from the book:
Realize a program uses an Event Horizion – just as you
can boil a frog by putting it in a pot of cool water and ever so slowly
raising the temp till its boiling and the frog will never realize till to
late. You can slowly beat a computer by slowly easing toward your
goal instead of trying to take the game in one move. Boris Gulko
has made similar statements in his articles.
If you don’t want to win slow. Learn to win the game in the midgame.
Most computer programs nowdays use tablebases or databases in the
endgame. If you can mate them before ever getting to the midgame
you have the advantage. Because they are having to work from a
finite set of calculations and not a reference point.
Try and Castle queens side most programs still have a problem with
this.
Learn to use decoys. If you can have multiple lines of attack going or
switch tactics for a moment you can really mess up a computers train
of thought. Wheras it may have on move 12 have your game planned
out for 40 moves if you switch threats for a move or two it will force it
to start back to square one and may miss your attack until too late.
Bank rank mates – most programs are weak in this area and you can
catch them with their pants down
The biggie – use a wasteful pawn move such as H3,H4, A1 as your
first move. This will immediately throw the computer out of its opening
database which can be very elaborate and force it calculate each
move. You on the other hand can drop into your favorite opening and
play it to the advantage. Someone asked in the forums a while back
why anyone would open H4, That’s why… Force the computer from the
familiar to strange territory.
This book rips away a lot of the mystique of playing a computer and
shows just how really vulnerable they are.
Some things I have learned but aren’t in the book.
Your #1 strength is you only have to beat a computer once. After you
have beaten Fritz,Shredder, Bubba whomever you can use the same
moves/strategy you did to beat it over and over again.
Each program has its own strength and weakness and once you learn
its footprints you can detect it and use those to beat it.
Best way to describe is say you are in a race and a fella has beat you
and your friends by riding a motorcyle instead of running. Realize you
goal is now to learn to use the terrain to beat the motorcycle. Because
once you can beat the motorcycle you can beat anyone riding the
same motorcycle. Human vs human is always a challenge because we
are unpredictable and can have flashes of brillance, but once you can
beat that motorcycle or that program you never have to fear it again.
I hope some of these tips help. This is a wonderful site to play in and
a bunch of good folks here. Programs have their place, To work on
your games and strategies and to help in post game analysis. But for
playing against another human nothing beats your pawn mind and
resources. A program doesn’t benefit or get smarter by playing a
human but We do. Good luck my friends and do not let the darkness
hide the light. Enjoy your adventures in the Black and white jungle.
Interesting stuff.
It triggers thoughts of an experiment I tried to run to figure out if
someone was using a progam. Since there are too many programs
and too many ways to modify their parameters, I felt it was impossible
to look at openings and midgames to see if your opponents moves
were the moves of a particular program.
But....
But the endgame tablebases programs use are identical. Essentially
everyone uses the Nalimov end tables these days. As the number of
pieces go up, the size of the corresponding databases get HUGE. So
at the current limit is 5 pieces total, say kr vs krp. Everything up to but
not including kpp vs kp takes about 6 GIGABYTES of hard drive. A few
6 piece tables have been done, and they alone are the size of a small
hard drive.
OK, I'm digressing. Point is, the table bases cover endings with only a
few pieces (and by piece I'm mean everything, King, pawn, piece); but
what they cover will be played perfectly by a computer - not a single
misstep, just a series of moves that wins the game in the absolute
shortest time possible. (or if the computer is playing the losing side,
the set of moves that will delay the loss for as long as possible).
So what I was going to do was just check if peoples endgames
followed the exact moves the Nalimov tables said to do.
I looked at about 100 games.
And here I ran into a problem.
OF THOSE ~100 GAMES, NOT A SINGLE ONE OF THEM GOT DOWN TO
FEW ENOUGH PIECES TO USE END GAME TABLES. So my strategy of
discovering program use didn't work. And it taught me something
about the relative importance of those "killer" end game table bases.
Of course it makes me wonder about writing programs that can also
learn to play to guaranteed endgame winning posistions in the
tablebases. IE, at some point drop as a goal "figure out how to
checkmate" and instead pick up as a goal "manuever the game into a
winning position in the database". Or do they already do that?
And I'm still curious about grandmaster games; are top players,
in "end game table base" positions, as "efficient" as as computers?
Well, just babbling away here.
Excellent post Michael!
Ok...I can help a little with your last question...players of master
strength are that way due to one thing...pattern recognition. I had a
knight and queen mating pattern in a game not too long ago...and
was asked how I could calculate that far into a position...it was actually
a mate in 22 with perfect play...ended up quicker than that...but
books like "Combination Challenge" by Hays and Hall.."1001 Brilliant
Checkmates" and "1001 Brilliant Sacrifices and Combinations" both by
Reinfeld...and the book I recommended to you.."The Art of Attack in
Chess" by Vuckovic ar books that teach these positions and patterns
to you.
GM's have these things memorized alot better than me....and they
also have endgame patterns memorized ALOT better than me...that is
what they have been trying to do with computers as of late. They say
Bobby Fischer didn't think "in moves"...but "in systems"...he would
always play to keep at least good pawn structure for the endgame.
They are just now getting computers to do this.
I can elaborate more if anyone wants.
Dave
Captain USA
Doc,
This is all way up above my understanding. If I knbow i am facing
a program. I will do one of three things. Move as slowly as a snail.
Convince my opponent of the fun of playing against a human
opponent. Or delete or resign the game congratulating his program.
John
PS: Another tactic. I let everyone I know in message log of what my
opponent is doing, but I wont post anything about it in the forums. I
will not give him/her any publicity.
John
Excellent way to handle it John...I always just say it in email or
through message logs as well.
Most people that do use "extra" means to win games either are very
quiet about it...or they come on to the forums and are very
arrogant...as we have all seen a few times on here.
The best thing to do is like you said...alert all of your buds about
it...and hopefully when the word gets around enough..the person will
realize they have been figured out and will leave.
And of course it is just humiliating to begin with to have to resort to
something like that to make yourself seem like you are something
you are not.
Dave
Captain USA
USCF NM 2219
- Joined
- 09 Aug 01
- Moves
- 54191
long-term strategic concepts are still difficult for machines to
evaluate. in one successful game against a suspected machine i
employed the minority attack ( the machine was truly clueless) ; in
another, i played the standard rxc6 exchange sac against the sicilian
and confused the suspected machine. so i opinion that strategic plans
is the way to play against machines.