Fischer vs. MIT's Greenblatt program, 1978
Robert Hyatt
: <!doctype html public "-//w3c//dtd html 4.0 transitional//en">
: <html>
: Here is the game I have. At the time, Greenblatt was state of the
: art in chess computers, probably playing around expert strength.
: Bobby makes quick work of it. Note the nice knight sacrifice at move
: 10.
: <p>1. e4 e5 2. f4 exf4 3. Bc4 d5 4. Bxd5 Nf6 5. Nc3 Bb4 6. Nf3 0-0 7. 0-0
: Nxd5 8. Nxd5 Bd6 9. d4 g5 10. Nxg5 Qxg5 11. e5 Bh3 12. Rf2 Bxe5 13. dxe5
: c6 14. Bxf4 Qg7 15. Nf6+ Kh8 16. Qh5 Rd8 17. Qxh3 Na6 18. Rf3 Qg6 19. Rc1
: Kg7 20. Rg3 Rh8 21. Qh6++ (1-0)</html>
Actually, mack hack was a 1500-level player. The first 'expert' program
didn't show up until the late 1970's in the body of chess 4.x from
Northwestern. Mack Hack dates to the late 60's and was _highly_ selective.
I used to have the source code for this thing many years ago (pdp 10
assembly language). It could search roughly 5 plies deep in tournament
time controls, and typically searched 15 moves at ply=1/2, 9 moves at ply=3/4
and 7 at ply=5. Seems very selective, but the computers back then were very
slow also.
--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
Sylvain Lacombe
Tracy Miller wrote in message <36D725E7...@infi.net>...
Here is the game I have. At the time, Greenblatt was state of the art
in chess computers, probably playing around expert strength. Bobby makes
quick work of it. Note the nice knight sacrifice at move 10.
1. e4 e5 2. f4 exf4 3. Bc4 d5 4. Bxd5 Nf6 5. Nc3 Bb4 6. Nf3 0-0 7. 0-0
Nxd5 8. Nxd5 Bd6 9. d4 g5 10. Nxg5 Qxg5 11. e5 Bh3 12. Rf2 Bxe5 13. dxe5 c6
14. Bxf4 Qg7 15. Nf6+ Kh8 16. Qh5 Rd8 17. Qxh3 Na6 18. Rf3 Qg6 19. Rc1 Kg7
I really don't get the move 19. ... Kg7??
I wonder why it played that move? Probably a bug.
You don't have to search very deep to see that this move is very bad.
Maybe there's something i don't get.
Monik 1.9 says, Nb4.
Ohall10
>art in chess computers, probably playing around expert strength.
>Bobby makes quick work of it. Note the nice knight sacrifice at move
Expert Strength?? You have got to be kidding! That thing is not playing even
close to expert strength. I Imagine fischer must have been very bored with the
game
Robert Hyatt
: Tracy Miller wrote in message <36D725E7...@infi.net>...
: Here is the game I have. At the time, Greenblatt was state of the art
: in chess computers, probably playing around expert strength. Bobby makes
: quick work of it. Note the nice knight sacrifice at move 10.
: 1. e4 e5 2. f4 exf4 3. Bc4 d5 4. Bxd5 Nf6 5. Nc3 Bb4 6. Nf3 0-0 7. 0-0
: Nxd5 8. Nxd5 Bd6 9. d4 g5 10. Nxg5 Qxg5 11. e5 Bh3 12. Rf2 Bxe5 13. dxe5 c6
: 14. Bxf4 Qg7 15. Nf6+ Kh8 16. Qh5 Rd8 17. Qxh3 Na6 18. Rf3 Qg6 19. Rc1 Kg7
: 20. Rg3 Rh8 21. Qh6++ (1-0)
: I really don't get the move 19. ... Kg7??
: I wonder why it played that move? Probably a bug.
: You don't have to search very deep to see that this move is very bad.
: Maybe there's something i don't get.
: Monik 1.9 says, Nb4.
See my previous note. The Greenblatt program was _very_ selective. Such
a move is a definite 'signature' of selectivity... if a move looks bad
at ply=1 it may _never_ be searched at all. And you can step right on
a land mine as a result...
I had a 'selective' program myself, same sort of ideas originally, and I
had the _same_ problem.... 39 brilliant moves, then one 'lemon'. Problem
was the 'lemon' was game-pivotal...
Sylvain Lacombe
>a move is a definite 'signature' of selectivity... if a move looks bad
>at ply=1 it may _never_ be searched at all. And you can step right on
>a land mine as a result...
>
>I had a 'selective' program myself, same sort of ideas originally, and I
>had the _same_ problem.... 39 brilliant moves, then one 'lemon'. Problem
>was the 'lemon' was game-pivotal...
>
Yes, after i read your note, it was quite obvious how it could have made
this move. I tried to write one myself when i was around 11. After 2 or 3
months, i gave up and made a poker game instead.
At that time, i didn't know anything about recursivity or search trees. If i
wanted to go to deep 3 i had to rewrite the search code for 3 plys. I was
doing this in basic. I had no internet to help me out and of course no
crafty code. :)
Sylvain.
RDavis101
>wanted to go to deep 3 i had to rewrite the search code for 3 plys. I was
>doing this in basic. I had no internet to help me out and of course no
>crafty code. :)
>
>Sylvain.
>
That's great. What was your evaluation function like?
Sylvain Lacombe
RDavis101 wrote in message <19990227055541...@ng43.aol.com>...
Very standard. The real problem was that i didn't knew Depth First Search.
The only search i could think of was the Width First Search. That's why i
had to keep only a few move per ply. But, with basic, i didn't have any
pointers and knew nothing about those. So it was a real problem.
Sylvain.
Rolf Tueschen
<7b7uae$juk$1...@juniper.cis.uab.edu>:
>I had a 'selective' program myself, same sort of ideas originally, and I
>had the _same_ problem.... 39 brilliant moves, then one 'lemon'. Problem
>was the 'lemon' was game-pivotal...
I have a revolutionary question for computerchess. How can it ever
happen that a 1500 player&programmer (e.g. Mr. Hyatt) can differentiate
between brilliant moves and 'lemons'? Any ideas? Isn't it more about
_hyperbole_?
_______________________
Extraordinary disclaimer:
I don't want any censoring answer/comment from Bob Hyatt on this post.
As my babysitter or watch patrol. This is a free forum for
computerchess. Thank you.
Igor Syry
SHUT THE F**K UP