Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old March 8th 04, 09:55 AM
Bruce Sterling Woodcock
 
Posts: n/a
Default Weakest CM9000 GM Personality?

I was curious if anyone here had opinions on what
the "weakest" GM Personality is in CM9000. They
all have different styles, so I'm sure it depends upon
what style you like to play. But although the game
rates them all equally, it seems likely that in a series
of head-to-head matchups some are easier to defeat
than others.

I found an old post that rated the CM6000 GMs
according to a tournament they ran, but CM9000
has new personalities not in that list and the others
may have been modified in strength. I picked some
of the glaringly weaker players from the old list and
combed through the descriptions of the CM9000
players to come up with a ten-person tournament of
what I guessed were the weakest GM personalities.
I played them 2 games each, 15 minute time controls,
with the following results:

Name Wins Draws Losses Games Score
Spassky 9 6 3 18 12.0
Shirov 7 9 2 18 11.5
Marshall 8 6 4 18 11.0
Pillsbury 5 12 1 18 11.0
Short 6 9 3 18 10.5
Tarrasch 6 5 7 18 8.5
Reshevsky 5 7 6 18 8.5
Petrosian 2 12 4 18 8.0
Leko 2 6 10 18 5.0
Flohr 0 8 10 18 4.0

It's interesting that the results fall into three distinct
strength brackets. While 18 games is by no means
definitive, I think it's obvious that the Flohr personality
is relatively very weak.

It's also interesting to note that the following have the
lowest Draw ratings; that is, they are likely willing to
draw even when they are up the corresponding number
of pawns:

Flohr: -4.0
Petrosian: -3.5
Leko: -3.0

And, surprise surprise, they are also the three lowest on
this list. Also, Pillsbury is a -1.0, and has a highest number
of draws and fewest wins in his "bracket".

It seems to me that for a human player, playing Flohr,
Petrosian, and Leko are going to be easier GMs to play,
not only because of their play, but also because if you
find yourself down a couple of pawns or even a whole
piece, you could ask for a draw. However, it is possible
that one of the other CM9000 personalities is still weaker,
just less draw-ish.

I'm curious if anyone else has done either any direct
testing via CM9000 tournaments or if you have any
first-hand opinions from playing them.

Bruce


  #2   Report Post  
Old March 8th 04, 12:56 PM
Bruce Sterling Woodcock
 
Posts: n/a
Default Weakest CM9000 GM Personality?

Yes, I'm following up to myself.

I just thought I'd add some more evidence...
I found this web site which included someone's
CM8000 tournament:

http://www.angelfire.com/home/bstjea...computers.html

Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.

Bruce


  #3   Report Post  
Old March 8th 04, 01:55 PM
Mikko Nummelin
 
Posts: n/a
Default Weakest CM9000 GM Personality?

On Mon, 8 Mar 2004, Bruce Sterling Woodcock wrote:

Yes, I'm following up to myself.


I just thought I'd add some more evidence...
I found this web site which included someone's
CM8000 tournament:


http://www.angelfire.com/home/bstjea...computers.html


Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.


One of the most difficult tasks when developing a chess engine is the
evaluation function and how different material and positional values
should be tuned. These "personalities" are, in fact, just saved evaluation
settings and I am pretty sure that the original developers are doing
constantly very intensive tests with far more different settings that are
actually available to the end user. Those testings might involve for
example large RR tourneys of tens of different copies of the same engine
with different settings run on multiprocessor architectures etc. Most
successful settings are chosen as default settings & for tournament use.


Mikko Nummelin
  #4   Report Post  
Old March 8th 04, 02:13 PM
David Richerby
 
Posts: n/a
Default Weakest CM9000 GM Personality?

Bruce Sterling Woodcock wrote:
I just thought I'd add some more evidence...
I found this web site which included someone's CM8000 tournament:

http://www.angelfire.com/home/bstjea...computers.html


Note that this is a Polish (1.b4) thematic tournament. Someo of the
personalities may not play well in the sort of position that comes from
that opening.


Dave.

--
David Richerby Mouldy Painting (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ Renaissance masterpiece but it's
starting to grow mushrooms!
  #5   Report Post  
Old March 8th 04, 02:19 PM
Bruce Sterling Woodcock
 
Posts: n/a
Default Weakest CM9000 GM Personality?


"Mikko Nummelin" wrote in message
i...
On Mon, 8 Mar 2004, Bruce Sterling Woodcock wrote:

Yes, I'm following up to myself.


I just thought I'd add some more evidence...
I found this web site which included someone's
CM8000 tournament:


http://www.angelfire.com/home/bstjea...computers.html


Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.


One of the most difficult tasks when developing a chess engine is the
evaluation function and how different material and positional values
should be tuned. These "personalities" are, in fact, just saved evaluation
settings and I am pretty sure that the original developers are doing
constantly very intensive tests with far more different settings that are
actually available to the end user. Those testings might involve for
example large RR tourneys of tens of different copies of the same engine
with different settings run on multiprocessor architectures etc. Most
successful settings are chosen as default settings & for tournament use.


You're partially right to start with... the personalities are also
different opening books, which may have weaknesses in
certain lines, as well as other tendancies to attack or defend
which may override what the situation ideally calls for.

As for testing, yes, they do far more extensive testing to rate
their personalities. But you're missing the point. There are
over 30 different grandmaster personalities, all "rated" the
same (grandmaster status treats them all as having the same
rating - 2700). But it seems obvious that if they all played
each-other hundreds of times, they aren't all going to come
out equal. The point of this thread is to estimate which are
the weakest to play against, especially for a human.

It is statistically very unlikely that Flohr played so poorly in
our tests, but somehow wins a bunch more games against
other opponents. At best, it is actually a "middle of the pack"
GM, and there are some other personalities who are weaker.

Perhaps in the end the actual relative difference between
the best GM personality and the weakest GM personality
is only 100 points or so... but that could be an important
difference for a strong player.

Bruce




  #6   Report Post  
Old March 9th 04, 03:47 AM
Entropyfoe
 
Posts: n/a
Default Weakest CM9000 GM Personality?

Bruce,

Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.


Recall that Flohr was one of the "weaker" grandmasters, and I hate to use that
term, for fear of starting flame wars. Indeed, the three quoted above were
never word champions, so cannot compare with Spasky, Tarrasch, or Petrosian,
who each dominated their contemporaries, at least for a while.
I have even [gasp] read that Petrosian was one of the weaker champions ! He
won less, but he was very very hard to beat !

As a later poster wrote, this may have to do with the opening books selected
for each personality, an how the kind of positions that arise are suited to the
computer's evaluation function.

-Jay
entr0pyf0e
  #7   Report Post  
Old March 9th 04, 06:59 AM
Bruce Sterling Woodcock
 
Posts: n/a
Default Weakest CM9000 GM Personality?


"Entropyfoe" wrote in message
...
Bruce,

Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.


Recall that Flohr was one of the "weaker" grandmasters, and I hate to use

that
term, for fear of starting flame wars.


Indeed, I just played Flohr and beat it quite easily. My first GM victory!
I've attached the PGN at the end of my post, with both my commentary
and Fritz's.

As for why Flohr is so weak in CM9000, I think it might be a combination
of a strong Defender preference with a strong Positional preference. Flohr
has some of the highest ratings in either, and the other players which rate
highly in both categories also seem to come out weaker in tests (Leko,
Reshevsky) compared to the other GMs. Maybe this just means that
Chessmaster is much better at tactical play than positional.

I don't think my play was particularly masterful... I just didn't make any
big mistakes, and I had been studying the Caro-Kann and decided that
Bf5 would put black in trouble right from the start. So when Flohr
played it, I was ready with a good response. Even if I had made a
mistake or two, with Flohr so willing to draw I'm sure I could have
gotten one.

Bruce

[Event "Chessmaster 9000 Rated Game"]
[Site "?"]
[Date "2004.03.08"]
[Round "?"]
[White "Bruce Woodcock"]
[Black "Flohr"]
[Result "1-0"]
[ECO "B19"]
[Annotator "Deep Fritz 8 (30s)"]
[PlyCount "97"]
[EventDate "2004.??.??"]

{B19: Classical Caro-Kann: 4...Bf5 main line} 1. e4 c6 2. d4 d5 3. Nc3 dxe4
4.
Nxe4 Bf5 {Here I think black makes his first mistake... I can now harass his
bishop and gain tempo. - BW} 5. Ng3 Bg6 6. h4 h6 7. Nf3 e6 8. Ne5 Bh7 {
Now that black is trapped, I can exchange bishops on my terms. - BW} 9. Bd3
Bxd3 10. Qxd3 Nf6 11. Bd2 Nbd7 12. f4 {
I not only reinforce e5 but I continue my march up the Kingside. - BW} Bd6 (
12... Be7 13. O-O-O O-O 14. Qe2 c5 15. dxc5 Nxc5 16. Bc3 Qc7 17. f5 exf5 18.
Nxf5 Rfe8 19. Qf3 Bf8 20. Nxh6+ gxh6 21. Ng4 Nxg4 22. Qxg4+ Kh7 23. Qf5+ Kg8
24. Qf6 Kh7 25. Qh8+ Kg6 26. h5+ {1-0 Adams,M-Leko,P/Linares 1999/CBM 70
(26)})
13. O-O-O O-O {He castles right into my coming attack! I think this was the
root of black's future problems. - BW} (13... Qc7 14. Ne4 Nxe4 15. Qxe4 Nf6
16.
Qe2 O-O-O 17. h5 c5 18. dxc5 Bxc5 19. Rh3 Rd4 20. Rd3 Rhd8 21. Be3 Rxd3 22.
Rxd3 Rxd3 23. Qxd3 Bxe3+ 24. Qxe3 a6 25. g4 Nd5 26. Qd4 Qa5 27. Qc4+ Kd8 28.
Nd3 {Carlsson,P-Hermansson,E/Stockholm 2002/CBM 91/1/2-1/2 (61)}) 14. Ne2
Qc7
$146 (14... Rc8 15. g4 Bxe5 16. fxe5 Nd5 17. c4 Ne7 18. Qe4 c5 19. d5 exd5
20.
cxd5 c4 21. Kb1 Nc5 22. Qxc4 Qxd5 23. Qxd5 Nxd5 24. Bxh6 Nb6 25. Bf4 Ne6 26.
b3
Rc5 27. Bg3 Rfc8 28. Rc1 Rxc1+ 29. Rxc1 {
Tseitlin,M-Alterman,B/Tel Aviv 1999/CBM 71/1/2-1/2 (97)}) (14... h5 15. Qf3
$14
) 15. g4 $16 {Now I start ratcheting up the pressure. - BW} Nd5 (15... Bxe5
16.
fxe5 Nxg4 17. Rhg1 $16) 16. g5 h5 17. g6 f5 18. c4 N5f6 19. c5 Be7 {Now I'm
rea
dy to shift all my forces to attacking Kingside. My plan is to take the h5
pawn with my night and get my Queen on the H file. I'm also thinking about
putting my other knight onto f7. But black has an unexpected response. -
BW}
20. Qf3 Rad8 (20... Rfe8 $5 $16) 21. Ng3 $18 Nxe5 {Oops. I didn't see this.
Now my knight to f7 is gone, and I've lost tempo because I have to retake
the
knight to protect my queen. This allows black to play... - BW} 22. fxe5
(22.
dxe5 $6 Ng4 $14) 22... Ng4 {... completely stopping my plan. I spend a lot
of
time now trying to formulate a new plan. But I'm thinking back to Josh
Waitzkin's lessons and I'm feeling pretty good. Look how active all my
pieces
are, whereas the computer's are hindered behind a lack of space. - BW} 23.
Bg5
(23. Nxh5 $6 Rxd4 24. Nf4 Qxe5 $11 (24... Bxc5 25. Nxe6 Qxe5 26. Rhe1 $18))
23... b6 {Black seems to be breaking down now, just giving away a pawn.
Also,
it gives me a great idea... I can play Nxh5, Nf4, and then Nxe6,
triple-forking his Queen and both Rooks! - BW} (23... Rd5 24. Kb1 b6 25.
Nxh5
bxc5 26. Nf4 cxd4 27. Nxe6 Qxe5 28. Nxf8 (28. Bxe7 $6 Qxe6 29. Bxf8 Kxf8
$16)
28... Bxf8 29. Rhe1 $18) 24. Nxh5 bxc5 25. Nf4 Rde8 {Here, I realized a flaw
in my plan... if Nxe6 then Bxf5+, and I'll have to move my King. Then he
can
move his Queen. Maybe I should have gone ahead with it anyway and picked up
a
Rook or his Bishop. - BW} 26. Kb1 Qc8 {A great move. Once again my plan is
thwarted. But still, I knew I was winning here. His Queen was isolated
away
from his King. His Rooks are blocked in. All my pieces are squeezing him
into a corner and I still haven't fully used my own Rooks yet. - BW} (26...
cxd4 27. Nxe6 Qxe5 28. Rhe1 $18) 27. Qb3 c4 {This move completely mystified
me.
It was like black was just giving me another pawn for free. - BW} (27...
cxd4 {
desperation} 28. Nxe6 Kh8 29. Nxf8 Bxf8 30. Rxd4 c5 $18) 28. Qxc4 Ba3 {
Oh... so that was what he had in mind. But to what end? He just gives up
another piece with no compensation. Was the computer trying to scare
me? -BW}
29. bxa3 Qd7 (29... Ne3 {is not the saving move} 30. Qb3 Qb8 31. Rc1 $18)
30.
Ka1 a5 {Here, I finally saw that black actually had a possible fork against
my
Rooks - Nf2. I can't believe I never saw this. Black for some reason never
played it, even though he's been sitting on g4 for a while. - BW} (30... Nf2
{
cannot change destiny} 31. Qe2 Ng4 32. h5 $18) 31. Rc1 Nf2 32. Rhf1 Ne4 33.
Rcd1 a4 {Here, I decided to go back to my previous plan. With his Knight
off
g4 I can put my Queen on h5. - BW} (33... Qc8 {
cannot change what is in store for ?} 34. Qe2 Ng3 35. Qf3 $18) 34. Qe2 Ng3
35.
Qf3 c5 {Black continues to fall apart? He just gives up his Knight. I
guess
he was hoping I'd take the pawn and then trade Queens. - BW} 36. Qxg3 Qb7 {
Yes, it looks like the computer wanted to trade Queens. I wasn't biting,
though. - BW} 37. Qf2 Qe4 {He doesn't give up! He's desperate the trade
Queens. But I just push him away again. - BW} 38. Rfe1 Qb7 39. Rb1 Qd7 40.
d5
$1 {Clearance to allow f2-c5 - I had no idea when I played this move that I
was sealing the deal... I was simply trying to prevent Qxd4+ which would
cost
me a Rook or worse when I played my prepared Qe2. - BW} Rf6 {Black makes
another bizarre move, just giving up a Rook. I guess it saw mate coming (I
wasn't sure, but I felt it was close) and was trying to do anything to throw
me off my game. - BW} (40... Qxd5 41. Nxd5 exd5 42. Qe2 Re6 43. Qh5 Rxg6 44.
Qxg6 f4 45. Rb7 Rf7 46. Qxf7+ Kh7 47. Qxg7#) (40... exd5 41. Qe2 Rf6 42. Qh5
Rxg6 43. Nxg6 Re6 44. Rb8+ Qe8 45. Qh8+ Kf7 46. Rb7+ Re7 47. Nxe7 Ke6 48.
Qxe8
d4 49. Rb6#) 41. exf6 gxf6 42. Bxf6 Rc8 43. Nxe6 Qxd5 (43... Qa7 {
is not much help} 44. Qf4 Qh7 45. gxh7+ Kf7 46. Rb7+ Rc7 47. Rxc7+ Kxf6 48.
Qg5#) 44. Qe2 {Finally! I play the move I wanted... but again black had an
answer I didn't see, since I had moved my Knight. - BW} f4 {Damn. So my
planned attack never happened despite multiple tries. Still, I knew the
game
was won... I was up a Rook and 2 minor pieces and my pawn on g6 was very
dangerous. - BW} (44... Rd8 {doesn't change the outcome of the game} 45.
Nxd8
Qd4+ 46. Bxd4 cxd4 47. Rb7 Kf8 48. Qe8#) 45. Nd8 $1 {Theme: Clearance for
e2-e6 - I was really proud that I found this move, because it sets up a back
rank mate even if he takes the Knight. - BW} Qd4+ {Black throws away another
piece. I had time on the clock so I spent quite a lot of time here making
sure I wasn't missing something my losing my bishop, but I ultimately
decided
even if I was, my superior material made mate inevitable, even if he could
win
another piece. - BW} (45... Rxd8 {Decoy to d8} 46. Qe8+ {A double attack}
Rxe8
47. Rxe8#) 46. Bxd4 cxd4 47. Qe6+ Kg7 48. Qf7+ Kh8 (48... Kh6 {
does not help much} 49. Qh7#) 49. Re8# (49. Qh7#) 1-0




  #8   Report Post  
Old March 9th 04, 10:35 AM
David Richerby
 
Posts: n/a
Default Weakest CM9000 GM Personality?

Entropyfoe wrote:
Recall that Flohr was one of the "weaker" grandmasters, and I hate to
use that term, for fear of starting flame wars. Indeed, the three
quoted above were never word champions, so cannot compare with Spasky,
Tarrasch, or Petrosian, who each dominated their contemporaries, at
least for a while.


Of course, Tarrasch wasn't world champion, either: he played Lasker for
the title in 1908 but lost heavily (+3-8=5). Nimzowitsch never played a
match for the championship; Reshevsky played in the 1948 World Champion-
ship Tournament and came equal third with Keres on 10.5/14; Flohr came
equal 8th out of ten (score +1-5=12) in the 1950 Candidates' Tournament.


Dave.

--
David Richerby Edible Transparent Clock (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a clock but you can see right
through it and you can eat it!
  #9   Report Post  
Old March 9th 04, 09:22 PM
Anders Lang
 
Posts: n/a
Default Weakest CM9000 GM Personality?


"Bruce Sterling Woodcock" skrev i meddelandet
om...

"Entropyfoe" wrote in message
...
Bruce,

Interesting, Flohr again rates near the bottom,
although surprisingly with Nimzowitsch and
Reshevsky.


Recall that Flohr was one of the "weaker" grandmasters, and I hate to

use
that
term, for fear of starting flame wars.


Indeed, I just played Flohr and beat it quite easily. My first GM

victory!
I've attached the PGN at the end of my post, with both my commentary
and Fritz's.



Congrats!!

I had a quick look at your game and am wondering if you know exactly where
"Flohr" went out of book? I would put my two cents worth on 14. - Qc7? which
simply ignores the threats you have on the board. Qc7 also happens to be the
move that Hiarcs 8 suggests which would support your suspicions as to the
"positional settings" being the root of the evil (Hiarcs being the engine
with a reputation for positional play). Both Junior and Fritz finds the
reply g4! in split seconds. Sometimes I think it will never matter how
"good" or "fast" a computer chessplayer gets, without book and endgame
tables they will still loose to most of us. I remember my first win against
Fritz 5 some years back where i sacrified a piece on move four (Cochrane's
Gambit) and the tin can never knew what had hit him.

Yours Sincerely

Anders Lang


  #10   Report Post  
Old March 10th 04, 03:51 AM
Entropyfoe
 
Posts: n/a
Default Weakest CM9000 GM Personality?

Dave was of course technically correct, Tarrasch was not a world champion, but
Tarasch was in the top 2-3 in the world, behind Lasker, and in the league with
Pillsbury and Steinetz, dominating the lower tier of grandmasters of the day.
My point was that he was stronger than the others mentioned, Flohr,
Nimzowitsch, and Reshevesky, who were whomped on by the top tier players of the
day.
-Jay
Read this web quote:
http://www.starfireproject.com/chess/tarrasch.html
At this point Tarrasch earned the right to be reckoned with, and sure enough,
received an offer to play Wilhelm Steinitz in a match for the World
Championship. Tarrasch declined to play because of his medical practice.
Steinitz, in dire need of monies, then queried Dr. Emanuel Lasker, who accepted
the World Championship match offer. Dr. Lasker won in 1894, and was Chess World
Champion for 27 years. While Dr. Lasker was alive, Dr. Tarrasch would be never
more than second fiddle, since Dr. Lasker was his chess superior in every way,
and one of the greatest chess players that ever lived.

At the famous Hastings tournament of 1895, Dr. Tarrasch came second to the
great Pillsbury, and at the great Nuremberg tournament of 1896, Dr. Tarrasch
came second to Dr. Lasker. He won first prize in the marathon tournament at
Vienna in 1898, but Dr. Lasker didn't participate. In 1905, he beat the US
Champion Frank Marshall overwhelmingly in their match, and in 1907, Dr.
Tarrasch won the Ostend Championship in classy style. In 1914, he beat his then
theoretical archrival Nimzowitch in a famous two bishop sacrifice, and went on
to be one of the 5 original Grandmasters named by Czar Nicholas (Lasker,
Capablanca, Alekhine, Tarrasch and Marshall).





Of course, Tarrasch wasn't world champion, either: he played Lasker for
the title in 1908 but lost heavily (+3-8=5). Nimzowitsch never played a
match for the championship; Reshevsky played in the 1948 World Champion-
ship Tournament and came equal third with Keres on 10.5/14; Flohr came
equal 8th out of ten (score +1-5=12) in the 1950 Candidates' Tournament.


Dave.

entropyfoe wrote:
Recall that Flohr was one of the "weaker" grandmasters, and I hate to
use that term, for fear of starting flame wars. Indeed, the three
quoted above were never word champions, so cannot compare with Spasky,
Tarrasch, or Petrosian, who each dominated their contemporaries, at
least for a while.

entr0pyf0e
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
My first win against a CM9000 GM Bruce Sterling Woodcock rec.games.chess.analysis (Chess Analysis) 3 March 12th 04 07:05 AM
CM9000: Turn off tutorial voice? Jet Jaguar rec.games.chess.computer (Computer Chess) 6 February 24th 04 06:33 PM
Teaching young children chess (also: CM9000 vs Fritz) Jestrada rec.games.chess.analysis (Chess Analysis) 0 January 8th 04 03:53 PM
Teaching young children chess (also: CM9000 vs Fritz) Jestrada rec.games.chess.computer (Computer Chess) 0 January 8th 04 03:53 PM
2-D board CM9000 laocmo rec.games.chess.computer (Computer Chess) 1 July 23rd 03 07:02 PM


All times are GMT +1. The time now is 03:18 PM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017