| |
Main
Date: 17 Mar 2006 15:24:10
From: irchans
Subject: Rybka : Why would an engine suddenly be 100 poins better?
|
At talkchess.com (http://216.25.93.108/forum/viewforum.php?f=2),
oneguy asks the following question:
--------quoted text--------------------
"On this list
http://www.husvankempen.de/nunn/cegtrating4040best.html
How is Rybka beta 100 points higher than all professional engines?
Isn't everything known about chess programming already implemented? I
also heard that PC chess software have been getting better for the last
10 or 15 years, not because programming technique got better but
because the hardware got faster. If true, how could Rybka suddenly
popup and score 100-150 points higher than other programs on the same
hardware?
If Rybka programmer tried something new that no one else ever thought
of, how is that possible given so many programmers have been working on
their engines for so many years?"
----end quoted text--------------------
What do you think?
I have recently purchased Rybka v1.1.w32 and I ran my first Arena.exe
tournament against Shredder 8. I was very impressed by the results.
108 wins for Rybka
30 wins for Shredder 8 UCI
62 ties.
According to Arena, this corresponds to rating difference of about 144
points! (There is still a lot of uncertainty with only 200 games. I
think the uncertainty is about + or - 25 points.) The large difference
in ratings might have been caused by the tournament conditions:
Blitz: 2 min + 1 second per move,
Rybka 79 MB
Shredder 71 MB
Windows XP, AMD 3200, 512 MB ram total.
Opening book: mainbook.abk provided by Arena.
Ponder Off, Adjudicate games after 250 moves.
I'm not sure if any endgame tablebases were used by Shredder. Rybka
did not use any table bases.
(Is there any other information needed?)
I set "Common hashtable size" to 64 MB, but that seemed to be ignored.
Does anyone know how to force the hashtable size?
I will try a new tournament with more time per move later.
Cheers,
Irchans
PS: This is my first post to this group. Any advice would be
appreciated.
|
|
| |
Date: 26 Mar 2006 11:19:14
From: Ange1o DePa1ma
Subject: Rybka vs. 1900 player
|
I've taken the game RD supplied and put it through Shredder.
What I found, as in my earlier investigation, is that the computer claiming
to be Fritz consistently made inferior moves (according to Shredder's
analysis).
For example:
8. ...a6
10...Nxd5
11...Ne5
28...h6 (by anyone's criteria a blunder)
31...Kh8 (another near-blunder)
I was struck at this point at how few bad moves the human made. At the same
time, I found it curious he was shifting his K back and forth from g1 to f2.
Was this part of his strategy?
36...Kh7
47...b5
52...g5 a horrible move that simply loses without a fight. Shredder doesn't
even think this is the 8th best move which, considering Black only has 10
moves he can make, is interesting.
55...Kxg5 another lemon. ...hxg5 puts up a bit more of a fight.
Perhaps there's some explanation as to why Black played so many inferior
moves. My theory is that this wasn't Fritz 6, or if it was it had been
handicapped in some way.
|
| | |
Date: 26 Mar 2006 21:52:26
From: James
Subject: Re: Rybka vs. 1900 player
|
Ange1o DePa1ma wrote :
>
> 52...g5 a horrible move that simply loses without a fight. Shredder doesn't
> even think this is the 8th best move which, considering Black only has 10
> moves he can make, is interesting.
>
It depends on the time shredder9 "thinks" on this move (and the state of
the hash tables). If you start from a fresh position (load the game,
quit shredder, restart shredder and make it think directly on that
move), it will take around 60s to find the right move (Kf5), and g5
remains in the top list for quite some time (on a P4-D 3.2GHz).
As I previously wrote, some chess engines find the right move
immediately, probably because they have better chess knowledge regarding
king opposition. Some, like crafty, keeps g5 on top of the list for many
seconds.
(BTW: Kf5 is the only move to get a draw, all others lose)
|
| | | |
Date: 26 Mar 2006 16:39:29
From: Ange1o DePa1ma
Subject: Re: Rybka vs. 1900 player
|
"James" <chess123@alliot.org > wrote
> (BTW: Kf5 is the only move to get a draw, all others lose)
I think Black loses anyway. He can't cross his 4th rank. White ditches the d
pawn and easily outflanks.
|
| | | | |
Date: 27 Mar 2006 01:01:19
From: James
Subject: Analysis please (was Re: Rybka vs. 1900 player)
|
Ange1o DePa1ma wrote :
> "James" <chess123@alliot.org> wrote
>
>> (BTW: Kf5 is the only move to get a draw, all others lose)
>
> I think Black loses anyway. He can't cross his 4th rank. White ditches the d
> pawn and easily outflanks.
>
>
8/6p1/p4k1p/2pP4/2P5/6KP/P7/8 b - - 0 52
Well, I would be happy to see the winning continuation after 52...Kf6f5
as I can't find any:
53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
and position is a draw according to me
other continuations seem to lead to the same point
53. a3 g6 54. h4 h5 55. a4 a5
53. a4 a5 54. Kh4 g6 55. Kg3 Ke5
etc.
Anyone?
|
| | | | | |
Date: 26 Mar 2006 20:42:04
From: Ange1o DePa1ma
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
"James" <chess123@alliot.org > wrote
>>> (BTW: Kf5 is the only move to get a draw, all others lose)
>>
>> I think Black loses anyway. He can't cross his 4th rank. White ditches
>> the d pawn and easily outflanks.
>
> Well, I would be happy to see the winning continuation after 52...Kf6f5 as
> I can't find any:
> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
> and position is a draw according to me
> other continuations seem to lead to the same point
> 53. a3 g6 54. h4 h5 55. a4 a5
> 53. a4 a5 54. Kh4 g6 55. Kg3 Ke5
After 52...Kf5 53.h4 g6 54.Kf3 . Now we can get 54...a5 55.a4 out of the way
and then ...Kf6 56.Kf5. Black is in zuggy. 56...g5 57.hg hg 58.Kg4 Kg6.
White ditches the d pawn and easily outflanks and wins Black's pawns. If
56....Kf7 then 57. Ke5 wins as in ...Kf7 58.Kd5 and white reaches c7 or
queens long before the pawns get rolling. If ...Kd7 or ...Kd8 then White
goes to f6, winning both pawns and easily outflanking as in previous
variations.
|
| | | | | | |
Date: 27 Mar 2006 07:49:41
From: James
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
Ange1o DePa1ma a écrit :
> "James" <chess123@alliot.org> wrote
>
>>>> (BTW: Kf5 is the only move to get a draw, all others lose)
>>> I think Black loses anyway. He can't cross his 4th rank. White ditches
>>> the d pawn and easily outflanks.
>> Well, I would be happy to see the winning continuation after 52...Kf6f5 as
>> I can't find any:
>> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
>> and position is a draw according to me
>> other continuations seem to lead to the same point
>> 53. a3 g6 54. h4 h5 55. a4 a5
>> 53. a4 a5 54. Kh4 g6 55. Kg3 Ke5
>
> After 52...Kf5 53.h4 g6 54.Kf3 . Now we can get 54...a5 55.a4 out of the way
> and then ...Kf6 56.Kf5. Black is in zuggy. 56...g5 57.hg hg 58.Kg4 Kg6.
> White ditches the d pawn and easily outflanks and wins Black's pawns. If
> 56....Kf7 then 57. Ke5 wins as in ...Kf7 58.Kd5 and white reaches c7 or
> queens long before the pawns get rolling. If ...Kd7 or ...Kd8 then White
> goes to f6, winning both pawns and easily outflanking as in previous
> variations.
>
As stated in a previous post, the correct answer to 53.h4 seems to be h5
and not g6 (but I might be mistaken).
Your move?
|
| | | | | |
Date: 27 Mar 2006 00:43:27
From: Ron
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
In article <44271d39$0$18343$8fcfb975@news.wanadoo.fr >,
James <chess123@alliot.org > wrote:
> 8/6p1/p4k1p/2pP4/2P5/6KP/P7/8 b - - 0 52
>
> Well, I would be happy to see the winning continuation after 52...Kf6f5
> as I can't find any:
> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
52. ... Kf5 53. h4 g6 54. Kf3 Ke5 55. Kg4 a5 56. a4 Kf6 57. Kf4 Kf7 58.
Ke5 Ke7 59. d6+ Kd7 60. Kf6 g5 61. hxg5 hxg5 62. Kxg5 Kxd6 63. Kf6 Kc7
64. Ke7 Kc6 65.Ke6 Kc7 66. Kd5 Kb6 67. Kd6
looks like a reasonable winning attempt for white. There's no reason to
send the king to the h-file. White simply has to find a way to liquidate
the g- and h- pawns favorably. Since black can't stray too far from the
e-pawn, this isn't too difficult.
I'm not claiming the above line is definitive. It's just what I came up
with messing around.
-Ron
|
| | | | | | |
Date: 27 Mar 2006 03:32:58
From: James
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
Ron a écrit :
> In article <44271d39$0$18343$8fcfb975@news.wanadoo.fr>,
> James <chess123@alliot.org> wrote:
>
>> 8/6p1/p4k1p/2pP4/2P5/6KP/P7/8 b - - 0 52
>>
>> Well, I would be happy to see the winning continuation after 52...Kf6f5
>> as I can't find any:
>> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
>
> 52. ... Kf5 53. h4 g6 54. Kf3 Ke5 55. Kg4 a5 56. a4 Kf6 57. Kf4 Kf7 58.
> Ke5 Ke7 59. d6+ Kd7 60. Kf6 g5 61. hxg5 hxg5 62. Kxg5 Kxd6 63. Kf6 Kc7
> 64. Ke7 Kc6 65.Ke6 Kc7 66. Kd5 Kb6 67. Kd6
>
> looks like a reasonable winning attempt for white. There's no reason to
> send the king to the h-file. White simply has to find a way to liquidate
> the g- and h- pawns favorably. Since black can't stray too far from the
> e-pawn, this isn't too difficult.
>
> I'm not claiming the above line is definitive. It's just what I came up
> with messing around.
>
> -Ron
Interesting but what if 53...h5 instead of g6. Now g4 is blocked for the
king and h5 is blocked for the pawn. I might very well be missing
something, but how do you liquidate the g and h pawn?
|
| | | | | | | |
Date: 27 Mar 2006 18:10:05
From: Ron
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
In article <442740c3$0$20168$8fcfb975@news.wanadoo.fr >,
James <chess123@alliot.org > wrote:
> >> 8/6p1/p4k1p/2pP4/2P5/6KP/P7/8 b - - 0 52
> >>
> >> Well, I would be happy to see the winning continuation after 52...Kf6f5
> >> as I can't find any:
> >> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
> >
> > 52. ... Kf5 53. h4 g6 54. Kf3 Ke5 55. Kg4 a5 56. a4 Kf6 57. Kf4 Kf7 58.
> > Ke5 Ke7 59. d6+ Kd7 60. Kf6 g5 61. hxg5 hxg5 62. Kxg5 Kxd6 63. Kf6 Kc7
> > 64. Ke7 Kc6 65.Ke6 Kc7 66. Kd5 Kb6 67. Kd6
> Interesting but what if 53...h5 instead of g6. Now g4 is blocked for the
> king and h5 is blocked for the pawn. I might very well be missing
> something, but how do you liquidate the g and h pawn?
I don't like h5 because it turns g6 into a target. White can now win by
infiltrating his king over on the kingside and gobbling pawns, and he'll
toss the d-pawn to distract the black king to do so.
Again, not claiming this is definitive, but:
52. ... Kf5 53. h4 h5 54. Kf3 Ke5 55. Ke3 Kf5 56. Kd3 Ke5 57. Ke3 Kf5
58. d6 Ke6 59. d7 Kxd7 60. Kf4 Ke6 61. Ke4 Kd6 (61. ... Kf6 62. Kd5 Kf5
63. Kxc5) 62. Kf5 Kd7 63. Kg6 Ke6 64. Kxh5 Ke5 65. Kg6 Kd4 66. Kxg7 Kxc4
67. h5 Kb4 68. h6 c4 69.h7 c3 70. h8=Q c2
|
| | | | | | | | |
Date: 28 Mar 2006 00:30:17
From: James
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
Ron wrote :
> In article <442740c3$0$20168$8fcfb975@news.wanadoo.fr>,
> James <chess123@alliot.org> wrote:
>
>>>> 8/6p1/p4k1p/2pP4/2P5/6KP/P7/8 b - - 0 52
>>>>
>>>> Well, I would be happy to see the winning continuation after 52...Kf6f5
>>>> as I can't find any:
>>>> 53. Kh4 g6 54. Kg3 a5 55. h4 a4 56. a3 h5
>>> 52. ... Kf5 53. h4 g6 54. Kf3 Ke5 55. Kg4 a5 56. a4 Kf6 57. Kf4 Kf7 58.
>>> Ke5 Ke7 59. d6+ Kd7 60. Kf6 g5 61. hxg5 hxg5 62. Kxg5 Kxd6 63. Kf6 Kc7
>>> 64. Ke7 Kc6 65.Ke6 Kc7 66. Kd5 Kb6 67. Kd6
>
>> Interesting but what if 53...h5 instead of g6. Now g4 is blocked for the
>> king and h5 is blocked for the pawn. I might very well be missing
>> something, but how do you liquidate the g and h pawn?
>
> I don't like h5 because it turns g6 into a target. White can now win by
> infiltrating his king over on the kingside and gobbling pawns, and he'll
> toss the d-pawn to distract the black king to do so.
>
> Again, not claiming this is definitive, but:
>
> 52. ... Kf5 53. h4 h5 54. Kf3 Ke5 55. Ke3 Kf5 56. Kd3 Ke5 57. Ke3 Kf5
> 58. d6 Ke6 59. d7 Kxd7 60. Kf4 Ke6 61. Ke4 Kd6 (61. ... Kf6 62. Kd5 Kf5
> 63. Kxc5) 62. Kf5 Kd7 63. Kg6 Ke6 64. Kxh5 Ke5 65. Kg6 Kd4 66. Kxg7 Kxc4
> 67. h5 Kb4 68. h6 c4 69.h7 c3 70. h8=Q c2
Sorry wrong again...
Continuation 61 Ke4 Kf6 62. Kd5 loses immediately (61. Ke4 is a major
mistake, 62. Kf4 is slightly better than 62. Kd5 but loses anyway)
62... g5! 63. h4xg5 Kxg5 and white can't stop the black h pawn and loses
the race.
But however, 58. d6 Ke6 59. d7 Kxd7 60. Kf4 Ke6 is not winning. The
best line seems to be:
61. Kg5 a5 62. a4 Ke5 63. Kxh5 Kd4 64. Kg6 Kxc4 and both pawns (h white
and c black) are promoted, with a slightly better position for black
(but should be a draw anyway).
Any other idea?
James
|
| | | | | | | | | |
Date: 28 Mar 2006 00:56:24
From: Ron
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
In article <44286777$0$20145$8fcfb975@news.wanadoo.fr >,
James <chess123@alliot.org > wrote:
>
> Any other idea?
Maybe 53 h4 h5 54.Kf2 (since the black king can't go after the h-pawn);
You're right, I completely missed black's ability to play g5 in the line
I suggested.
I don't have time now to look at this idea in any detail. There are
certainly some tricks here for white, but it's complicated.
|
| | | | | | | | | | |
Date: 28 Mar 2006 04:18:30
From: James
Subject: Re: Analysis please (was Re: Rybka vs. 1900 player)
|
Ron a écrit :
> In article <44286777$0$20145$8fcfb975@news.wanadoo.fr>,
> James <chess123@alliot.org> wrote:
>
>> Any other idea?
>
> Maybe 53 h4 h5 54.Kf2 (since the black king can't go after the h-pawn);
Right it can't, but it hasn't to either
54...Ke5
and now we are back to where we started
55. a4 Kf6 56. Ke3 Ke5 and the position is blocked
55. Ke3 a5 56. a4 Kf5 idem
ad libitum...
> I don't have time now to look at this idea in any detail. There are
> certainly some tricks here for white, but it's complicated.
Interesting position indeed...
Thanks
|
| |
Date: 20 Mar 2006 14:26:04
From: irchans
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
RD wrote:
>
> I had similar results... Rybka 1.1 vs Shredder 9.1 with scores like
> 11 wins 2 loses and 7 draws for Rybka after 20 games. Rybka is clearly
> stronger.
>
> Rybka still needs improvements in some positions (especially end
> game). It's still a new program. I suspect after one or two years
> Rybka 2 or Rybka 3 might be 200-250+ points stronger than present-day
> Fritz and Shredder on the current hardware. That would be big jump.
Wow 200-250 points! Before Rybka, I would have said that it was nearly
impossible for computer chess software to improve more than 100 points
on the same hardware in a year. I am very curious about how Vasik
accomplished this. (I asked him in an email, but he replied to other
parts of the email and understandably avoided that question.) Lots of
other strong players have tried. Will we go back to the old 30 point
per year improvements? Will Vasik lead the way to 100 point
improvements every year so that in 2010 we have chess computers rated
3300 on 2006 hardware?
Cheers,
Irchans
PS: RD, did you use blitz time controls and the 32-bit Rybka 1.1?
|
| |
Date: 18 Mar 2006 16:09:55
From: CeeBee
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
"irchans" <infinitgames@yahoo.com > wrote in rec.games.chess.computer:
> --------quoted text--------------------
> "On this list
>
> http://www.husvankempen.de/nunn/cegtrating4040best.html
>
> How is Rybka beta 100 points higher than all professional engines?
>
> Isn't everything known about chess programming already implemented? I
> also heard that PC chess software have been getting better for the last
> 10 or 15 years, not because programming technique got better but
> because the hardware got faster. If true, how could Rybka suddenly
> popup and score 100-150 points higher than other programs on the same
> hardware?
>
> If Rybka programmer tried something new that no one else ever thought
> of, how is that possible given so many programmers have been working on
> their engines for so many years?"
> ----end quoted text--------------------
>
> What do you think?
I think the Rybka programmmer is a significantly stronger chess player than
the other "good" programmers. For a knowledge based program this could even
make more difference than 100 rating point.
Most programmers are good programmers who play a decent game of chess;
seldomly they're good chess players with good programming skills. Vasik
Rajlich has a 2300+ Elo and an IM title and he is a professional software
developer.
BTW, "professional" doesn't always mean "better"; the open source and
freeware society has shown us that much.
--
CeeBee
*** The Cookie Has Spoken ***
|
| |
Date: 18 Mar 2006 06:01:32
From: irchans
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Yes, I played both on the same PC.
The engines are restared every game and the black player is always
started first, so if you are right, then black should win more often
than expected.
I also turned "ponder" off. I think that means that the engine only
uses the cpu during its turn. The size of the hash tables ( Rybka 79
MB Shredder 71 MB) did not seem to depend on the order that engines
were started, but I will check that.
Thanks for the feedback!
Irchans
|
| |
Date: 18 Mar 2006 10:10:50
From: Terry
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
"irchans" <infinitgames@yahoo.com > wrote in message
news:1142637850.481509.35790@g10g2000cwb.googlegroups.com...
> At talkchess.com (http://216.25.93.108/forum/viewforum.php?f=2),
> oneguy asks the following question:
>
> --------quoted text--------------------
> "On this list
>
> http://www.husvankempen.de/nunn/cegtrating4040best.html
>
> How is Rybka beta 100 points higher than all professional engines?
>
> Isn't everything known about chess programming already implemented? I
> also heard that PC chess software have been getting better for the last
> 10 or 15 years, not because programming technique got better but
> because the hardware got faster. If true, how could Rybka suddenly
> popup and score 100-150 points higher than other programs on the same
> hardware?
>
> If Rybka programmer tried something new that no one else ever thought
> of, how is that possible given so many programmers have been working on
> their engines for so many years?"
> ----end quoted text--------------------
>
> What do you think?
>
> I have recently purchased Rybka v1.1.w32 and I ran my first Arena.exe
> tournament against Shredder 8. I was very impressed by the results.
>
> 108 wins for Rybka
> 30 wins for Shredder 8 UCI
> 62 ties.
>
> According to Arena, this corresponds to rating difference of about 144
> points! (There is still a lot of uncertainty with only 200 games. I
> think the uncertainty is about + or - 25 points.) The large difference
> in ratings might have been caused by the tournament conditions:
>
> Blitz: 2 min + 1 second per move,
> Rybka 79 MB
> Shredder 71 MB
> Windows XP, AMD 3200, 512 MB ram total.
> Opening book: mainbook.abk provided by Arena.
> Ponder Off, Adjudicate games after 250 moves.
> I'm not sure if any endgame tablebases were used by Shredder. Rybka
> did not use any table bases.
>
> (Is there any other information needed?)
>
> I set "Common hashtable size" to 64 MB, but that seemed to be ignored.
> Does anyone know how to force the hashtable size?
>
> I will try a new tournament with more time per move later.
>
> Cheers,
> Irchans
>
> PS: This is my first post to this group. Any advice would be
> appreciated.
>
Did you play both engines on the same PC.
Often the first one to run will steal the resources and the
2nd engine gets very little.
Regards
|
| | |
Date: 24 Mar 2006 07:34:18
From: Eric Hallsworth
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
I was interested to read the remark
>they still do not think nor necessarily play as world
>champions do
In view of the fact that the team of Hydra, Deep Junior, and Fritz (on a
Centrino laptop!) recently beat 3 ex World Champions (Ponomariov,
Khazimdzha and Khalifman) by 8-4, we should say 'World Champions
still do not think nor necessarily play like the top chess programs'?!
I typed the above with a :-) on my face
Eric
In article <442255b3$0$56211$ae4e5890@news.nationwide.net >, Johnny T
<not@home.com > writes
>Ange1o DePa1ma wrote:
>> Again, reality rears its ugly head. Johnny wrote:
>>
>> "Already you hear about people saying these programs play "differently"
>> than computer programs of the past. And I suspect we shall see more
>> "beauty" in the play, because we shall see strength in not being just
>> irrefutable, but strength based on the structure and knowledge of the
>> game."
>>
>> My challenge to Johnny (and any other computer skeptic) is to quantify
>> "beauty" in chess in any way other than "good move."
>
>Ok, I would like to return this thread to it's topic. Why would a
>engine be suddenly 100 points better?
>
>For a while there was a race for depth, because irrefutable moves as
>deep as possible was clearly the best way to build "strong" programs.
>But that the race for depth has serious drawbacks because it is a
>logarithmic problem, and we have reached a point where even the next ply
>is a long ways away on desktop machines.
>
>On the whim of those that threw money at the possible, the Deep Blue and
>Hydra teams showed that enough irrefutable depth could seriously
>challenge the champions of the world. And that strong observers of the
>game were amazed by some of the moves, and the control of the game the
>machines sometimes demonstrated. And it may be that with rather limited
>knowledge and sufficient depth that machines are indeed world champions.
> And yet they still do not think nor necessarily play as world
>champions do. There has been criticism of opponents of hydra that the
>games while strong seemed to have "weakness" in their play, that being
>irrefutable is not the same as having a positive strategy.
>
>But back in the world of desktop machines, we are years and years away
>of getting the next level of depth. But, we have more than enough
>power to get to the previous level of depth, and we have the luxury of
>excess computing power to utilize for the evaluation function. This is
>the source of the "extra 100 points", and the point being is that the
>moves are not just irrefutable within the computers horizon (which turns
>out to be deep enough to beat most people), but that the moves are also
>"better" resulting in both the 100 point gain, but a newly displayed
>aesthetic that appeals to human players, and validates years of theory
>on what good moves "look" like.
>
>This is the next level of computer play. Just as I thought computer
>chess would be quiescent for awhile because of the success of Hydra, I
>think it is clear that we are at the start of a new arms race. But
>more importantly for me, is what chessbase does to incorporate this into
>their "auto-fritz" functionality. I want a smarter, better analyzer
>that can bring more knowledge to a given game. For me, .2 pawns means
>nothing, understanding the structure, threats and options that are
>nearly as good mean more to improving my game.
--
With best wishes from Eric Hallsworth
45 Stretham Road, Wilburton, Cambs CB6 3RX, England
Publisher of Selective Search, the UK's only Computer Chess Magazine, est. 1985.
Computer Chess web pages+photos... http://www.elhchess.demon.co.uk
|
| | | |
Date: 25 Mar 2006 15:47:55
From: HD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Eric Hallsworth skrev:
> I was interested to read the remark
>
>> they still do not think nor necessarily play as world
>> champions do
>
> In view of the fact that the team of Hydra, Deep Junior, and Fritz (on a
> Centrino laptop!) recently beat 3 ex World Champions (Ponomariov,
> Khazimdzha and Khalifman) by 8-4, we should say 'World Champions
> still do not think nor necessarily play like the top chess programs'?!
>
> I typed the above with a :-) on my face
>
> Eric
>
Nice comment, Eric :)
|
| | | |
Date: 24 Mar 2006 15:14:30
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Eric Hallsworth" <eric@elhchess.demon.co.uk > wrote
>I was interested to read the remark
>
>>they still do not think nor necessarily play as world
>>champions do
>
> In view of the fact that the team of Hydra, Deep Junior, and Fritz (on a
> Centrino laptop!) recently beat 3 ex World Champions (Ponomariov,
> Khazimdzha and Khalifman) by 8-4, we should say 'World Champions
> still do not think nor necessarily play like the top chess programs'?!
If this newsgroup were active you'd get all sorts of crap about those three
not being real world champions, and that the humans didn't really take this
match seriously. We forget that with half a million dollars on the line the
best two human players in the world could not muster up the "beauty,"
"smartnes" or "goodness" to beat off the shelf programs operating from
desktops. (But somehow we're supposed to believe that 1900 players could
have).
All this talk about "beauty," "smart," "good" goes away if you just look at
the games and forget who (or what) is playing them. If you taught an alien
how to play chess and then asked them which entity is stronger today, the
best computers or the best humans, they would not even hesitate to answer
that it's the former.
|
| | | | |
Date: 24 Mar 2006 21:06:24
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Fri, 24 Mar 2006 15:14:30 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>(But somehow we're supposed to believe that 1900 players could
>have).
Some 1900 that were real good at beating comps used to beat them
regularly; some of them still can beat them once in a while .. like
this one 2 days ago
[Event "Blitz:5'+5""]
[Site "Medellin"]
[Date "2006.03.22"]
[Round "?"]
[White "Restrepo, Pablo"]
[Black "Fritz 8"]
[Result "1-0"]
[ECO "E14"]
[Annotator "Restrepo,Pablo"]
[PlyCount "143"]
[TimeControl "300+5"]
{70MB, Fritz8.ctg, YOUR-D1ECA258F0} 1. d4 {[%emt 0:00:21]} Nf6 {[%emt
0:00:00]}
2. e3 {[%emt 0:00:04]} c5 {[%emt 0:00:00]} 3. Nf3 {[%emt 0:00:06]} e6
{
[%emt 0:00:00]} 4. c4 {[%emt 0:00:07]} b6 {[%emt 0:00:00]} 5. Nc3 {
[%emt 0:00:17]} d6 {[%emt 0:00:00]} 6. Be2 {[%emt 0:00:27]} Be7 {[%emt
0:00:32]
} 7. O-O {[%emt 0:00:15]} O-O {[%emt 0:00:18]} 8. d5 {[%emt 0:00:52]}
a6 {
[%emt 0:00:51]} 9. e4 {[%emt 0:00:16]} Nbd7 {[%emt 0:00:21]} 10. b3 {
[%emt 0:00:28]} exd5 {[%emt 0:00:32]} 11. exd5 {[%emt 0:00:26]} Ne5 {
[%emt 0:00:00]} 12. Nxe5 {[%emt 0:00:15]} dxe5 {[%emt 0:00:04]} 13.
Qc2 {
[%emt 0:00:29]} Re8 {[%emt 0:00:16]} 14. Bb2 {[%emt 0:00:38]} Bd6 {
[%emt 0:00:20]} 15. Rad1 {[%emt 0:00:27]} e4 {[%emt 0:00:19]} 16. h3 {
[%emt 0:00:28]} Be5 {[%emt 0:00:16]} 17. Nb1 {[%emt 0:00:52]} Bf4 {
[%emt 0:00:32]} 18. Bc1 {[%emt 0:00:10]} Bc7 {[%emt 0:00:25]} 19. f4 {
[%emt 0:02:12]} exf3 {[%emt 0:00:00]} 20. Bxf3 {[%emt 0:00:06]} Be5 {
[%emt 0:00:33]} 21. Bb2 {[%emt 0:00:20]} Bxb2 {[%emt 0:00:20]} 22.
Qxb2 {
[%emt 0:00:08]} Qd6 {[%emt 0:00:08]} 23. Qf2 {[%emt 0:00:35]} Bf5 {
[%emt 0:00:00]} 24. Nd2 {[%emt 0:00:31]} Re5 {[%emt 0:00:14]} 25. Rfe1
{
[%emt 0:00:05]} Rae8 {[%emt 0:00:17]} 26. Rxe5 {[%emt 0:00:16]} Rxe5 {
[%emt 0:00:00]} 27. Re1 {[%emt 0:00:07]} Rxe1+ {[%emt 0:00:36]} 28.
Qxe1 {
[%emt 0:00:04]} h6 {[%emt 0:00:21]} 29. Qe3 {[%emt 0:01:13]} Bc2 {
[%emt 0:00:31]} 30. Qc3 {[%emt 0:00:19]} Bg6 {[%emt 0:00:18]} 31. Qe3
{
[%emt 0:00:06]} Kh8 {[%emt 0:00:10]} 32. Kf2 {[%emt 0:00:39]} Bf5 {
[%emt 0:00:12]} 33. Kg1 {[%emt 0:00:06]} Qg3 {[%emt 0:00:17]} 34. Nf1
{
[%emt 0:00:32]} Qc7 {[%emt 0:00:15]} 35. Nd2 {[%emt 0:00:05]} Bg6 {
[%emt 0:00:10]} 36. Kf2 {[%emt 0:00:08]} Kh7 {[%emt 0:00:11]} 37. Kg1
{
[%emt 0:00:05]} Qd6 {[%emt 0:00:08]} 38. Kf2 {[%emt 0:00:10]} Bc2 {
[%emt 0:00:14]} 39. Qc3 {[%emt 0:00:06]} Bf5 {[%emt 0:00:24]} 40. Qe3
{
[%emt 0:00:05]} Bd7 {[%emt 0:00:14]} 41. Kg1 {[%emt 0:00:10]} Kg8 {
[%emt 0:00:16]} 42. Ne4 {[%emt 0:00:05]} Nxe4 {[%emt 0:00:34]} 43.
Qxe4 {
[%emt 0:00:08]} f5 {[%emt 0:00:39]} 44. Qe3 {[%emt 0:00:36]} f4 {[%emt
0:00:36]
} 45. Qe2 {[%emt 0:00:19]} Kf7 {[%emt 0:00:00]} 46. Bg4 {[%emt
0:00:11]} Bxg4 {
[%emt 0:00:12]} 47. Qxg4 {[%emt 0:00:03]} b5 {[%emt 0:00:11]} 48. Kf2
{
[%emt 0:00:25]} bxc4 {[%emt 0:00:00]} 49. bxc4 {[%emt 0:00:07]} Kf6 {
[%emt 0:00:17]} 50. g3 {[%emt 0:00:07]} fxg3+ {[%emt 0:00:15]} 51.
Qxg3 {
[%emt 0:00:00]} Qxg3+ {[%emt 0:00:11]} 52. Kxg3 {[%emt 0:00:00]} g5 {
[%emt 0:00:11]} 53. h4 {[%emt 0:00:12]} a5 {[%emt 0:00:01]} 54. a4 {
[%emt 0:00:06]} Kf5 {[%emt 0:00:16]} 55. hxg5 {[%emt 0:03:07]} Kxg5 {
[%emt 0:00:00]} 56. Kh3 {[%emt 0:02:45]} h5 {[%emt 0:00:00]} 57. Kg3 {
[%emt 0:00:04]} Kf5 {[%emt 0:00:28]} 58. Kh4 {[%emt 0:00:15]} Kg6 {
[%emt 0:00:14]} 59. d6 {[%emt 0:00:21]} Kf6 {[%emt 0:00:10]} 60. Kxh5
{
[%emt 0:00:07]} Ke6 {[%emt 0:00:11]} 61. Kg5 {[%emt 0:00:05]} Ke5 {
[%emt 0:00:28]} (61... Kxd6 {[%emt 0:00:00]} 62. Kf5 {[%emt 0:00:00]}
Kd7 {
[%emt 0:00:00]} 63. Ke5 {[%emt 0:00:00]} Kc6 {[%emt 0:00:03]} 64. Ke6
{
[%emt 0:00:03]} Kc7 {[%emt 0:00:01]} 65. Kd5 {[%emt 0:00:01]} Kb6 {
[%emt 0:00:01]} 66. Kd6 {[%emt 0:00:01]} Kb7 {[%emt 0:00:01]} 67. Kxc5
{
[%emt 0:00:01]} Ka6 {[%emt 0:00:03]} 68. Kc6 {[%emt 0:00:00]} Ka7 {
[%emt 0:00:01]} 69. c5 {[%emt 0:00:00]} Ka6 {[%emt 0:00:01]} 70. Kd6 {
[%emt 0:00:00]} Kb7 {[%emt 0:00:01]} 71. c6+ {[%emt 0:00:00]} Kc8 {
[%emt 0:00:01]} 72. c7 {[%emt 0:00:00]} Kb7 {[%emt 0:00:01]} 73. Kd7 {
[%emt 0:00:01]} Ka6 {[%emt 0:00:01]} 74. c8=Q+ {[%emt 0:00:01]}) 62.
d7 {
[%emt 0:00:20]} Kd4 {[%emt 0:00:20]} 63. d8=Q+ {[%emt 0:00:02]} Kc3 {
[%emt 0:00:10]} 64. Qxa5+ {[%emt 0:00:05]} Kd4 {[%emt 0:00:05]} 65.
Qb5 {
[%emt 0:00:01]} Kd3 {[%emt 0:00:01]} 66. a5 {[%emt 0:00:07]} Ke4 {
[%emt 0:00:01]} 67. a6 {[%emt 0:00:00]} Kd3 {[%emt 0:00:02]} 68. a7 {
[%emt 0:00:00]} Ke3 {[%emt 0:00:01]} 69. a8=Q {[%emt 0:00:00]} Kf2 {
[%emt 0:00:00]} 70. Qa3 {[%emt 0:00:03]} Kg1 {[%emt 0:00:00]} 71. Qbb2
{
[%emt 0:00:01]} Kf1 {[%emt 0:00:01]} 72. Qaa1# {[%emt 0:00:01]} 1-0
>All this talk about "beauty," "smart," "good" goes away if you just look at
>the games and forget who (or what) is playing them. If you taught an alien
>how to play chess and then asked them which entity is stronger today, the
>best computers or the best humans, they would not even hesitate to answer
>that it's the former.
Again, computers are stronger than humans because they can calculate
4/5 moves perfectly. That's about it. This 'brute force' technique
only works in chess because the branches are fewer. In games like 'Go'
that have larger branches, computers are very weak compared to humans.
There is no hope in foreseeable future that computers will ever beat
'Go' masters using techniques like 'brute force' (counting beans
basically). I would like to see better quality and smarter programs
(not bean counters).
|
| | | | | |
Date: 27 Mar 2006 13:34:20
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
RD <fdgldfj@hotmails.com > wrote:
> "Ange1o DePa1ma" <angelodpnospam@nospam.gmail.com> wrote:
>> (But somehow we're supposed to believe that 1900 players could
>> have).
>
> Some 1900 that were real good at beating comps used to beat them
> regularly; some of them still can beat them once in a while .. like
> this one 2 days ago
By `1900 player' do you mean 1900 OTB FIDE / national federation Elo
or 1900 on whatever server the games were played on? If it's a server
rating, we can assume the computers are rated 2500+ and this suggests
that the computers beat him much more than he beats them.
Dave.
--
David Richerby Erotic Gnome (TM): it's like a smiling
www.chiark.greenend.org.uk/~davidr/ garden ornament but it's genuinely
erotic!
|
| | | | | |
Date: 24 Mar 2006 23:33:08
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
"RD" <fdgldfj@hotmails.com > wrote
> [Event "Blitz:5'+5""]
> [Site "Medellin"]
> [Date "2006.03.22"]
> [Round "?"]
> [White "Restrepo, Pablo"]
> [Black "Fritz 8"]
> [Result "1-0"]
> [ECO "E14"]
> [Annotator "Restrepo,Pablo"]
> [PlyCount "143"]
> [TimeControl "300+5"]
>
> {70MB, Fritz8.ctg, YOUR-D1ECA258F0} 1. d4 {[%emt 0:00:21]} Nf6 {[%emt
> 0:00:00]}
> 2. e3 {[%emt 0:00:04]} c5 {[%emt 0:00:00]} 3. Nf3 {[%emt 0:00:06]} e6
> {
> [%emt 0:00:00]} 4. c4 {[%emt 0:00:07]} b6 {[%emt 0:00:00]} 5. Nc3 {
> [%emt 0:00:17]} d6 {[%emt 0:00:00]} 6. Be2 {[%emt 0:00:27]} Be7 {[%emt
> 0:00:32]
> } 7. O-O {[%emt 0:00:15]} O-O {[%emt 0:00:18]} 8. d5 {[%emt 0:00:52]}
> a6 {
> [%emt 0:00:51]} 9. e4 {[%emt 0:00:16]} Nbd7 {[%emt 0:00:21]} 10. b3 {
> [%emt 0:00:28]} exd5 {[%emt 0:00:32]} 11. exd5 {[%emt 0:00:26]} Ne5 {
> [%emt 0:00:00]} 12. Nxe5 {[%emt 0:00:15]} dxe5 {[%emt 0:00:04]} 13.
> Qc2 {
> [%emt 0:00:29]} Re8 {[%emt 0:00:16]} 14. Bb2 {[%emt 0:00:38]} Bd6 {
> [%emt 0:00:20]} 15. Rad1 {[%emt 0:00:27]} e4 {[%emt 0:00:19]} 16. h3 {
> [%emt 0:00:28]} Be5 {[%emt 0:00:16]} 17. Nb1 {[%emt 0:00:52]} Bf4 {
> [%emt 0:00:32]} 18. Bc1 {[%emt 0:00:10]} Bc7 {[%emt 0:00:25]} 19. f4 {
> [%emt 0:02:12]} exf3 {[%emt 0:00:00]} 20. Bxf3 {[%emt 0:00:06]} Be5 {
> [%emt 0:00:33]} 21. Bb2 {[%emt 0:00:20]} Bxb2 {[%emt 0:00:20]} 22.
> Qxb2 {
> [%emt 0:00:08]} Qd6 {[%emt 0:00:08]} 23. Qf2 {[%emt 0:00:35]} Bf5 {
> [%emt 0:00:00]} 24. Nd2 {[%emt 0:00:31]} Re5 {[%emt 0:00:14]} 25. Rfe1
> {
> [%emt 0:00:05]} Rae8 {[%emt 0:00:17]} 26. Rxe5 {[%emt 0:00:16]} Rxe5 {
> [%emt 0:00:00]} 27. Re1 {[%emt 0:00:07]} Rxe1+ {[%emt 0:00:36]} 28.
> Qxe1 {
> [%emt 0:00:04]} h6 {[%emt 0:00:21]} 29. Qe3 {[%emt 0:01:13]} Bc2 {
> [%emt 0:00:31]} 30. Qc3 {[%emt 0:00:19]} Bg6 {[%emt 0:00:18]} 31. Qe3
> {
> [%emt 0:00:06]} Kh8 {[%emt 0:00:10]} 32. Kf2 {[%emt 0:00:39]} Bf5 {
> [%emt 0:00:12]} 33. Kg1 {[%emt 0:00:06]} Qg3 {[%emt 0:00:17]} 34. Nf1
> {
> [%emt 0:00:32]} Qc7 {[%emt 0:00:15]} 35. Nd2 {[%emt 0:00:05]} Bg6 {
> [%emt 0:00:10]} 36. Kf2 {[%emt 0:00:08]} Kh7 {[%emt 0:00:11]} 37. Kg1
> {
> [%emt 0:00:05]} Qd6 {[%emt 0:00:08]} 38. Kf2 {[%emt 0:00:10]} Bc2 {
> [%emt 0:00:14]} 39. Qc3 {[%emt 0:00:06]} Bf5 {[%emt 0:00:24]} 40. Qe3
> {
> [%emt 0:00:05]} Bd7 {[%emt 0:00:14]} 41. Kg1 {[%emt 0:00:10]} Kg8 {
> [%emt 0:00:16]} 42. Ne4 {[%emt 0:00:05]} Nxe4 {[%emt 0:00:34]} 43.
> Qxe4 {
> [%emt 0:00:08]} f5 {[%emt 0:00:39]} 44. Qe3 {[%emt 0:00:36]} f4 {[%emt
> 0:00:36]
> } 45. Qe2 {[%emt 0:00:19]} Kf7 {[%emt 0:00:00]} 46. Bg4 {[%emt
> 0:00:11]} Bxg4 {
> [%emt 0:00:12]} 47. Qxg4 {[%emt 0:00:03]} b5 {[%emt 0:00:11]} 48. Kf2
> {
> [%emt 0:00:25]} bxc4 {[%emt 0:00:00]} 49. bxc4 {[%emt 0:00:07]} Kf6 {
> [%emt 0:00:17]} 50. g3 {[%emt 0:00:07]} fxg3+ {[%emt 0:00:15]} 51.
> Qxg3 {
> [%emt 0:00:00]} Qxg3+ {[%emt 0:00:11]} 52. Kxg3 {[%emt 0:00:00]} g5 {
> [%emt 0:00:11]} 53. h4 {[%emt 0:00:12]} a5 {[%emt 0:00:01]} 54. a4 {
> [%emt 0:00:06]} Kf5 {[%emt 0:00:16]} 55. hxg5 {[%emt 0:03:07]} Kxg5 {
> [%emt 0:00:00]} 56. Kh3 {[%emt 0:02:45]} h5 {[%emt 0:00:00]} 57. Kg3 {
> [%emt 0:00:04]} Kf5 {[%emt 0:00:28]} 58. Kh4 {[%emt 0:00:15]} Kg6 {
> [%emt 0:00:14]} 59. d6 {[%emt 0:00:21]} Kf6 {[%emt 0:00:10]} 60. Kxh5
> {
> [%emt 0:00:07]} Ke6 {[%emt 0:00:11]} 61. Kg5 {[%emt 0:00:05]} Ke5 {
> [%emt 0:00:28]} (61... Kxd6 {[%emt 0:00:00]} 62. Kf5 {[%emt 0:00:00]}
> Kd7 {
> [%emt 0:00:00]} 63. Ke5 {[%emt 0:00:00]} Kc6 {[%emt 0:00:03]} 64. Ke6
> {
> [%emt 0:00:03]} Kc7 {[%emt 0:00:01]} 65. Kd5 {[%emt 0:00:01]} Kb6 {
> [%emt 0:00:01]} 66. Kd6 {[%emt 0:00:01]} Kb7 {[%emt 0:00:01]} 67. Kxc5
> {
> [%emt 0:00:01]} Ka6 {[%emt 0:00:03]} 68. Kc6 {[%emt 0:00:00]} Ka7 {
> [%emt 0:00:01]} 69. c5 {[%emt 0:00:00]} Ka6 {[%emt 0:00:01]} 70. Kd6 {
> [%emt 0:00:00]} Kb7 {[%emt 0:00:01]} 71. c6+ {[%emt 0:00:00]} Kc8 {
> [%emt 0:00:01]} 72. c7 {[%emt 0:00:00]} Kb7 {[%emt 0:00:01]} 73. Kd7 {
> [%emt 0:00:01]} Ka6 {[%emt 0:00:01]} 74. c8=Q+ {[%emt 0:00:01]}) 62.
> d7 {
> [%emt 0:00:20]} Kd4 {[%emt 0:00:20]} 63. d8=Q+ {[%emt 0:00:02]} Kc3 {
> [%emt 0:00:10]} 64. Qxa5+ {[%emt 0:00:05]} Kd4 {[%emt 0:00:05]} 65.
> Qb5 {
> [%emt 0:00:01]} Kd3 {[%emt 0:00:01]} 66. a5 {[%emt 0:00:07]} Ke4 {
> [%emt 0:00:01]} 67. a6 {[%emt 0:00:00]} Kd3 {[%emt 0:00:02]} 68. a7 {
> [%emt 0:00:00]} Ke3 {[%emt 0:00:01]} 69. a8=Q {[%emt 0:00:00]} Kf2 {
> [%emt 0:00:00]} 70. Qa3 {[%emt 0:00:03]} Kg1 {[%emt 0:00:00]} 71. Qbb2
> {
> [%emt 0:00:01]} Kf1 {[%emt 0:00:01]} 72. Qaa1# {[%emt 0:00:01]} 1-0
This game is a farce. I don't know where you got it from. I played through
the first 30 moves using Shredder 7 in infinite analysis mode, with six
lines of analysis and 20 seconds per move. With all the cheese in there I
couldn't bear to play through more of this game. But the following table
summarizes my results. The number before the period is the move number, and
the number after is the ranking my computer gave your computer's moves. The
designation >6 means your move was not included in my computer's six lines
(in other words, it was a lemon). An asterisk after the move is an obvious
move or capture that a 900-rated player would make:
1. 1
2. 4
3. 2
4.2
5.5
6.3
7.1
8. >6
9.3
10. >6
11. >6
12.1*
13.4
14.1
15.1
16. >6
17. >6
18.1
19. >6
20.4
21.1*
22.1
23.1
24.1
25.1*
26.1*
27.3
28.5
29. >6
30.1*
I realize that different engines often give different moves, but this
"Fritz" played more moves that were off the chart (in a negative way) than I
normally play during a game. I must conclude that the engine was either
handicapped during this game, or that the White player was getting help, or
that this game is a total fabrication.
Your experience, of 1900 players regularly beating computers, is so far out
of the ordinary that I must conclude it is fantasy. I am a 1900 player and I
have played thousands of games against Shredder and Fritz, and I have never
one a single game.
If there are any other 1900 players out there with similar experiences, I
would love to hear from you.
|
| | | | | | |
Date: 25 Mar 2006 07:14:07
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Fri, 24 Mar 2006 23:33:08 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>The number before the period is the move number, and
>the number after is the ranking my computer gave your computer's moves. The
>designation >6 means your move was not included in my computer's six lines
>(in other words, it was a lemon). An asterisk after the move is an obvious
>move or capture that a 900-rated player would make:
>
>1. 1
>2. 4
>3. 2
>4.2
>5.5
>6.3
>7.1
>8.>6
If a move is not on this first 6 choices of your shredder 7, then
that must be a lemon? LOL. I am beginning to doubt that you are even a
chess player. Computers are not gods. They really do play some
positions horribly. Anyway, the 8th move was a6 by Fritz 8 in this
position. I don't have Fritz 8, but I just checked on Fritz 9 and a6
is the first choice by Fritz 9 on move 8.
Anyone else who has Fritz 8 or 9 can confirm that.
|
| | | | | | | |
Date: 25 Mar 2006 16:06:08
From: James
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
RD wrote :
> On Fri, 24 Mar 2006 23:33:08 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>> The number before the period is the move number, and
>> the number after is the ranking my computer gave your computer's moves. The
>> designation >6 means your move was not included in my computer's six lines
>> (in other words, it was a lemon). An asterisk after the move is an obvious
>> move or capture that a 900-rated player would make:
>>
>> 1. 1
>> 2. 4
>> 3. 2
>> 4.2
>> 5.5
>> 6.3
>> 7.1
>> 8.>6
>
> If a move is not on this first 6 choices of your shredder 7, then
> that must be a lemon? LOL. I am beginning to doubt that you are even a
> chess player. Computers are not gods. They really do play some
> positions horribly. Anyway, the 8th move was a6 by Fritz 8 in this
> position. I don't have Fritz 8, but I just checked on Fritz 9 and a6
> is the first choice by Fritz 9 on move 8.
>
> Anyone else who has Fritz 8 or 9 can confirm that.
>
Right.
However, Fritz8 was not set at its best regarding the opening book.
4...b6 is marked in Fritz8 book as a serious mistake and shouldn't have
been played at its strongest settings. Moreover Fritz8 book lists
8...e5, so there was no reason to play 8...a6.
I replayed the entire game and Fritz8 disagrees with the moves quite
often. I tried different time controls, and the discrepancy remains. It
would be interesting to know the time controls of the actual game.
However such discrepancies could come from a different way of setting
hash table parameters.
There are some moves which are quite strange. For example,
49...Kf7f6 (+1.32) is bad and g7g5 (+0.63) is found in only 4s
52...g7g5 (+3.60) is a huge mistake. Kf5 (+0.90) is correct and the game
is probably a draw, the white king being unable to penetrate the black
position.
When starting from a "fresh" position, the engine needs 150s to find
Kf5. This time can be reduced depending on the use of hash tables and
the time spent searching past positions to fill these tables. In
"standard conditions", Kf5 is found in around 15s.
Kf5 is found (from scratch) in 60s by Shredder9, and instantaneously by
Rybka beta, Fruit2.1 or Glaurung Mainz.
After g7g5, the game is lost but there are other strange things such as:
53...a5 which is a way to lose quicker and last (but not least)
61...Ke5 which is a very strange and huge "beans counting" blunder that
no engine could ever make (correct is Kd7 or at least Kxd6) except in
highly abnormal situations (extremely short time controls for example).
Fritz8 might have played this game, but probably not with its stronger
settings (book, hash tables, time controls).
And I am really impressed by its opponent, because time controls were
probably extremely short, which makes things highly difficult for
humans, who are, as you previously said, prone to make tactical
blunders, especially under stress...
|
| | | | | | | | |
Date: 25 Mar 2006 10:34:55
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Unless we know the state of the computer and can certify that the human did
not receive help, this game is scientifically meaningless. There are many
different ways to manipulate server games, as anyone who has played 35,000
(as I have) of them can attest.
I stand by my statement that 1900 players cannot beat un-handicapped,
out-of-the-box, top chess engines and have not been able to do so regularly
for many years. This is so clearly a matter of fact that any "evidence" to
the contrary cannot be taken seriously. If you really believe a guy in Chile
who I can beat half the time can also beat Fritz, I have a nice little
bridge I'd like to sell you.
"James" <chess123@alliot.org > wrote in message
news:44255c5c$0$19709$8fcfb975@news.wanadoo.fr...
> RD wrote :
>> On Fri, 24 Mar 2006 23:33:08 -0500, "Ange1o DePa1ma"
>> <angelodpnospam@nospam.gmail.com> wrote:
>>
>>> The number before the period is the move number, and the number after is
>>> the ranking my computer gave your computer's moves. The designation >6
>>> means your move was not included in my computer's six lines (in other
>>> words, it was a lemon). An asterisk after the move is an obvious move or
>>> capture that a 900-rated player would make:
>>>
>>> 1. 1
>>> 2. 4
>>> 3. 2
>>> 4.2
>>> 5.5
>>> 6.3
>>> 7.1
>>> 8.>6
>>
>> If a move is not on this first 6 choices of your shredder 7, then
>> that must be a lemon? LOL. I am beginning to doubt that you are even a
>> chess player. Computers are not gods. They really do play some
>> positions horribly. Anyway, the 8th move was a6 by Fritz 8 in this
>> position. I don't have Fritz 8, but I just checked on Fritz 9 and a6
>> is the first choice by Fritz 9 on move 8. Anyone else who has Fritz 8 or
>> 9 can confirm that.
>
> Right.
>
> However, Fritz8 was not set at its best regarding the opening book. 4...b6
> is marked in Fritz8 book as a serious mistake and shouldn't have been
> played at its strongest settings. Moreover Fritz8 book lists 8...e5, so
> there was no reason to play 8...a6.
>
> I replayed the entire game and Fritz8 disagrees with the moves quite
> often. I tried different time controls, and the discrepancy remains. It
> would be interesting to know the time controls of the actual game. However
> such discrepancies could come from a different way of setting hash table
> parameters.
>
> There are some moves which are quite strange. For example,
> 49...Kf7f6 (+1.32) is bad and g7g5 (+0.63) is found in only 4s
>
> 52...g7g5 (+3.60) is a huge mistake. Kf5 (+0.90) is correct and the game
> is probably a draw, the white king being unable to penetrate the black
> position.
> When starting from a "fresh" position, the engine needs 150s to find Kf5.
> This time can be reduced depending on the use of hash tables and the time
> spent searching past positions to fill these tables. In "standard
> conditions", Kf5 is found in around 15s.
> Kf5 is found (from scratch) in 60s by Shredder9, and instantaneously by
> Rybka beta, Fruit2.1 or Glaurung Mainz.
>
> After g7g5, the game is lost but there are other strange things such as:
> 53...a5 which is a way to lose quicker and last (but not least) 61...Ke5
> which is a very strange and huge "beans counting" blunder that no engine
> could ever make (correct is Kd7 or at least Kxd6) except in highly
> abnormal situations (extremely short time controls for example).
>
> Fritz8 might have played this game, but probably not with its stronger
> settings (book, hash tables, time controls).
>
> And I am really impressed by its opponent, because time controls were
> probably extremely short, which makes things highly difficult for humans,
> who are, as you previously said, prone to make tactical blunders,
> especially under stress...
|
| | | | | | | | | |
Date: 25 Mar 2006 18:00:06
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Sat, 25 Mar 2006 10:34:55 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>I stand by my statement that 1900 players cannot beat un-handicapped,
>out-of-the-box, top chess engines and have not been able to do so regularly
>for many years. This is so clearly a matter of fact that any "evidence" to
>the contrary cannot be taken seriously. If you really believe a guy in Chile
>who I can beat half the time can also beat Fritz, I have a nice little
>bridge I'd like to sell you.
Many years? How about the book "Chess: A Psychiatrist Matches Wits
with Fritz"
http://www.almudo.com/cgi-bin/chess.cgi?item_id=1929331045&search_type=AsinSearch&locale=us
There are over 100 games of below 2000 player beating Fritz 6
consistently.
|
| | | | | | |
Date: 25 Mar 2006 06:55:19
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Fri, 24 Mar 2006 23:33:08 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>This game is a farce. I don't know where you got it from. I played through
>the first 30 moves using Shredder 7 in infinite analysis mode, with six
>lines of analysis and 20 seconds per move. With all the cheese in there I
>couldn't bear to play through more of this game. But the following table
>summarizes my results. The number before the period is the move number, and
>the number after is the ranking my computer gave your computer's moves. The
>designation >6 means your move was not included in my computer's six lines
>(in other words, it was a lemon). An asterisk after the move is an obvious
>move or capture that a 900-rated player would make:
The game is not 'farce.' The game is by Pablo who regularly plays
against commuters on playchess server (including a couple of draws
against Hydra last year as I remember). Many people on this forum
probably know him from CCC forum. The program used in this game was
Fritz 8, and you were using Shredder 7 -- that might explain why your
comp chooses different moves. Here is the game again after clean up
(you can simply copy and past the score to Shredder GUI).
Blitz:5'+5
1. d4 Nf6 2. e3 c5 3. Nf3 e6 4. c4 b6 5. Nc3 d6 6. Be2 Be7 7. O-O O-O
8. d5 a6 9. e4 Nbd7 10. b3 exd5 11. exd5 Ne5 12. Nxe5 dxe5 13. Qc2
Re8 14. Bb2 Bd6 15. Rad1 e4 16. h3 Be5 17. Nb1 Bf4 18. Bc1 Bc7 19. f4
exf3 20. Bxf3 Be5 21. Bb2 Bxb2 22. Qxb2 Qd6 23. Qf2 Bf5 24. Nd2 Re5
25. Rfe1 Rae8 26. Rxe5 Rxe5 27. Re1 Rxe1+ 28. Qxe1 h6 29. Qe3 Bc2
30. Qc3 Bg6 31. Qe3 Kh8 32. Kf2 Bf5 33. Kg1 Qg3 34. Nf1 Qc7 35. Nd2
Bg6 36. Kf2 Kh7 37. Kg1 Qd6 38. Kf2 Bc2 39. Qc3 Bf5 40. Qe3 Bd7 41.
Kg1 Kg8 42. Ne4 Nxe4 43. Qxe4 f5 44. Qe3 f4 45. Qe2 Kf7 46. Bg4 Bxg4
47. Qxg4 b5 48. Kf2 bxc4 49. bxc4 Kf6 50. g3 fxg3+ 51. Qxg3 Qxg3+ 52.
Kxg3 g5 53. h4 a5 54. a4 Kf5 55. hxg5 Kxg5 56. Kh3 h5 57. Kg3 Kf5 58.
Kh4 Kg6 59. d6 Kf6 60. Kxh5 Ke6 61. Kg5 Ke5 Kxd6 62. Kf5 Kd7 63. Ke5
Kc6 64. Ke6 Kc7 65. Kd5 Kb6 66. Kd6 Kb7 67. Kxc5 Ka6 68. Kc6Ka7 69. c5
Ka6 70. Kd6 Kb7 71. c6+ Kc8 72. c7 Kb7 73. Kd7 Ka6 74. c8=Q+ 62. d7
Kd4 63. d8=Q+ Kc3 64. Qxa5+ Kd4 65. Qb5 Kd3 66. a5 Ke4 67. a6 Kd3
68. a7 Ke3 69. a8=Q Kf2 70. Qa3 Kg1 71. Qbb2 Kf1 72. Qaa1# 0-1
|
| | | | | | | |
Date: 25 Mar 2006 09:17:19
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Thanks. I do not know how to get those moves into the program without
entering them manually. I tried saving in notepad as .pgn but that didn't
work. Help?
"RD" <fdgldfj@hotmails.com > wrote in message
news:67p922pgbeo718echk8ie2dprkp41jlbpc@4ax.com...
> On Fri, 24 Mar 2006 23:33:08 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>>This game is a farce. I don't know where you got it from. I played through
>>the first 30 moves using Shredder 7 in infinite analysis mode, with six
>>lines of analysis and 20 seconds per move. With all the cheese in there I
>>couldn't bear to play through more of this game. But the following table
>>summarizes my results. The number before the period is the move number,
>>and
>>the number after is the ranking my computer gave your computer's moves.
>>The
>>designation >6 means your move was not included in my computer's six lines
>>(in other words, it was a lemon). An asterisk after the move is an obvious
>>move or capture that a 900-rated player would make:
>
> The game is not 'farce.' The game is by Pablo who regularly plays
> against commuters on playchess server (including a couple of draws
> against Hydra last year as I remember). Many people on this forum
> probably know him from CCC forum. The program used in this game was
> Fritz 8, and you were using Shredder 7 -- that might explain why your
> comp chooses different moves. Here is the game again after clean up
> (you can simply copy and past the score to Shredder GUI).
>
> Blitz:5'+5
>
> 1. d4 Nf6 2. e3 c5 3. Nf3 e6 4. c4 b6 5. Nc3 d6 6. Be2 Be7 7. O-O O-O
> 8. d5 a6 9. e4 Nbd7 10. b3 exd5 11. exd5 Ne5 12. Nxe5 dxe5 13. Qc2
> Re8 14. Bb2 Bd6 15. Rad1 e4 16. h3 Be5 17. Nb1 Bf4 18. Bc1 Bc7 19. f4
> exf3 20. Bxf3 Be5 21. Bb2 Bxb2 22. Qxb2 Qd6 23. Qf2 Bf5 24. Nd2 Re5
> 25. Rfe1 Rae8 26. Rxe5 Rxe5 27. Re1 Rxe1+ 28. Qxe1 h6 29. Qe3 Bc2
> 30. Qc3 Bg6 31. Qe3 Kh8 32. Kf2 Bf5 33. Kg1 Qg3 34. Nf1 Qc7 35. Nd2
> Bg6 36. Kf2 Kh7 37. Kg1 Qd6 38. Kf2 Bc2 39. Qc3 Bf5 40. Qe3 Bd7 41.
> Kg1 Kg8 42. Ne4 Nxe4 43. Qxe4 f5 44. Qe3 f4 45. Qe2 Kf7 46. Bg4 Bxg4
> 47. Qxg4 b5 48. Kf2 bxc4 49. bxc4 Kf6 50. g3 fxg3+ 51. Qxg3 Qxg3+ 52.
> Kxg3 g5 53. h4 a5 54. a4 Kf5 55. hxg5 Kxg5 56. Kh3 h5 57. Kg3 Kf5 58.
> Kh4 Kg6 59. d6 Kf6 60. Kxh5 Ke6 61. Kg5 Ke5 Kxd6 62. Kf5 Kd7 63. Ke5
> Kc6 64. Ke6 Kc7 65. Kd5 Kb6 66. Kd6 Kb7 67. Kxc5 Ka6 68. Kc6Ka7 69. c5
> Ka6 70. Kd6 Kb7 71. c6+ Kc8 72. c7 Kb7 73. Kd7 Ka6 74. c8=Q+ 62. d7
> Kd4 63. d8=Q+ Kc3 64. Qxa5+ Kd4 65. Qb5 Kd3 66. a5 Ke4 67. a6 Kd3
> 68. a7 Ke3 69. a8=Q Kf2 70. Qa3 Kg1 71. Qbb2 Kf1 72. Qaa1# 0-1
>
|
| | | | | | | | |
Date: 25 Mar 2006 18:02:05
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Sat, 25 Mar 2006 09:17:19 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>Thanks. I do not know how to get those moves into the program without
>entering them manually. I tried saving in notepad as .pgn but that didn't
>work. Help?
You can do copy the game (select text then copy) and in Shredder GUI
"edit/past game".
|
| | | | | | | | | |
Date: 25 Mar 2006 15:16:15
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Ok, thanks.
What was the time control for that game? I would like to use something
similar.
adp
"RD" <fdgldfj@hotmails.com > wrote in message
news:s91b22hsti8lgn20b4tjsdadafmetrbgt2@4ax.com...
> On Sat, 25 Mar 2006 09:17:19 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>>Thanks. I do not know how to get those moves into the program without
>>entering them manually. I tried saving in notepad as .pgn but that didn't
>>work. Help?
>
> You can do copy the game (select text then copy) and in Shredder GUI
> "edit/past game".
>
|
| | | | | |
Date: 25 Mar 2006 01:53:12
From: James
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
RD wrote :
> On Fri, 24 Mar 2006 15:14:30 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>> (But somehow we're supposed to believe that 1900 players could
>> have).
>
>
> Again, computers are stronger than humans because they can calculate
> 4/5 moves perfectly. That's about it. This 'brute force' technique
> only works in chess because the branches are fewer. In games like 'Go'
> that have larger branches, computers are very weak compared to humans.
> There is no hope in foreseeable future that computers will ever beat
> 'Go' masters using techniques like 'brute force' (counting beans
> basically). I would like to see better quality and smarter programs
> (not bean counters).
>
Your point is not that simple. What about Othello programs which use
learning technics? I wrote one of them a few years ago (Otage) which
reached the top ten on the IOS, and used to beat every human opponent.
This program learned to play by itself, using only patterns recognition
and reinforcement learning (with of course an MTD(f) algorithm).
Is it a bean counter or a smart program?
Even regarding chess, it is not that clear. There is a lot of
intelligence in the best programs. Reading Ed Shroder, or any other top
chess programmer, you realize that futility pruning, null move
heuristic, etc, etc are extremely clever things, difficult to implement
efficiently (even bitboards based move generators are not that easy to
write).
Writing a "bean counter" chess program is easy, writing a smart program
is difficult. I think that denying "beauty" or "intelligence" to chess
played by engines is, in a way, denying intelligence to the people who
make them.
Computers are not intelligent, good programmers are.
|
| | | | | | |
Date: 27 Mar 2006 13:28:24
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
James <chess123@alliot.org > wrote:
> Even regarding chess, it is not that clear. There is a lot of
> intelligence in the best programs. Reading Ed Shroder, or any other
> top chess programmer, you realize that futility pruning, null move
> heuristic, etc, etc are extremely clever things,
The point originally under discussion is that Rybka has far more
chess-based knowledge than other engines. Pretty much any good
program has good pruning.
Dave.
--
David Richerby Miniature Dictator (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ totalitarian leader but you can hold
in it your hand!
|
| | |
Date: 19 Mar 2006 09:27:43
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Sat, 18 Mar 2006 10:10:50 -0000, "Terry"
<terry@tbean.freeserve.co.uk > wrote:
>Did you play both engines on the same PC.
I had similar results... Rybka 1.1 vs Shredder 9.1 with scores like
11 wins 2 loses and 7 draws for Rybka after 20 games. Rybka is clearly
stronger.
Rybka still needs improvements in some positions (especially end
game). It's still a new program. I suspect after one or two years
Rybka 2 or Rybka 3 might be 200-250+ points stronger than present-day
Fritz and Shredder on the current hardware. That would be big jump.
|
| | |
Date: 18 Mar 2006 14:00:53
From: Johnny T
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Simply
Because computer hardware is sufficiently strong enough to add more
chess knowledge, but insufficiently strong to add more search depth.
More complicated.
In the "early days" computers could not search very deep at a given
time, just because there were not enough clock cycles to do so. So
computers were given more "knowledge" and were quite exploitable (to use
a game theory term), mostly due to either horizon problems (the computer
can't see beyond a certain depth and misses relatively short term
tactical problems), or due to knowledge problems, this could be caused
by a whole host of reasons, including lack of understanding of the
programmers.
In the recent past, computers started to become fast enough that it was
possible to get to an 8 move horizon, as a matter of fact getting that
last possible ply, with simplistic knowledge, was enough to play "world
class" chess. Not the very best chess, chess that still seemed not
pretty, or knowledgeably, but chess that was difficult to refute. Still
seemingly exploitable to the worlds best.
Recently, two things have happened. Either massive increases in search
depth (Deep Blue and Hydra), which still with simplistic knowledge, but
deep enough to find the "right answers" anyways, or more interestingly,
EXCESS computing power.
Todays desktop machines are significantly more powerful than the
machines of 4 years ago, but no where NEAR powerful enough to add plys
to a chess search (because it is an exponential problem), but there is
now enough excess cpu cycles that can't be usefully exploited on getting
deeper, that now can be utilized for being smarter. Rybka, and arguably
Fritz9 are the two clearest examples of that now.
It is actually a very interesting time, because unlike most chess
optimizations up until now which were all about getting deeper, and all
those techniques were largely shared and showed up in open source like
crafty, and all the engines were in a fairly narrow ballpark (and the
differences were generally attributed to being exploitations of other
engine tendencies), given that base level, the next level will be based
on chess knowledge, and the quality of exploitation, and that we shall
see some significant ELO changes and differences between the engines.
Already you hear about people saying these programs play "differently"
than computer programs of the past. And I suspect we shall see more
"beauty" in the play, because we shall see strength in not being just
irrefutable, but strength based on the structure and knowledge of the
game.
|
| | | |
Date: 19 Mar 2006 22:06:16
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
I must defer to your obviously superior knowledge of computers and chess
programming. However, I strongly disagree with everyone who snickers at the
lack of "beauty" in computer chess play.
Bobby Fischer once said there he does not believe in psychology, only good
moves. I'd rather win every game by playing ugly, computer moves than win
the way I play (creatively, unsoundly, weakly). Objectively speaking, it
makes no sense to speak of "beauty" in a style of play when the moves speak
for themselves.
Hydra crushed Adams because it made better moves which may or may not have
met someone's criteria for loveliness. The computer won because it simply
made stronger moves.
I think it's time we humans, especially those of us who have no chance of
winning even one game against Frtiz, to break out of our delusions about
chess engines, that they somehow understand less about chess than we do.
Even saying a computer plays closed positions, or endgames, worse than the
average GM is irrelevant. If Michael Adams, rated in the mid-2700s at the
time, could not find a way to steer his games into familiar (for Mickey)
territory, then nobody likely can do it.
We have reached the point where our chess-playing inventions out-play us.
*All* of us. Let's get over it and give credit where it is due.
adp
"Johnny T" <not@home.com > wrote in message
news:441c8312$0$56215$ae4e5890@news.nationwide.net...
> Simply
>
> Because computer hardware is sufficiently strong enough to add more chess
> knowledge, but insufficiently strong to add more search depth.
>
> More complicated.
> In the "early days" computers could not search very deep at a given time,
> just because there were not enough clock cycles to do so. So computers
> were given more "knowledge" and were quite exploitable (to use a game
> theory term), mostly due to either horizon problems (the computer can't
> see beyond a certain depth and misses relatively short term tactical
> problems), or due to knowledge problems, this could be caused by a whole
> host of reasons, including lack of understanding of the programmers.
>
> In the recent past, computers started to become fast enough that it was
> possible to get to an 8 move horizon, as a matter of fact getting that
> last possible ply, with simplistic knowledge, was enough to play "world
> class" chess. Not the very best chess, chess that still seemed not
> pretty, or knowledgeably, but chess that was difficult to refute. Still
> seemingly exploitable to the worlds best.
>
> Recently, two things have happened. Either massive increases in search
> depth (Deep Blue and Hydra), which still with simplistic knowledge, but
> deep enough to find the "right answers" anyways, or more interestingly,
> EXCESS computing power.
>
> Todays desktop machines are significantly more powerful than the machines
> of 4 years ago, but no where NEAR powerful enough to add plys to a chess
> search (because it is an exponential problem), but there is now enough
> excess cpu cycles that can't be usefully exploited on getting deeper, that
> now can be utilized for being smarter. Rybka, and arguably Fritz9 are the
> two clearest examples of that now.
>
> It is actually a very interesting time, because unlike most chess
> optimizations up until now which were all about getting deeper, and all
> those techniques were largely shared and showed up in open source like
> crafty, and all the engines were in a fairly narrow ballpark (and the
> differences were generally attributed to being exploitations of other
> engine tendencies), given that base level, the next level will be based
> on chess knowledge, and the quality of exploitation, and that we shall see
> some significant ELO changes and differences between the engines.
>
> Already you hear about people saying these programs play "differently"
> than computer programs of the past. And I suspect we shall see more
> "beauty" in the play, because we shall see strength in not being just
> irrefutable, but strength based on the structure and knowledge of the
> game.
|
| | | | |
Date: 20 Mar 2006 07:37:32
From: Johnny T
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Ange1o DePa1ma wrote:
> I must defer to your obviously superior knowledge of computers and chess
> programming. However, I strongly disagree with everyone who snickers at the
> lack of "beauty" in computer chess play.
I appreciate your comment, but I think you misunderstood. I didn't talk
much about the Hydra and before that Deep Blue acceleration of depth.
And specifically Hydra.
Being deep enough still seems to be sufficient to play world champion
beating chess, and that if you are deep enough you start playing in ways
that even humans appreciate. Though there have been some grumblings
about Hydra.
But it will be many many years, if ever (Moore's law simply doesn't get
us there without breaking the laws of physics), we will have the kind of
hardware available to us everyday consumers and chess players that will
be able to get to the depth of Hydra.
So as stated before we have this in-between, where our desktop machines
can't get deeper, but they can get "smarter", and this "smarter" is
qualitatively changing the choice of moves, and there is beginning to be
evidence that this is resulting in quantitatively better chess.
|
| | | | | |
Date: 20 Mar 2006 16:23:06
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Johnny T <not@home.com > wrote:
> Moore's law simply doesn't get us there without breaking the laws of
> physics
Moore's law laughs at your laws of physics! :-)
The feature size of chips has, for several years, been small enough
compared to the wavelength of the ultra-violet used for etching that
the shadow-mask *ought* to be acting like a diffraction grating rather
than casting shadows. Somehow, it doesn't and photolithography still
works.
Dave.
--
David Richerby Frozen Radioactive Clock (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a clock but it'll make you glow
in the dark and it's frozen in a block
of ice!
|
| | | | |
Date: 20 Mar 2006 14:24:56
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Ange1o DePa1ma <angelodpnospam@nospam.gmail.com > wrote:
> I think it's time we humans, especially those of us who have no
> chance of winning even one game against Frtiz, to break out of our
> delusions about chess engines, that they somehow understand less
> about chess than we do.
But, um, chess engines do understand less about chess than we do. We
understand at least a little bit about chess, while the computer
doesn't `understand' in any real sense. It just plays, albeit with
better results than our understanding gives us.
> If Michael Adams, rated in the mid-2700s at the time, could not find
> a way to steer his games into familiar (for Mickey) territory, then
> nobody likely can do it.
As I recall, even Adams himself has admitted that he was under-
prepared for that match.
> We have reached the point where our chess-playing inventions
> out-play us. *All* of us.
Not quite. I wrote a chess program a couple of years ago and was
surprised to find that it was almost exactly as good as me. If I
played something that I recognized (say a move later) as a bad move,
it would beat me; if I didn't, I would beat it. I didn't try very
hard to make it better after that point, though.
Dave.
--
David Richerby Mouldy Chair (TM): it's like a chair
www.chiark.greenend.org.uk/~davidr/ but it's starting to grow mushrooms!
|
| | | | |
Date: 20 Mar 2006 03:45:52
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Sun, 19 Mar 2006 22:06:16 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>Bobby Fischer once said there he does not believe in psychology, only good
>moves. I'd rather win every game by playing ugly, computer moves than win
>the way I play (creatively, unsoundly, weakly). Objectively speaking, it
>makes no sense to speak of "beauty" in a style of play when the moves speak
>for themselves.
By 'beauty he meant structure and knowledge of chess. Several years
ago computers played 'ugly' (ugly = just bad and stupid chess) but
still did fairly well because the humans blundered (even when the
human player had better position).
Frankly, I wasn't impressed even by the quality of some of the games
of Deep Blue 2 in 1997.
>Hydra crushed Adams because it made better moves which may or may not have
>met someone's criteria for loveliness.
Hydra actually played very 'beautiful' (see the definition above)
chess.
Even just a few years ago, I wasn't at all impressed by computer
chess. They played ugly and stupid chess and usually won when the GM
would blunder in better/winning positions (even Kramnnik-Fritz match
only a few years ago had games like that).
I think this has begun to change now. Computers are playing much
better now. Hopefully, they would finally start playing 'good'
('beautiful') chess and win because of that (and not just rely on
humans to blunder to win bad positions).
|
| | | | | |
Date: 20 Mar 2006 09:33:26
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
There is something to be said for playing sound moves throughout a game. I
am most proud of the few games I've won where I did not make any - 0.3
pawn-equivalent blunders.
If the moves by which most computers win are ugly, then why aren't those who
pass these esthetic judgements able to parry them over the board? What is
the difference between a "harmonious" or "beautiful" move that humans can
solve (i.e. by maintaining the evaluation status) OTB and an "ugly" move
which is difficult for Kramnik, Adams, or Kasparov to meet, which causes
them to "blunder"?
Objectively, scientifically, you would have to say that the latter move is
stronger. There is no other explanation. If our top players represent the
paragon of chess beauty, and they are consistently beaten by computers, then
beauty is irrelevant.
We all make the mistake of referring to various stages of the game as if
they had some independent significance. "He was busted!!" (but I lost
anyway), or "I was clearly winning but I made one bad move and got mated"
are the excuses of patzers. "Fritz is no good at endgames" or "Shredder
plays horribly in closed positions" are mere excuses. If you can
consistently get Fritz to an endgame that you can win, or steer your games
against Shredder into locked pawn structures and I will admit you are equal
to or better those engines.
To paraphrase Fischer there are no beautiful moves, only strong moves.
Angelo DePa1ma
"RD" <fdgldfj@hotmails.com > wrote in message
news:7h8s125hphh51hmukrlrk79vmfa4iqr43n@4ax.com...
> On Sun, 19 Mar 2006 22:06:16 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>>Bobby Fischer once said there he does not believe in psychology, only good
>>moves. I'd rather win every game by playing ugly, computer moves than win
>>the way I play (creatively, unsoundly, weakly). Objectively speaking, it
>>makes no sense to speak of "beauty" in a style of play when the moves
>>speak
>>for themselves.
>
> By 'beauty he meant structure and knowledge of chess. Several years
> ago computers played 'ugly' (ugly = just bad and stupid chess) but
> still did fairly well because the humans blundered (even when the
> human player had better position).
>
> Frankly, I wasn't impressed even by the quality of some of the games
> of Deep Blue 2 in 1997.
>
>>Hydra crushed Adams because it made better moves which may or may not have
>>met someone's criteria for loveliness.
>
> Hydra actually played very 'beautiful' (see the definition above)
> chess.
>
> Even just a few years ago, I wasn't at all impressed by computer
> chess. They played ugly and stupid chess and usually won when the GM
> would blunder in better/winning positions (even Kramnnik-Fritz match
> only a few years ago had games like that).
>
> I think this has begun to change now. Computers are playing much
> better now. Hopefully, they would finally start playing 'good'
> ('beautiful') chess and win because of that (and not just rely on
> humans to blunder to win bad positions).
>
|
| | | | | | |
Date: 20 Mar 2006 19:40:46
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Mon, 20 Mar 2006 09:33:26 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>To paraphrase Fischer there are no beautiful moves, only strong moves.
You missed the point completely of my post and repeated the same thing
again. Computers played ugly and bad chess. They didn't win because
they made stronger moves. They usually won because human blundered.
Please read the post again.
|
| | | | | | | |
Date: 27 Mar 2006 15:21:34
From: irchans
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
David, very nice post. I particularly liked your reference to the ICC
finger notes which I had not seen before.
David Richerby wrote:
> irchans <infinitgames@yahoo.com> wrote:
> > David Richerby wrote:
> >> The point originally under discussion is that Rybka has far
> >> more chess-based knowledge than other engines. Pretty much any
> >> good program has good pruning.
> >
> > It seems to me that increasing chess-based knowledge improves
> > pruning either directly (for example you could have a rule to prune
> > queen exchanges early in the game) or indirectly by giving better
> > evaluations at leaf nodes causing the program to prune more effectively
> > without changing the pruning algorithm.
>
> I would have thought that the primary advantage of a better evaluation
> function would be not increased pruning but decreased required search
> depth.
I would like to stick with the phrase "more chess-based knowledge"
(hereafter CBK) rather than "better evaluation function". I don't
think it is entirely safe to assume that the added CBK is only applied
in the evaluation function. To me the project goals "knowledge-guided,
human-like search and evaluation" seems to imply that maybe Vasik is
adding CBK to the pruning.
> For example, if my evaluation function says that a position
> with a white pawn on a2 and the black king on a1 is won for White (to
> move) then I don't need to search another nine ply to find the pawn
> promotion and the win. (You could argue that this is a form of
> pruning but, as I understand the term, it refers to decreasing the
> width of the tree, rather than the depth.)
Before reading your example, I thought it was easy to separate pruning
and evaluation, but now I wonder.
I would like to explore this topic further, but I am not feeling well
today. I will post again in a few days when I have recovered enough
for more than shallow thinking.
Cheers,
Irchans
|
| | | | | | | | |
Date: 28 Mar 2006 09:20:49
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
irchans <infinitgames@yahoo.com > wrote:
> David Richerby wrote:
>> irchans <infinitgames@yahoo.com> wrote:
>>> It seems to me that increasing chess-based knowledge improves
>>> pruning [...]
>>
>> I would have thought that the primary advantage of a better
>> evaluation function would be not increased pruning but decreased
>> required search depth.
>
> I would like to stick with the phrase "more chess-based knowledge"
> (hereafter CBK) rather than "better evaluation function". I don't
> think it is entirely safe to assume that the added CBK is only applied
> in the evaluation function.
Good point. I don't think that affects the truth of the sentence you
quoted but it does mean that it's not the whole story.
> To me the project goals "knowledge-guided, human-like search and
> evaluation" seems to imply that maybe Vasik is adding CBK to the
> pruning.
I'd not thought of that but it seems very likely now that you mention
it.
> I would like to explore this topic further, but I am not feeling well
> today. I will post again in a few days when I have recovered enough
> for more than shallow thinking.
Get well soon!
Dave.
--
David Richerby Portable Salted Peanut (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a roasted nut but it's covered in salt
and you can take it anywhere!
|
| | | | | | | | |
Date: 28 Mar 2006 02:17:01
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
"irchans" <infinitgames@yahoo.com > wrote in message
news:1143501694.241651.140470@g10g2000cwb.googlegroups.com...
> David, very nice post. I particularly liked your reference to the ICC
> finger notes which I had not seen before.
>
>
> David Richerby wrote:
>> irchans <infinitgames@yahoo.com> wrote:
>> > David Richerby wrote:
>> >> The point originally under discussion is that Rybka has far
>> >> more chess-based knowledge than other engines. Pretty much any
>> >> good program has good pruning.
>> >
>> > It seems to me that increasing chess-based knowledge improves
>> > pruning either directly (for example you could have a rule to prune
>> > queen exchanges early in the game) or indirectly by giving better
>> > evaluations at leaf nodes causing the program to prune more effectively
>> > without changing the pruning algorithm.
>>
>> I would have thought that the primary advantage of a better evaluation
>> function would be not increased pruning but decreased required search
>> depth.
>
> I would like to stick with the phrase "more chess-based knowledge"
> (hereafter CBK) rather than "better evaluation function". I don't
> think it is entirely safe to assume that the added CBK is only applied
> in the evaluation function. To me the project goals "knowledge-guided,
> human-like search and evaluation" seems to imply that maybe Vasik is
> adding CBK to the pruning.
I am not trying to bust your chops here David. But how do you instill
"knowledge" into a software program without either relying on (better) brute
force or hard-wiring it?
|
| | | | | | | |
Date: 27 Mar 2006 05:21:51
From: irchans
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
>The point originally under discussion is that Rybka has far
>more chess-based knowledge than other engines. Pretty much any
>good program has good pruning.
Dave,
It seems to me that increasing chess-based knowledge improves
pruning either directly (for example you could have a rule to prune
queen exchanges early in the game) or indirectly by giving better
evaluations at leaf nodes causing the program to prune more effectively
without changing the pruning algorithm.
What evidence do we have that Rybka has more chess-based knowledge?
At this point my main evidence for that hypothesis is 1) Vasik is a
strong chess player, 2) I have heard rumors that some grandmasters have
said that Rybka plays more like a human, and 3) Rybka plays better than
other engines despite the fact that it searches fewer nodes. It seems
possible that 1), 2), and 3) could be true even if Rybka did not have
more chess knowledge. If Vasik has invented some new general AI search
method, it might just appear that Rybka has more explicit chess
knowledge programmed in, but in reality it just had a better search
method.
|
| | | | | | | | |
Date: 27 Mar 2006 16:12:11
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
irchans <infinitgames@yahoo.com > wrote:
> David Richerby wrote:
>> The point originally under discussion is that Rybka has far
>> more chess-based knowledge than other engines. Pretty much any
>> good program has good pruning.
>
> It seems to me that increasing chess-based knowledge improves
> pruning either directly (for example you could have a rule to prune
> queen exchanges early in the game) or indirectly by giving better
> evaluations at leaf nodes causing the program to prune more effectively
> without changing the pruning algorithm.
I would have thought that the primary advantage of a better evaluation
function would be not increased pruning but decreased required search
depth. For example, if my evaluation function says that a position
with a white pawn on a2 and the black king on a1 is won for White (to
move) then I don't need to search another nine ply to find the pawn
promotion and the win. (You could argue that this is a form of
pruning but, as I understand the term, it refers to decreasing the
width of the tree, rather than the depth.)
> What evidence do we have that Rybka has more chess-based knowledge?
The author's says it does. The ICC finger notes for Rybka read,
``1: Computer account for Rybka chess engine
2: CCT8 Winner http://www.cctchess.info/
3: Project goals: knowledge-guided, human-like search and evaluation
4: Author: Vasik Rajlich (finger vrajlich)''
There are also some hints at
http://www.rybkachess.com/free/Rybka11FAQ.htm
where he says that Rybka 1.1 has increased middlegame evaluation
knowledge over the previous beta and that a major priority for version
1.2 is `More chess knowledge'. He also refers to the programs
`endgame knowledge'.
Dave.
--
David Richerby Carnivorous Aluminium Tool (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a handy household tool that's
really light but it eats flesh!
|
| | | | | | | |
Date: 20 Mar 2006 20:21:00
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Mon, 20 Mar 2006 19:40:46 GMT, RD <fdgldfj@hotmails.com > wrote:
>>To paraphrase Fischer there are no beautiful moves, only strong moves.
>
>You missed the point completely of my post and repeated the same thing
>again. Computers played ugly and bad chess. They didn't win because
>they made stronger moves. They usually won because human blundered.
>Please read the post again.
I meant to say that computers used to play bad (ugly) chess with no
(or hardly any) tactical blunders. This is how they usually won. They
didn't win because they played better/stronger chess -- they won
because they didn't make tactical mistakes and humans (in 40 move
game) would occasionally blunder tactically and thus lose even though
human player had a better position . This isn't 'strong' chess. It's
very boring/unimpressive/ugly chess.
I am rated around 2200. When I used to play comps a few years ago, it
would almost always play stupid/illogical game and would get itself
into worse position. Despite that, it would win most of time because I
would blunder tactically. Okay, comp didn't blunder (tactically) as
often as I do, but what is so impressive about that kind of chess?
What 'strong' moves? Beautiful game (the entire game) is played
strongly. Both are same thing. Playing weakly/illogically most of the
game and then winning because your opponent blundered tactically isn't
strong (nor 'beutiful').
Having said all that, comp definitely are improving their positional
play for last several years.
|
| | | | | | | | |
Date: 21 Mar 2006 09:28:48
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
"RD" <fdgldfj@hotmails.com > wrote
>>You missed the point completely of my post and repeated the same thing
>>again. Computers played ugly and bad chess.
When? Computers have been GM strength for at least 10 years.
>>They didn't win because
>>they made stronger moves. They usually won because human blundered.
>>Please read the post again.
It is impossible to win a chess game without having played better moves.
> I meant to say that computers used to play bad (ugly) chess with no
> (or hardly any) tactical blunders. This is how they usually won. They
> didn't win because they played better/stronger chess --
If you're really a master strength player you should realize by now that
this is impossible.
> they won
> because they didn't make tactical mistakes and humans (in 40 move
> game) would occasionally blunder tactically and thus lose even though
> human player had a better position . This isn't 'strong' chess. It's
> very boring/unimpressive/ugly chess.
You may find it boring, but it beats almost everyone on earth.
I have done a lot of experimentation against my various engines. In my
opinion computers blurr the distinction between tactics and positional chess
(I have not believed in that distinction for several years now). To me, what
we call "positional chess" is nothing more than tactics whose goal is better
piece placement, occupation of squares, prophylaxis, etc. These "mini
tactics" don't mate in 3 or win a Knight, but if you execute enough of them
pretty soon you have an overwhelming position and something gives. My
engines -- I can't speak for anyone else's -- do this very well. I've played
hundreds of games in which the computer built up several advantages until it
broke through.
If you consistently play the third best move against the computer, before
you know it you are going to lose a pawn or a piece, and subsequently the
game. For at least as long as I have owned a chess playing computer,
outright blunders have not been the prerequisite for losing.
> I am rated around 2200. When I used to play comps a few years ago, it
> would almost always play stupid/illogical game and would get itself
> into worse position. Despite that, it would win most of time because I
> would blunder tactically.
In other words the computer would "often" get into a worse position against
you, but defend sufficiently to hold the position until you "blundered." In
other words, it outplayed you. I would be satisfied to win my next 100 games
in that fashion.
> Having said all that, comp definitely are improving their positional
> play for last several years.
adp
|
| | | | | | | | | |
Date: 22 Mar 2006 02:07:06
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Tue, 21 Mar 2006 09:28:48 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>When? Computers have been GM strength for at least 10 years.
I don't believe that for a second (excluding DB who played very few
games anyway to make a judgment). In 1995/1996 Roman used to play a
lot of blitz on ICC against crafty. Computers are supposed to be much
stronger than humans at blitz, but the comp would often end up with
bad positions and lose. Also, old comps didn't understand king safety.
Even 1800 rated players could beat them often using stonewall setup.
How often can an 1800 players beat GMs using stonewall?
Forget 1996, I just saw yesterday a horrible end game positional
blunder by Fritz 9 that even 1600 player won't make. Time control was
5 3 (blitz). Here is the position
8/p4p2/1p6/1Pp2kp1/5P1P/1P2K3/1P5P/8 b - - 0 48
Fritz 9 here played gxf4. What the heck? Doesn't it understand the
concept of outside passed pawn in the endgame? (doesn't matter if gxh4
is winning/losing/drawing -- gxh4 must be played -- gxf4 is just bad
positional move). No calculations are needed. Any reasonable player
would immediately see that gxf4 is a bad choice, but yet Fritz 9 (as
strong as GM?) plays that kind of move? (it was blitz so maybe Fritz
would have seen gxf4 is a blunder in slower game -- but still a human
wouldn't make positional blunder like that even in one minute game).
No doubt comps are playing much better now, but there is still lots of
room for improvement. Next stage is making them positionally much
stronger.
|
| | | | | | | | | | |
Date: 22 Mar 2006 11:51:05
From: David Richerby
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
RD <fdgldfj@hotmails.com > wrote:
> Forget 1996, I just saw yesterday a horrible end game positional
> blunder by Fritz 9 that even 1600 player won't make. Time control was
> 5 3 (blitz). Here is the position
>
> 8/p4p2/1p6/1Pp2kp1/5P1P/1P2K3/1P5P/8 b - - 0 48
White has quite the Nigel Short pawn structure, there.
> Fritz 9 here played gxf4. What the heck? [...] (it was blitz so
> maybe Fritz would have seen gxf4 is a blunder in slower game -- but
> still a human wouldn't make positional blunder like that even in one
> minute game).
I'd be hat-eatingly surprised if Fritz didn't spot that gxf4+ is a
blunder given only a little more time.
Dave.
--
David Richerby Moistened Postman (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ man who delivers the mail but it's
moist!
|
| | | | | | | | | | |
Date: 22 Mar 2006 00:26:19
From: Ange1o DePa1ma
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
Are you saying that the best desktop engine circa 1996 was not at least as
strong as a slew of second- or third-tier GMs, and could not achieve a GM
norm in rated tournaments?
I quote from: http://www.geocities.com/SiliconValley/Lab/7378/comphis.htm:
"The 11th AEGON Computer Chess Tournament (Mankind vs Machine) was held on
April 10-17, 1996 in The Hague, Netherlands. There were 50 masters,
International Masters, and Grandmasters and 50 computers (most playing on HP
Pentium-166 machines with 16MB of RAM). Yasser Seirawan won the event with 6
straight wins and no losses. The best computer was QUEST, with 4.5/6 and a
2652 performance rating. The machines won with 162.5 points versus the
humans with 137.5 points."
You and I should play such ugly chess.
Of course computers are deficient in many areas of chess, but their
strengths more than make up for those defects. My favorite example is from
the first time I beat a 2100 player (I was rated in the mid-1800s at the
time). I saw a very nice outflanking manoever, 6 or 7 moves ahead, that
enabled me to win an ending even though I was a pawn down. When I feed the
position to Shredder it thinks and thinks and thinks and says my opponent is
better. There are many other examples too.
"RD" <fdgldfj@hotmails.com > wrote in message
news:0f51229qf1rhchk5alpfj3ii5em8sbpl6d@4ax.com...
> On Tue, 21 Mar 2006 09:28:48 -0500, "Ange1o DePa1ma"
> <angelodpnospam@nospam.gmail.com> wrote:
>
>>When? Computers have been GM strength for at least 10 years.
>
> I don't believe that for a second (excluding DB who played very few
> games anyway to make a judgment). In 1995/1996 Roman used to play a
> lot of blitz on ICC against crafty. Computers are supposed to be much
> stronger than humans at blitz, but the comp would often end up with
> bad positions and lose. Also, old comps didn't understand king safety.
> Even 1800 rated players could beat them often using stonewall setup.
> How often can an 1800 players beat GMs using stonewall?
>
> Forget 1996, I just saw yesterday a horrible end game positional
> blunder by Fritz 9 that even 1600 player won't make. Time control was
> 5 3 (blitz). Here is the position
>
> 8/p4p2/1p6/1Pp2kp1/5P1P/1P2K3/1P5P/8 b - - 0 48
>
> Fritz 9 here played gxf4. What the heck? Doesn't it understand the
> concept of outside passed pawn in the endgame? (doesn't matter if gxh4
> is winning/losing/drawing -- gxh4 must be played -- gxf4 is just bad
> positional move). No calculations are needed. Any reasonable player
> would immediately see that gxf4 is a bad choice, but yet Fritz 9 (as
> strong as GM?) plays that kind of move? (it was blitz so maybe Fritz
> would have seen gxf4 is a blunder in slower game -- but still a human
> wouldn't make positional blunder like that even in one minute game).
>
> No doubt comps are playing much better now, but there is still lots of
> room for improvement. Next stage is making them positionally much
> stronger.
>
|
| | | | | | | | | | | |
Date: 22 Mar 2006 06:23:42
From: RD
Subject: Re: Rybka : Why would an engine suddenly be 100 poins better?
|
On Wed, 22 Mar 2006 00:26:19 -0500, "Ange1o DePa1ma"
<angelodpnospam@nospam.gmail.com > wrote:
>"The 11th AEGON Computer Chess Tournament (Mankind vs Machine) was held on
>April 10-17, 1996 in The Hague, Netherlands. There were 50 masters,
>International Masters, and Grandmasters and 50 computers (most playing on HP
>Pentium-166 machines with 16MB of RAM). Yasser Seirawan won the event with 6
>straight wins and no losses. The best computer was QUEST, with 4.5/6 and a
>2652 performance rating. The machines won with 162.5 points versus the
>humans with 137.5 points."
Performance rating means nothing in a 6 game tournament. Shakhriyar
Mamedyarov had a performance rating of 2953 in the World Junior Chess
Championship (http://chessbase.com/newsdetail.asp?newsid=2745) .. |
|
