about Botvinnik theory...

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 19:49:56:

In a basis of a method - " struggle for a material ". First it's choice of the purpose for attack.

Object of attack can be not only a figure of the opponent, but also a weak squares, squares next to king and etc.
Then the figure goes to the chosen object (on the shortest trajectories determined beforehand) and figures of the
opponent prevent this.
Even if in process search also it will be found out that the purpose is not achievable, in process
struggle can be achieved the general improvement of a position.
The program can change object of attack if in process search.
Actually it is necessary to build multilevel expert system. Where each level has the purposes.
The purposes of the bottom level are movements of the figures chosen the program on trajectories.
Higher level considers{examines} any of these purposes and puts bottom
to levels the new purposes, etc.
On the basis of the chosen purpose and the auxiliary purposes both opponents the zone of game is formed.
The following level watches{keeps up} interaction zones also unites cooperating zones in one.
And at last the highest level supervises search also carries out a choice of a move.
It is difficult for understanding at once, especially with such bad English.
For the best concept I shall result an example.
When you play a chess you do not see all of 64 fields. You interests as a rule
one or two sites of a board. On other sites of a board you simply do not pay
any attention. Thus you subconsciously form these zones.
1. To choose the purpose
2. To define{determine} trajectories of movement
3...

Method complex{difficult} enough however it is absolutely not necessary to apply it at once entirely.
In a result from search a plenty of figures leaves.
I do not think that this method is ideal. Certainly is not present. You can set many questions.
Unfortunately even in thick books Botvinnik there are no answers to all questions.
Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

[Günther Simon]

Geschrieben von: / Posted by: Günther Simon at 30 June 2003 20:04:53:
Als Antwort auf: / In reply to: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 19:49:56:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 

Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 20:24:59:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Günther Simon at 30 June 2003 20:04:53:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

Hello Günther
I shall check up thanks.
Best.
Vladimir.

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 20:27:31:
Als Antwort auf: / In reply to: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 19:49:56:

In a basis of a method - " struggle for a material ". First it's choice of the purpose for attack.

Object of attack can be not only a figure of the opponent, but also a weak squares, squares next to king and etc.
Then the figure goes to the chosen object (on the shortest trajectories determined beforehand) and figures of the
opponent prevent this.
Even if in process search also it will be found out that the purpose is not achievable, in process
struggle can be achieved the general improvement of a position.
The program can change object of attack if in process search.
Actually it is necessary to build multilevel expert system. Where each level has the purposes.
The purposes of the bottom level are movements of the figures chosen the program on trajectories.
Higher level considers{examines} any of these purposes and puts bottom
to levels the new purposes, etc.
On the basis of the chosen purpose and the auxiliary purposes both opponents the zone of game is formed.
The following level watches{keeps up} interaction zones also unites cooperating zones in one.
And at last the highest level supervises search also carries out a choice of a move.
It is difficult for understanding at once, especially with such bad English.
For the best concept I shall result an example.
When you play a chess you do not see all of 64 fields. You interests as a rule
one or two sites of a board. On other sites of a board you simply do not pay
any attention. Thus you subconsciously form these zones.
1. To choose the purpose
2. To define{determine} trajectories of movement
3...

Method complex{difficult} enough however it is absolutely not necessary to apply it at once entirely.
In a result from search a plenty of figures leaves.
I do not think that this method is ideal. Certainly is not present. You can set many questions.
Unfortunately even in thick books Botvinnik there are no answers to all questions.
Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

I am almost confident that Fritz uses something similar. Otherwise how to explain that in the middle
game at him depth = 13-15(9-12(Except for the some) in amateur programs) at 500-800 Kn/s. I do not think that it uses too simple
evaluation function.
Best
vladimir

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 20:39:58:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Günther Simon at 30 June 2003 20:04:53:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

[Günther Simon]

Geschrieben von: / Posted by: Günther Simon at 30 June 2003 20:54:04:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 20:39:58:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

Well, may be I need more beers to understand this
after e4+ which you gave, I simply play Kxg5 and none of
whites pawns survive if not sth really weird happens,
which I cannot even imagine )
I only see that g6 will win, because White cannot stop
White from getting a Q on g8 despite sacrificing its Bishop?
Houston I have a problem
Cheers,
Günther

[Andreas Herrmann]

Geschrieben von: / Posted by: Andreas Herrmann at 30 June 2003 20:57:17:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 20:39:58:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

ok, after e4 i would play Kxg5 (not checked with a prog). Can you show here a winning line for white.
Andreas

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 21:03:27:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Günther Simon at 30 June 2003 20:54:04:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

Well, may be I need more beers to understand this
after e4+ which you gave, I simply play Kxg5 and none of
whites pawns survive if not sth really weird happens,
which I cannot even imagine )
I only see that g6 will win, because White cannot stop
White from getting a Q on g8 despite sacrificing its Bishop?
Houston I have a problem
Cheers,
Günther

see main variant
g5-g6 Kf5-f6 g6-g7 Bc2-h7 e3-e4!!!
sorry i'm need beer
i remember to say that 3.e3-e4

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 30 June 2003 21:05:10:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Andreas Herrmann at 30 June 2003 20:57:17:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

ok, after e4 i would play Kxg5 (not checked with a prog). Can you show here a winning line for white.
Andreas

see main variant:
g5-g6 Kf5-f6 g6-g7 Bc2-h7 e3-e4!!!
sorry i'm need beer
i remember to say that 3.e3-e4

[Günther Simon]

Geschrieben von: / Posted by: Günther Simon at 30 June 2003 21:21:55:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 21:03:27:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

Well, may be I need more beers to understand this
after e4+ which you gave, I simply play Kxg5 and none of
whites pawns survive if not sth really weird happens,
which I cannot even imagine )
I only see that g6 will win, because White cannot stop
White from getting a Q on g8 despite sacrificing its Bishop?
Houston I have a problem
Cheers,
Günther

see main variant
g5-g6 Kf5-f6 g6-g7 Bc2-h7 e3-e4!!!
sorry i'm need beer
i remember to say that 3.e3-e4

Ahhh! 3.e4 thats a really fine move!
In my previous analysis, where I saw of course that g6 is the
only road to win, I completely missed 2...Bh7!! which works
because of 3.Kxh7 3...Nf3!!
[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Neue Partie"]
[Black "?"]
[Result ""]
[Annotator "Simon,Günther"]
[SetUp "1"]
[FEN "7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - 0 1"]
[PlyCount "27"]
[SourceDate "2002.08.17"]
1. g6 Kf6 2. g7 Bh7! 3. Kxh7? (3. e4!! Nf3 4. e5+ Kxe5 5. Kxh7 Ng5+ 6. Kg6 +-) 3...Nf3!! 4. g8=Q Ng5+ 5. Qxg5+ Kxg5 6. h6 c4 7. Kg7 c3 8. h7 c2 9. h8=Q c1=Q 10.Qh6+ Kg4 11. Qxe6+ Kg3 (should be drawn)
Cheers,
Günther

[Andreas Herrmann]

Geschrieben von: / Posted by: Andreas Herrmann at 30 June 2003 21:31:57:
Als Antwort auf: / In reply to: Re: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 21:05:10:

Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w generated with GenDiag 1.7 (C) 2002 by Andreas Herrmann
 
Hello Vladimir,
Is there something wrong with the FEN, I am quite sure that
e3e4 will never win (g7g6 should but looks easy)?
Regards,
Günther

write please:
White: King:h8, pawns:e3,g5,h5
Black: King:f5, Bishop:c2, Knight:e1, pawns:c7,c5,e6
white begin and win.
Congratulate me I started to write verses on english

ok, after e4 i would play Kxg5 (not checked with a prog). Can you show here a winning line for white.
Andreas

see main variant:
g5-g6 Kf5-f6 g6-g7 Bc2-h7 e3-e4!!!
sorry i'm need beer
i remember to say that 3.e3-e4

ok that makes more sense A realy nice position.
Holmes plays here the correct moves, but sees the e4 first after g6 and Kf6 was playd.
Andreas

[Sune Fischer]

Geschrieben von: / Posted by: Sune Fischer at 01 July 2003 00:34:03:
Als Antwort auf: / In reply to: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 19:49:56:

I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

Vladimir

Hi Vladimir
Thanks for sharing.
I won't deny that's it's still somewhat fuzzy to me exactly how it's done.
But I find it interesting and I do get the basic idea.
It will be exciting to how see if/when you release your engine, how successful it is.
Cheers,
-S.

[Ross Boyd]

Geschrieben von: / Posted by: Ross Boyd at 01 July 2003 02:20:25:
Als Antwort auf: / In reply to: about Botvinnik theory... geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 19:49:56:

In a basis of a method - " struggle for a material ". First it's choice of the purpose for attack.

Object of attack can be not only a figure of the opponent, but also a weak squares, squares next to king and etc.
Then the figure goes to the chosen object (on the shortest trajectories determined beforehand) and figures of the
opponent prevent this.
Even if in process search also it will be found out that the purpose is not achievable, in process
struggle can be achieved the general improvement of a position.
The program can change object of attack if in process search.
Actually it is necessary to build multilevel expert system. Where each level has the purposes.
The purposes of the bottom level are movements of the figures chosen the program on trajectories.
Higher level considers{examines} any of these purposes and puts bottom
to levels the new purposes, etc.
On the basis of the chosen purpose and the auxiliary purposes both opponents the zone of game is formed.
The following level watches{keeps up} interaction zones also unites cooperating zones in one.
And at last the highest level supervises search also carries out a choice of a move.
It is difficult for understanding at once, especially with such bad English.
For the best concept I shall result an example.
When you play a chess you do not see all of 64 fields. You interests as a rule
one or two sites of a board. On other sites of a board you simply do not pay
any attention. Thus you subconsciously form these zones.
1. To choose the purpose
2. To define{determine} trajectories of movement
3...

Method complex{difficult} enough however it is absolutely not necessary to apply it at once entirely.
In a result from search a plenty of figures leaves.
I do not think that this method is ideal. Certainly is not present. You can set many questions.
Unfortunately even in thick books Botvinnik there are no answers to all questions.
Unfortunately after death Botvinnik him have accused of a juggling of results of work of his program
on construction of a short tree search.
His program has found a move (e3-e4) and has generated a tree of all in 200 moves!
Look below.
I do not believe that there was a juggling.
7K/2p5/4p3/2p2kPP/8/4P3/2b5/4n3 w - - bm ;

(win)
Kaissa found e3e4 in several seconds. Botvinnik write that was 3 hous. 45 min.(on his Pioneer)
World Champion 1977 Chess 4.6 play Kh8-h7(draw)
I shall be pleased if written by me it is useful to somebody. Sorry I used translator Prompt.

I think I understand the general idea here.... but how to implement it is not so easy.
I believe David Slate planned a version of Nuchess/Chess which would make short range goals like manoevering to an outpost and similar board control ideas.
This is kind of similar isn't it???
And was it David Kittinger that spoke of doing root processing which would pre-scan the board for tactical motifs and apply static bonuses for achieving pieces on certain squares.... at deep search levels this became less and less reliable... and this was the flaw... but...
the difference is you are doing this type of processing _in_ the search.
I don't suppose you will be publishing Kaissa's source code?!? !!
No, I didn't think so...

TRACE finds 3.e3-e4 in a few seconds but only AFTER g6 and g7 have both been played... its a good test position.
Thanks for sharing your ideas!
I think this is the way to a GM level program... and as you mention elsewhere, the best programs are probably doing some of these things already.
Looking forward to Kaissa 3... maybe one day TRACE will start beating Kaissa 2...
Ross

[Sergei S. Markoff]

Geschrieben von: / Posted by: Sergei S. Markoff at 01 July 2003 10:28:59:
Als Antwort auf: / In reply to: little remarks geschrieben von: / posted by: Vladimir Yelin at 30 June 2003 20:27:31:

Hello Vladimir!

I am almost confident that Fritz uses something similar. Otherwise how to
explain that in the middle game at him depth = 13-15(9-12(Except for the some)
in amateur programs) at 500-800 Kn/s. I do not think that it uses too simple
evaluation function.

((:
I think no. You can see on SmarThink depths. For example see 0.10a. The extremely high depths of search is the result of aggressive pruning. For example SmarThink uses not only futility pruning/razoring but several other methods. SmarThink can decrease the depth is some node in some cases.
Newer versions of ST also uses a lot of new methods of aggressive search extending. Average depth means almost _nothing_. For example you can be surprised by high search depth of Junior. But Amir Ban wrote that Junior uses completely another depth concept. Depth in Junior is measuring in 1/2 ply! Depth 20 in Junior means depth 10 in SmarThink. Junior isn't agressively "pruner" but aggressively "extender". Also Amir wrote that some other commercials also uses non-Crafty depth measuring scheme.
The high nps-rate can be the result of effective lazy evaluation and some non well-known techniques. For example Vladimit Rybinkin says that Mirage calls eval. function practically only in the root of quiescence search. Mirage has completely another quiescence concept. There are several ways to skip nodes not evaluating it at all or, at least, completely.
Best wishes,
Sergei



http://www.aigroup.narod.ru

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 01 July 2003 11:32:40:
Als Antwort auf: / In reply to: Re: little remarks geschrieben von: / posted by: Sergei S. Markoff at 01 July 2003 10:28:59:

Hello Vladimir!

I am almost confident that Fritz uses something similar. Otherwise how to
explain that in the middle game at him depth = 13-15(9-12(Except for the some)
in amateur programs) at 500-800 Kn/s. I do not think that it uses too simple
evaluation function.

((:
I think no. You can see on SmarThink depths. For example see 0.10a. The extremely high depths of search is the result of aggressive pruning. For example SmarThink uses not only futility pruning/razoring but several other methods. SmarThink can decrease the depth is some node in some cases.
Newer versions of ST also uses a lot of new methods of aggressive search extending. Average depth means almost _nothing_. For example you can be surprised by high search depth of Junior. But Amir Ban wrote that Junior uses completely another depth concept. Depth in Junior is measuring in 1/2 ply! Depth 20 in Junior means depth 10 in SmarThink. Junior isn't agressively "pruner" but aggressively "extender". Also Amir wrote that some other commercials also uses non-Crafty depth measuring scheme.
The high nps-rate can be the result of effective lazy evaluation and some non well-known techniques. For example Vladimit Rybinkin says that Mirage calls eval. function practically only in the root of quiescence search. Mirage has completely another quiescence concept. There are several ways to skip nodes not evaluating it at all or, at least, completely.
Best wishes,
Sergei

Hello Sergei!
I had no in view of a concrete method.
You are probably right. We simply do not know in what units Fritz measures depth (1/2 ply èëè 2*ply:-) And as we do not know what methods it applies.
The fact that it not only pruning/razoring.
About metod Vladimit Rybinkin. I am not familiar with his method, but I know that some manufacturers chess
computers used something similar. Eval() function calls once in root and in depth of variants the first not losing material move gets out.
Best wishes.
Vladimir.

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 01 July 2003 12:37:10:
Als Antwort auf: / In reply to: Re: little remarks geschrieben von: / posted by: Sergei S. Markoff at 01 July 2003 10:28:59:

Hello Vladimir!

I am almost confident that Fritz uses something similar. Otherwise how to
explain that in the middle game at him depth = 13-15(9-12(Except for the some)
in amateur programs) at 500-800 Kn/s. I do not think that it uses too simple
evaluation function.

((:
I think no. You can see on SmarThink depths. For example see 0.10a. The extremely high depths of search is the result of aggressive pruning. For example SmarThink uses not only futility pruning/razoring but several other methods. SmarThink can decrease the depth is some node in some cases.
Newer versions of ST also uses a lot of new methods of aggressive search extending. Average depth means almost _nothing_. For example you can be surprised by high search depth of Junior. But Amir Ban wrote that Junior uses completely another depth concept. Depth in Junior is measuring in 1/2 ply! Depth 20 in Junior means depth 10 in SmarThink. Junior isn't agressively "pruner" but aggressively "extender". Also Amir wrote that some other commercials also uses non-Crafty depth measuring scheme.
The high nps-rate can be the result of effective lazy evaluation and some non well-known techniques. For example Vladimit Rybinkin says that Mirage calls eval. function practically only in the root of quiescence search. Mirage has completely another quiescence concept. There are several ways to skip nodes not evaluating it at all or, at least, completely.
Best wishes,
Sergei

[Event "Computer chess game"]
[Site "Vladimir"]
[Date "2003.07.01"]
[Round "1"]
[White "SmarThink 0.16b[eta]++"]
[Black "Kaissa2 2.0"]
[Result "0-1"]
[TimeControl "300"]
1. e4 c5 2. Nc3 d6 3. f4 Nc6 4. Nf3 f5 5. e5 dxe5 6. fxe5 e6 7. Bb5 Ne7 8.
O-O a6 9. Bc4 g5 10. d3 b5 11. Bb3 g4 12. Ng5 Nd4 13. Nge4 fxe4 14. Nxe4
Nef5 15. Nf6+ Kf7 16. Qxg4 Nxb3 17. Qh5+ Ke7 18. axb3 h6 19. Rf2 Bb7 20. g4
Ng7 21. Qh4 c4 22. Bxh6 Qd4 23. c3 Nf5 24. Ng8+ Kd7 25. Qh5 Qxe5 26. d4 Qd6
27. Bxf8 Rxf8 28. Qxh8 Qd5 29. Qh7+ Kd8 30. Qxb7 Qxb7 31. Nh6 cxb3 32. d5
Qb6 33. Nxf5 exf5 34. g5 Re8 35. Kf1 Qc5 36. g6 Qxd5 37. Kg1 Qd6 38. g7 Rg8
39. Rg2 b4 40. cxb4 Qd4+ 41. Kf1 Rxg7 42. Rxg7 Qxg7 43. Rd1+ Kc7 44. Rd2
Qe5 45. Re2 Qf4+ 46. Kg2 Qg4+ 47. Kf2 Qxb4 48. Ke3 Qd6 49. Kf3 Qd3+ 50. Kf2
f4 51. Re7+ Kd6 52. Re2 Kd5 53. h4 Qg3+ 54. Kf1 f3
{White resigns} 0-1

[Sergei S. Markoff]

Geschrieben von: / Posted by: Sergei S. Markoff at 01 July 2003 13:44:19:
Als Antwort auf: / In reply to: Re: little remarks geschrieben von: / posted by: Vladimir Yelin at 01 July 2003 11:32:40:

Hi!

I had no in view of a concrete method.
You are probably right. We simply do not know in what units Fritz measures
depth (1/2 ply èëè 2*ply:-)
And as we do not know what methods it applies.
The fact that it not only pruning/razoring.
About metod Vladimit Rybinkin. I am not familiar with his method, but I know
that some manufacturers chess
computers used something similar. Eval() function calls once in root and in
depth of variants the first not losing material move gets out.

((:
Yes. I think that it's a some modification of fail-high reduction. When analyzing commercials output on several test suites I make a conclusion that a lot of them reducts depth in nodes which is not leaf/pre-leaf/pre-pre-leaf. The main idea that a lot of quite moves can't return lost material. There are many ways to imlement this strategy. SmarThink uses my own "history-based pruning".
Yes, it's one of possible ways.
Best wishes,
Sergei



http://www.aigroup.narod.ru

[Sergei S. Markoff]

Geschrieben von: / Posted by: Sergei S. Markoff at 01 July 2003 14:02:12:
Als Antwort auf: / In reply to: first game against SmarThink geschrieben von: / posted by: Vladimir Yelin at 01 July 2003 12:37:10:

[White "SmarThink 0.16b[eta]++"]
[Black "Kaissa2 2.0"]

Classical case (; SmarThink 0.17 easily understanding the critical position with goal on h1.
It's interesting when Kaissa2 found winning score.



http://www.aigroup.narod.ru

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 01 July 2003 16:06:47:
Als Antwort auf: / In reply to: Re: first game against SmarThink geschrieben von: / posted by: Sergei S. Markoff at 01 July 2003 14:02:12:

[White "SmarThink 0.16b[eta]++"]
[Black "Kaissa2 2.0"]

Classical case (; SmarThink 0.17 easily understanding the critical position with goal on h1.
It's interesting when Kaissa2 found winning score.

I guess that you mean goal Qh1#

Movei believes that kaissa had a very bad position
instead of Bxh6 Be3 was very good for white
depth=12 +2.84 c1e3 c4b3 f6h7 e7e8 h4h6 h8h7 h6h7 d8d5 h7g6 e8d7 f2f7 f8e7 f7e7 d7e7 g6g7 e7e8 g7h8 e8d7 h8h7 d7c8 h7g8 c8c7 g8f7 c7b8 f7f8 b7c8
Nodes: 274872018 NPS: 189152
Time: 00:24:13.18
Later Bxf8 was probably a blunder and Nf6+ seems to be the right move.

r4bNr/1b1k4/p2qp2B/1p3n1Q/2pP2P1/1PP5/1P3R1P/R5K1 w - - 0 27
am Bxf8
6 3 218 410457 h5f7 f8e7 h6f4 a8g8 f4d6 g8g4 g1f1 f5d6
7 -138 654 1230274 h5f7 f8e7 g8f6 d7d8 h6g5 f5e3 g5e3 d6g3 g1f1 g3e3 b3c4 b5c4
7 -137 670 1258925 h6f8
7 -108 679 1273327 h6f8
7 -27 696 1302342 h6f8 d6f8 h5h8 f8g8 h8h5 c4b3 g4g5 d7d8
7 -26 1356 2556156 g8f6
7 -9 1505 2824382 g8f6 d7c7 h6f4 h8h5 f4d6 f5d6 f6h5 c4b3 h5f4 b7d5
7 -9 1597 3013222 g8f6 d7c7 h6f4 h8h5 f4d6 f5d6 f6h5 c4b3 h5f4 b7d5
8 14 2028 3798416 g8f6 d7c6 d4d5 c6b6 h6f4 d6f4 f2f4 h8h5 g4h5 f8c5 g1h1 c5e3
8 14 2417 4571330 g8f6 d7c6 d4d5 c6b6 h6f4 d6f4 f2f4 h8h5 g4h5 f8c5 g1h1 c5e3
After Bxf8 it is not hard to find Rxf8 with a winning score.
Movei version with no goals(it is something that I may add in the future but I have more important things to do first and I plan to add knowledge to movei to detect fast information about tactics like forks(the attack tables that I have today may help me but they are not good enough)):
8 -9 108 175614 d6f8 h5h8 f8g8 h8g8 a8g8 h2h3 c4b3 a1a3 d7d6 a3b3
8 -8 194 319050 a8f8
8 21 245 398473 a8f8
8 70 369 597626 a8f8 h5h8 d6c6 h8h7 d7d8 d4d5 c6d5 g1f1 d5h1 f1e2 h1a1 h7b7 a1b
2 e2f1
8 70 426 692113 a8f8 h5h8 d6c6 h8h7 d7d8 d4d5 c6d5 g1f1 d5h1 f1e2 h1a1 h7b7 a1b
2 e2f1
9 204 1343 2331442 a8f8 h5h8 d6d5 h8h7 d7d8 h7b7 d5b7 g8h6 f8h8 h6f5 e6f5 f2f5
c4b3
9 204 1431 2487597 a8f8 h5h8 d6d5 h8h7 d7d8 h7b7 d5b7 g8h6 f8h8 h6f5 e6f5 f2f5
c4b3
10 280 2874 5131425 a8f8 h5h8 d6c6 d4d5 c6d5 h8h7 d7e8 h7g6 e8d8 g6g5 d8c7 f2d2
d5h1 g1f2 h1a1 b3c4
10 280 3050 5432867 a8f8 h5h8 d6c6 d4d5 c6d5 h8h7 d7e8 h7g6 e8d8 g6g5 d8c7 f2d2
d5h1 g1f2 h1a1 b3c4
11 354 7180 12753963 a8f8 h5h8 d6d5 h8h7 d7e8 g8f6 f8f6 h7b7 d5b7 g4f5 e6f5 a1e
1 e8d7 b3c4 b5c4
11 354 7597 13437156 a8f8 h5h8 d6d5 h8h7 d7e8 g8f6 f8f6 h7b7 d5b7 g4f5 e6f5 a1e
1 e8d7 b3c4 b5c4
Uri

[Vladimir Yelin]

Geschrieben von: / Posted by: Vladimir Yelin at 01 July 2003 17:47:12:
Als Antwort auf: / In reply to: Re: first game against SmarThink geschrieben von: / posted by: Sergei S. Markoff at 01 July 2003 14:02:12:

[White "SmarThink 0.16b[eta]++"]
[Black "Kaissa2 2.0"]

Classical case (; SmarThink 0.17 easily understanding the critical position with goal on h1.
It's interesting when Kaissa2 found winning score.

move 22.Bh6 Kaissa at all did not consider, including a field d4 critical she
counted the best 22.Be3(Bc5 in planning)
Best.
Vladimir.