The problem of opening theory in computer chess

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: The problem of opening theory in computer chess

Post by bob »

Rolf wrote:
bob wrote:
MatsW wrote:Robert, I did not "insult" Deep Blue. One can't insult a computer. I am educated in computer science myself, and I have been very interested in opening studies. I try to address issues concerning creativity in chess, and how to strengten this aspect in view of the general problem of computerization. It is problematic.

Facts are, in the final game against Deep Blue, Kasparov made his error already in the seventh move: 7...h6?? Deep Blue did not think, but sacrificed his knight on e6 according to the book. Game over!

Then you make quotes which are not from me.
/Mats
If you are educated in computer science, you ought to be able to see that _my_ post was a response to Rolf. It really isn't that hard to follow a thread and see who is replying to who. I even quoted _his_ comments and replied to them.

So unless you are "Rolf in disguise" you are way off base...
Excuse me, Bob, but when could you begin to admit if your were mistaken? You are caught in the act of replying - provenly ! - to Mats and now you claim having responded to Rolf which is provenly WRONG!

Here is the decisive quote and please respect that Rolf didnt comment or write about Deep Blue at all! So, by consequence and simple logic you spoled it. Mats spoke about DB, not Rolf! And since you accused the one who had commented on DB, you can only have meant Mats, not Rolf.

Here is the proof:

QUOTE

Rolf wrote:
MatsW wrote:
The problem of opening theory in computer chess. What's the point in
measuring the opening book knowledge when arranging matches between
chess programs? By example, Deep Blue never got out of the opening
book before it was practically over in the final game against
Kasparov. Would the chess public stand up and applause if a human
player sat with an opening book in his lap, replicating the moves in
the book, and winning against the world champion? Of course not, they
would boo at him. I realize, of course, that it is a difficult
question, not the least because human players can play from memory as
well, although not nearly as good. Nevertheless, it is high time that
this practice is questioned in computer chess, at least in matches
between chess programs.

[now here is the direct judgement of R. Hyatt:]
Most of us greatly enjoyed our break from this kind of nonsensical post. You apparently can't write a single post without tossing an insult at Deep Blue. Deep blue. If my memory is correct, and this can be checked by looking at the DB log for game 6 for anyone interested, they were out of book around move 10. The game went to move 20 or something, The game was not "practically over" at that point.

I have seen GM players play 30 moves before they stop to think in a game. Do they cheat?


QUOTE END


Moral:

You shouldnt put yourself on such a highly patronizing false horse and insult e.g. Rolf if Mats has written what you then commented on. Thanks, Bob.

Now show us the next shot from the hip... :P
I suggest you first shut up, then look at the post by me. And look _right at the top_ where it says "Rolf wrote:"

Get it now?

My post _directly_ follows one from you. Here is the link to your post:

http://www.talkchess.com/forum/viewtopi ... 22&t=28011

Ad here is the link to my post, which _directly_ follows your post:

http://www.talkchess.com/forum/viewtopi ... 64&t=28011

And here is the top of my post which followed yours:
Bob wrote:
Rolf wrote:
MatsW wrote:The problem of opening theory in computer chess. What's the point in
measuring the opening book knowledge when arranging matches between
chess programs? By example, Deep Blue never got out of the opening
book before it was practically over in the final game against
Kasparov. Would the chess public stand up and applause if a human
player sat with an opening book in his lap, replicating the moves in
the book, and winning against the world champion? Of course not, they
would boo at him. I realize, of course, that it is a difficult
question, not the least because human players can play from memory as
well, although not nearly as good. Nevertheless, it is high time that
this practice is questioned in computer chess, at least in matches
between chess programs.


Most of us greatly enjoyed our break from this kind of nonsensical post. You apparently can't write a single post without tossing an insult at Deep Blue. Deep blue. If my memory is correct, and this can be checked by looking at the DB log for game 6 for anyone interested, they were out of book around move 10. The game went to move 20 or something, The game was not "practically over" at that point.
So, my conclusion is, whatever you are on, you need to take _less_ of, and whatever you are not on, you need more of it. My post is directly addressed to you, not to anyone else. It directly follows your post, not anybody elses. And it directly quotes your post, in addition to the one you replied to.

If you can't follow that, then perhaps you should go back to where you have been for the past month or two and stay there..
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: The problem of opening theory in computer chess

Post by bob »

Uri Blass wrote:
MatsW wrote:Robert, I did not "insult" Deep Blue. One can't insult a computer. I am educated in computer science myself, and I have been very interested in opening studies. I try to address issues concerning creativity in chess, and how to strengten this aspect in view of the general problem of computerization. It is problematic.

Facts are, in the final game against Deep Blue, Kasparov made his error already in the seventh move: 7...h6?? Deep Blue did not think, but sacrificed his knight on e6 according to the book. Game over!

Then you make quotes which are not from me.
/Mats
This is not correct

Deep blue used time to play the moves.
You can see Deep blue log files.
You have to be careful here. DB always "thinks" when waiting on the opponent. I assume they use the same algorithm Murray added to HiTech, and which I also use in Crafty. Namely when it is my move, I look it up in the book and then make it. I then switch sides, find the set of known book moves, and remove those from the set of legal moves and then search that sub-set to find the best "non-book" move my opponent can play. I then use that to ponder so that I don't waste time pondering something where I have an immediate book response, but instead ponder a likely opponent move if he is now out of book.

In looking at the copy of the DB logs I have, it appears that move 10 by DB was book, move 11 was not although I looked very quickly. And DB did not think the game was over, having an eval of something like +0.60 or so... It appears it might "understand" this opening issue better than Rybka based on the analysis you gave by Rybka.



It is also not clear that 7...h6 is losing

Some analysis after 8.Nxe6 by rybka3
It may be interesting if somebody give multi-rybka to analyze it for many hours so we can know better and I do not trust humans to know the truth.

New game - so k, 120'/40
[D]r1bqkb1r/pp1n1pp1/2p1Nn1p/8/3P4/3B1N2/PPP2PPP/R1BQK2R b KQkq - 0 1

Analysis by Rybka 3 1-cpu 32-bit :

8...Qd8-e7
= (0.24) Depth: 2 00:00:00
8...f7xe6
= (0.06) Depth: 2 00:00:00
8...f7xe6
= (-0.02) Depth: 3 00:00:00
8...f7xe6
= (0.17) Depth: 4 00:00:00
8...f7xe6 9.Bd3-g6+
= (0.18) Depth: 5 00:00:00
8...Qd8-e7 9.0-0
= (0.01) Depth: 5 00:00:00
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.Bc1-f4 Nf6-d5
= (0.16) Depth: 6 00:00:00 2kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.Bc1-f4 Nf6-d5 12.Bf4-g3
= (0.18) Depth: 7 00:00:00 5kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.Bc1-f4 b7-b6 12.c2-c4 Bc8-b7
= (0.19) Depth: 8 00:00:00 15kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.Bc1-f4 Nf6-d5 12.Bf4-g3 Qe7-b4 13.Qd1-e2
= (0.19) Depth: 9 00:00:01 26kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.c2-c4 Qe7-b4 12.Qd1-e2 Kd8-c7 13.Bc1-f4+ Bf8-d6 14.a2-a3
² (0.34) Depth: 10 00:00:04 109kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.c2-c4 Qe7-b4 12.Qd1-e2 b7-b6 13.Bc1-f4 Bc8-a6 14.b2-b3 Qb4-e7
² (0.26) Depth: 11 00:00:06 165kN
8...Qd8-e7 9.0-0 f7xe6 10.Bd3-g6+ Ke8-d8 11.c2-c4 b7-b6 12.Bc1-f4 Bc8-b7
² (0.38) Depth: 12 00:00:12 334kN
8...f7xe6 9.Bd3-g6+ Ke8-e7
² (0.27) Depth: 12 00:00:20 540kN
8...f7xe6 9.Bd3-g6+ Ke8-e7 10.0-0 Qd8-c7
² (0.27) Depth: 13 00:00:27 708kN
8...f7xe6 9.Bd3-g6+ Ke8-e7 10.0-0 Qd8-c7 11.Rf1-e1 Ke7-d8 12.Bc1-d2 Bf8-e7 13.c2-c4 Nd7-f8 14.Bg6-d3 Bc8-d7 15.Nf3-e5 Kd8-c8
= (0.25) Depth: 14 00:01:05 1789kN
8...f7xe6 9.Bd3-g6+ Ke8-e7 10.0-0 Qd8-c7 11.c2-c4 Ke7-d8 12.g2-g3 Bf8-e7 13.Bc1-f4 Qc7-a5 14.Qd1-d3 Nd7-f8 15.Nf3-e5 Nf8xg6
² (0.36) Depth: 15 00:01:53 2997kN
8...f7xe6 9.Bd3-g6+ Ke8-e7 10.0-0 Qd8-c7 11.c2-c4 Ke7-d8 12.g2-g3 Bf8-e7 13.Bc1-f4 Qc7-a5 14.Qd1-d3 Nd7-f8 15.Nf3-e5 Nf8xg6
² (0.36) Depth: 16 00:03:37 5869kN
8...f7xe6 9.Bd3-g6+ Ke8-e7 10.0-0 Qd8-c7 11.c2-c4 Ke7-d8 12.g2-g3 Bf8-e7 13.Bc1-f4 Qc7-a5 14.a2-a3 Nd7-f8 15.b2-b4 Qa5-a6 16.Bg6-d3
= (0.22) Depth: 17 00:06:10 10232kN

(so k, 24.05.2009)

Uri
User avatar
Graham Banks
Posts: 41455
Joined: Sun Feb 26, 2006 10:52 am
Location: Auckland, NZ

Re: The problem of opening theory in computer chess

Post by Graham Banks »

bob wrote:
Graham Banks wrote:
Marc Lacrosse wrote: One day we could perfectly have an engine who is the best in the world because it has found a way to always win playing 1.f4 while its result are poor with 1.e4 or 1.d4. As it may choose to allways play 1.f4 it will win every championship.
The engine can only play what is programmed into it, so it is not effectively choosing to play openings as humans do (not yet anyway).

Cheers,
Graham.
I'm not sure what this means. Probably the best "book learner" around was mchess pro. Marty did the usual book-learning stuff to learn which moves were good and which were bad. But then after finishing a game, he went back and added to this book by using the moves played in the game just finished, so that the program would "learn" to extend the book by playing moves where it won, and avoid playing moves that led to a loss...
Yes - I forgot about the learning feature.
The problem with book learning though is that it's quite often a whole line of continuation that's faulty rather than a single move. The single move that the programme culls might in fact be perfectly okay if it preceded the faulty continuation, or alternatively it could be that an earlier move or moves were the ones at fault.
Hope all that makes sense?
However, what Marc says could in fact happen sometime in the future. It's just that such sophistication would seem to be a long way off.
gbanksnz at gmail.com
Uri Blass
Posts: 10297
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: The problem of opening theory in computer chess

Post by Uri Blass »

bob wrote: You have to be careful here. DB always "thinks" when waiting on the opponent. I assume they use the same algorithm Murray added to HiTech, and which I also use in Crafty. Namely when it is my move, I look it up in the book and then make it. I then switch sides, find the set of known book moves, and remove those from the set of legal moves and then search that sub-set to find the best "non-book" move my opponent can play. I then use that to ponder so that I don't waste time pondering something where I have an immediate book response, but instead ponder a likely opponent move if he is now out of book.

In looking at the copy of the DB logs I have, it appears that move 10 by DB was book, move 11 was not although I looked very quickly. And DB did not think the game was over, having an eval of something like +0.60 or so... It appears it might "understand" this opening issue better than Rybka based on the analysis you gave by Rybka.
My comments:
1)Based on the log file Deep blue clearly pondered about book moves and not about the best non book moves.
for example it pondered about 1.e4 c5 when kasparov played 1...c6



2)Deep blue evaluated the position after 8...fxe6 that kasparov did not play(kasparov chose 8...Qe7) as equal after 22 seconds

Here is the relevant part from the log file:
8(6) #[Bg6](0) 0 T=17
bd3g6 Ke8e7 o-o Pb7b5 rf1e1 Nf6d5 qd1e2 Nd7f6 nf3e5

and later:
0 T=22
bd3g6 Ke8e7 o-o Pb7b5 rf1e1 Nf6d5 qd1e2 Ke7d6 nf3e5 Nd7e5n qe2e5N Kd6e7 bc1d2 Qd8b6 pb2b4 Qb6c7 qe5h5

Deep blue also evaluated the position after move 9...fxe6 of the game as better for black

7(5) #[Bg6](-10) -10 T=4
bd3g6 Ke8d8 rf1e1 Kd8c7 bc1f4 Kc7b6 pc2c4 Pa7a6 pc4c5 Kb6a7 bg6f5

8(6)<ch> 'fe'
[0 sec (main.c:6589)] -10
T=13
bd3g6 Ke8d8 rf1e1 Kd8c7 bc1f4 Kc7b6 pc2c4 Pa7a6 ra1b1 Kb6a7 pb2b4
---------------------------------------
-->
10. Bg6 <-- 30/110:39




3)The eval of deep blue out of book at move 11 was 0.32 pawns for white
based on this part of the logfile:

9(6)[TIMEOUT] 32
T=190
bc1f4 Nf6e8 bf4g3 Ne8d6 rf1e1 Pa7a5 re1e3 Kd8c7 qd1e1 Qe7d8 pc2c4 Pc6c5 re3e6P Ra8a6

Rybka suggests at this point 11.c4

so k - Rybka 3 1-cpu 32-bit, Friend mode
r1bk1b1r/pp1nq1p1/2p1pnBp/8/3P4/5N2/PPP2PPP/R1BQ1RK1 w - - 0 1

Analysis by Rybka 3 1-cpu 32-bit :

11.Bc1-f4
= (0.06) Depth: 2 00:00:00
11.Bc1-f4
= (0.18) Depth: 3 00:00:00
11.Bc1-f4
= (0.10) Depth: 4 00:00:00
11.Bc1-f4
² (0.30) Depth: 5 00:00:00
11.Bc1-f4 Nf6-d5
= (0.15) Depth: 5 00:00:00
11.Bc1-f4 b7-b6 12.Qd1-e2
² (0.26) Depth: 6 00:00:00 2kN
11.Bc1-f4 b7-b6 12.Qd1-e2 Bc8-b7
² (0.29) Depth: 7 00:00:00 4kN
11.Bc1-f4 b7-b6 12.Qd1-e2 Bc8-b7 13.c2-c4
² (0.32) Depth: 8 00:00:00 10kN
11.c2-c4 b7-b6 12.Bc1-f4 Bc8-b7 13.Ra1-c1 Kd8-c8 14.d4-d5
² (0.28) Depth: 9 00:00:02 61kN
11.c2-c4 Qe7-b4 12.Qd1-e2 b7-b5
² (0.35) Depth: 10 00:00:05 157kN
11.c2-c4 b7-b6 12.Bc1-f4 Nf6-h7
² (0.35) Depth: 11 00:00:15 405kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 b7-b6 14.Rf1-e1 Bc8-b7 15.a2-a4
² (0.33) Depth: 12 00:00:22 617kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-d3 Nd7-b6 15.Ra1-e1 Bc8-d7
² (0.44) Depth: 13 00:00:33 915kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-e2 Nd7-b6 15.h4-h5 Rh8-g8 16.Nf3-d2
² (0.44) Depth: 14 00:00:53 1461kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-e2 Nd7-b6 15.h4-h5 Bc8-d7 16.Nf3-d2 e6-e5
² (0.48) Depth: 15 00:02:55 4810kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-e2 Nd7-b6 15.h4-h5 Bc8-d7 16.Nf3-d2 Kd8-c8 17.a2-a4 a7-a5 18.Rf1-d1 Ng4-f6 19.Nd2-b3 Nf6-e8
² (0.47) Depth: 16 00:04:57 7681kN
11.c2-c4 Nf6-g8 12.Qd1-d3 Qe7-d6
² (0.46) Depth: 17 00:09:58 16520kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-d3 Qe7-b4 15.Qd3-c2 Bf8-e7 16.a2-a3 Qb4-a5 17.b2-b4 Qa5-a6 18.b4-b5 Qa6-b6 19.b5xc6
² (0.52) Depth: 18 00:28:58 51101kN
11.c2-c4 Nf6-h7 12.h2-h4 Nh7-f6 13.Bc1-f4 Nf6-g4 14.Qd1-d3 Qe7-b4 15.Qd3-c2 Bf8-e7 16.a2-a3 Qb4-a5 17.b2-b4 Qa5-a6 18.b4-b5 Qa6-b6 19.b5xc6
² (0.52) Depth: 19 00:36:55 64011kN

(so k, 24.05.2009)
User avatar
Rolf
Posts: 6081
Joined: Fri Mar 10, 2006 11:14 pm
Location: Munster, Nuremberg, Princeton

Re: The problem of opening theory in computer chess

Post by Rolf »

bob wrote: So, my conclusion is, whatever you are on, you need to take _less_ of, and whatever you are not on, you need more of it. My post is directly addressed to you, not to anyone else. It directly follows your post, not anybody elses. And it directly quotes your post, in addition to the one you replied to.

If you can't follow that, then perhaps you should go back to where you have been for the past month or two and stay there..

Perhaps another violation of the charter here is it when someone (Prof. Hyatt!) had it _all_ confused, content and author wise, and to then imply with doctored quotes that the one author (Mats) could well be Rolf in disguise.
-Popper and Lakatos are good but I'm stuck on Leibowitz
Sven
Posts: 4052
Joined: Thu May 15, 2008 9:57 pm
Location: Berlin, Germany
Full name: Sven Schüle

Re: The problem of opening theory in computer chess

Post by Sven »

Rolf wrote:
bob wrote: So, my conclusion is, whatever you are on, you need to take _less_ of, and whatever you are not on, you need more of it. My post is directly addressed to you, not to anyone else. It directly follows your post, not anybody elses. And it directly quotes your post, in addition to the one you replied to.

If you can't follow that, then perhaps you should go back to where you have been for the past month or two and stay there..

Perhaps another violation of the charter here is it when someone (Prof. Hyatt!) had it _all_ confused, content and author wise, and to then imply with doctored quotes that the one author (Mats) could well be Rolf in disguise.
As a neutral observer in this case, I can only state that it has been trivial for me to prove - for instance by using thread view - that Bob had indeed responded to you, Rolf. Also it is obvious for me that Bob has done so by intent, so I do not see any point in your claim that Bob had confused content and/or author of some post in this thread.

I propose that you cross-check this first, and then come back to tell us whether my observation is correct. In case you accept I'd assume that your conclusions were without cause.

Sven
Sven
Posts: 4052
Joined: Thu May 15, 2008 9:57 pm
Location: Berlin, Germany
Full name: Sven Schüle

Re: The problem of opening theory in computer chess

Post by Sven »

Actually the impression that Bob would have mixed something up might have come from this little technical fault he made by not closing his quoting brackets before starting his own reply. But I consider this a minor issue since I think is trivial for any experienced reader of this board to discover that.

It is of course possible that this technical flaw had caused Mats to get confuse d about Bob's post himself.

Therefore my proposal to Bob would be to try taking more care about the exact "quoting syntax".

Sven
Sven
Posts: 4052
Joined: Thu May 15, 2008 9:57 pm
Location: Berlin, Germany
Full name: Sven Schüle

Re: The problem of opening theory in computer chess

Post by Sven »

Sven Schüle wrote:Actually the impression that Bob would have mixed something up might have come from this little technical fault he made by not closing his quoting brackets before starting his own reply. But I consider this a minor issue since I think is trivial for any experienced reader of this board to discover that.

It is of course possible that this technical flaw had caused Mats to get confuse d about Bob's post himself.

Therefore my proposal to Bob would be to try taking more care about the exact "quoting syntax".

Sven
"Edit" (past 15 minutes :evil: ):
"... since I think it is trivial ..."
"... to get confused about ..."

Sven
Adam Hair
Posts: 3226
Joined: Wed May 06, 2009 10:31 pm
Location: Fuquay-Varina, North Carolina

Re: The problem of opening theory in computer chess

Post by Adam Hair »

bob wrote:
I suggest you first shut up, then look at the post by me. And look _right at the top_ where it says "Rolf wrote:"

Get it now?

My post _directly_ follows one from you. Here is the link to your post:

http://www.talkchess.com/forum/viewtopi ... 22&t=28011

Ad here is the link to my post, which _directly_ follows your post:

http://www.talkchess.com/forum/viewtopi ... 64&t=28011

And here is the top of my post which followed yours:
Bob wrote:
Rolf wrote:
MatsW wrote:The problem of opening theory in computer chess. What's the point in
measuring the opening book knowledge when arranging matches between
chess programs? By example, Deep Blue never got out of the opening
book before it was practically over in the final game against
Kasparov. Would the chess public stand up and applause if a human
player sat with an opening book in his lap, replicating the moves in
the book, and winning against the world champion? Of course not, they
would boo at him. I realize, of course, that it is a difficult
question, not the least because human players can play from memory as
well, although not nearly as good. Nevertheless, it is high time that
this practice is questioned in computer chess, at least in matches
between chess programs.


Most of us greatly enjoyed our break from this kind of nonsensical post. You apparently can't write a single post without tossing an insult at Deep Blue. Deep blue. If my memory is correct, and this can be checked by looking at the DB log for game 6 for anyone interested, they were out of book around move 10. The game went to move 20 or something, The game was not "practically over" at that point.
So, my conclusion is, whatever you are on, you need to take _less_ of, and whatever you are not on, you need more of it. My post is directly addressed to you, not to anyone else. It directly follows your post, not anybody elses. And it directly quotes your post, in addition to the one you replied to.

If you can't follow that, then perhaps you should go back to where you have been for the past month or two and stay there..
Clearly the confusion here is who were you directing this statement
towards:

" Most of us greatly enjoyed our break from this kind of nonsensical post. You apparently can't write a single post without tossing an insult at Deep Blue. Deep blue. If my memory is correct, and this can be checked by looking at the DB log for game 6 for anyone interested, they were out of book around move 10. The game went to move 20 or something, The game was not "practically over" at that point.

I have seen GM players play 30 moves before they stop to think in a game. Do they cheat? "

In the quote box above, it is clear that you were responding to Rolf.
Yet, I have to say that your post
http://www.talkchess.com/forum/viewtopi ... 64&t=28011
seems to imply that Rolf wrote that statement. It is inside his quote
box, not outside.
Adam Hair
Posts: 3226
Joined: Wed May 06, 2009 10:31 pm
Location: Fuquay-Varina, North Carolina

Re: The problem of opening theory in computer chess

Post by Adam Hair »

Sven Schüle wrote:Actually the impression that Bob would have mixed something up might have come from this little technical fault he made by not closing his quoting brackets before starting his own reply. But I consider this a minor issue since I think is trivial for any experienced reader of this board to discover that.

It is of course possible that this technical flaw had caused Mats to get confuse d about Bob's post himself.

Therefore my proposal to Bob would be to try taking more care about the exact "quoting syntax".

Sven
The use of the preview function would help in such situations.