Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / How opening books have destroyed computer chess
- - By Ray (****) Date 2010-02-22 07:38 Edited 2010-02-22 07:42
http://www.talkchess.com/forum/viewtopic.php?t=32801&start=30

Hiarcs vs Sjeng CCT12

Both engines in book up to move 50. How on earth do you call that a computer match ? The engines only started to think on move 51. In my humble opinion, that's crazy.
Roll on Mainz and chess960. Conventional chess tournaments are of no interest to me when you get this happening
Parent - By Kreuzfahrtschiff (***) Date 2010-02-22 10:43 Edited 2010-02-22 10:46
thats not crazy, thats stupid, they should play another move before

or do u think all moves were right?

or do "you think that’s air you’re breathing now?"  :)
Parent - - By Sedat Canbaz (*****) Date 2010-02-22 10:55 Edited 2010-02-22 11:05
Hello Ray,

Both engines in book up to move 50. How on earth do you call that a computer match ?
Maybe after 50th move ... the both engines are started to use the endgames :)

Anyway in my opinion ( in case of running  a such important tournament),only the book's winning games should be allowed to play up to 50 or 100 moves book depth
The draw games should be imported up to 20-30 moves book depth

Note:not only the hardware speed,but also the opening book plays a very important role to destroy the opponent
(of course it depends on what kind of book you are using)

Regards,
Sedat
Parent - By Harvey Williamson (*****) Date 2010-02-22 11:30 Edited 2010-02-22 11:43

> not only the hardware speed,but also the opening book plays a very important role to destroy the opponent


I would not mind a book limit. Although books are cheap, but time consuming to prepare, and add a level of skill. Hardware is expensive and not all can afford the best.
Parent - - By BrandonSi (***) Date 2010-02-22 13:48
I agree with you about the draw games. This is a little off-topic, but many people do not include opening lines from games that resulted in draws in their books. I've never understood that. I do agree that you obviously would not want to import the first 50 moves of a drawn game, but the first 20 or so can be very beneficial. If it happened to be a lesser played opening, like A03, and a bookmaker ignored this game because of a draw, it's very possible that the book may not even have data past move 8 or 9. Better to have data from a drawn game in your book than no data at all.

Just my opinion. :)
Parent - - By Sedat Canbaz (*****) Date 2010-02-22 15:26
I agree with you about the draw games...
I recommend to not give many hints...,otherwise your Uptown book will be not classified at the top in SCCT :)

Regards,
Sedat
Parent - - By BrandonSi (***) Date 2010-02-22 16:18
It's OK. Information should be free. I would rather teach others and come in last place, than keep all the secrets and win. :)
Parent - - By Nelson Hernandez (Gold) Date 2010-02-22 19:10
Oh man, that is 180 degrees contrary to my philosophy!  :)
Parent - - By BrandonSi (***) Date 2010-02-22 19:25
Understandable! I'm not saying one ideal is right, and one is wrong, they're both fine. It's a personal preference.

It's just my opinion that more competition results in better books for everyone. The only way to increase competition is to lower the barrier to entry, in this case, knowledge/information.

I'm considering making a video on hand tuning an opening book and putting it on youtube. Just thought I'd give everyone fair warning so you can get the assassins in place before I let out too many secrets. :)
Parent - - By Uly (Gold) Date 2010-02-22 19:28

> Understandable! I'm not saying one ideal is right, and one is wrong, they're both fine.


But with Nelson's you get many more wins! :)
Parent - By BrandonSi (***) Date 2010-02-22 19:47
This is most certainly true. :)
Parent - - By Highendman (****) Date 2010-02-22 20:00
LOL. I belong to your camp too.

Though I have often stated, in other disciplines of life, that information is overrated.
Basically in many domains all or many have access to the same information - what makes winners is knowing what to do with it, and having the will to execute repeatedly.
Parent - By NATIONAL12 (Gold) Date 2010-02-22 23:38
Basically in many domains all or many have access to the same information - what makes winners is knowing what to do with it(HEM)

agree.
Parent - By Dr.Wael Deeb (***) Date 2010-02-22 16:59
Fully agreed here....
Dr.D
Parent - By InspectorGadget (*****) Date 2010-02-22 18:54

> Both engines in book up to move 50. How on earth do you call that a computer match ?
> Maybe after 50th move ... the both engines are started to use the endgames :-)


That's a good one Sedat, it might as well happen :)
Parent - - By M ANSARI (*****) Date 2010-02-22 11:48
I agree, there should be some limit over this.  I will admit though that books have been incredibly important in engine match results since inception of engine tournaments.  Maybe a 15 move limit should be implemented, or maybe a maximum size of book.  I bet it would be very hard to verify though as circumventing the checks would seem very easy to do ... maybe that is why there is no limit on this in the first place.
Parent - By Sedat Canbaz (*****) Date 2010-02-22 12:00
Maybe a 15 move limit should be implemented, or maybe a maximum size of book.
Normally,yes...its a very good idea that all games to be run on identical conditions and not only the book move limit should be implemented,but also the hardware's  speed must be identical too

Best,
Sedat
Parent - By Dr.Wael Deeb (***) Date 2010-02-22 13:52
Opening book home preparation is essential....I prepared Spark's opening book for the CCT12 and when it played against Hiarcs,which is famous for it's fine tuned opening book generaly,Hiarcs didn't come out of book so healthy and it managed only to draw Spark....so Spark's opening book did cost Hiarcs additional half point which could make the outcome of the event different for Hiarcs....
Dr.D
Parent - By rookie (**) Date 2010-02-22 15:04
I agree. Openings analysed out to 40 or 50 moves are not so much openings any more as they are pre-played games. Fischer, crazy as he was, had it right when he said something to the effect that opening theory had ruined what he came to think of as 'old chess'. (Yes, I know, partly responsible for his tremendous scores was his opening preparation.) As much as I love computer chess, this is the one thing I regret about it--how easy it now is to advance opening theory to the absurd point that you noted.
Parent - - By Uly (Gold) Date 2010-02-22 16:07
I think these tournaments should be about the strongest chess entities available, this includes opening books. Too few games are played anyway.

Do you want to know the engine's strengths? Look at the rating lists, they do take measures so the book effects should be minimized.
Parent - - By josias mauricio (**) Date 2010-02-22 17:41
Vytron ... you make reference of the game that Petrosian x Korchnoi 1963 moscow
movement 35.Rxh6??
Parent - - By Uly (Gold) Date 2010-02-22 18:05
Heh, it's an obscure joke, in context it said that when you're on an OTB game you don't usually think "‘This reminds me of that Petrosian-Korchnoi game from ’63 where grabbing the pawn proved fatal’", you just recall the patterns, in the article I read it I found it hilarious and had to put it in my sig :), but probably your reference is correct.
Parent - By josias mauricio (**) Date 2010-02-22 18:15
pero son correctas en la puesta.
el juego en cuestión es el único y realmente capturar el movimiento
de los peatones en mov 35 llevó a la derrota de los blancos.:-)

http://www.chessgames.com/perl/chessgame?gid=1081527
Parent - - By Nelson Hernandez (Gold) Date 2010-02-22 19:21
I totally disagree, Ray.  Perfection, in my opinion, is a game that you win without the engine ever being engaged.  Note that someone has to win.  In that case the winner outwitted the loser and applied just about the worst kind of humiliation imaginable in a tournament, apart from not having the right EGTBs in the endgame and turning a win to a loss (a very rare event).

However, two books playing 50 moves in book to a draw, that's just mutually-assured stupidity in an engine tournament.  One of the two books, at least, should be able to stop this foolish outcome.  Both players ought to be red-faced after this game.  This is just sloppiness in the game settings.  You can use 0/100 in an engine match if you want and are using .ctg, but not in a big public tournament!  It has to be at least 1/100.

But as far as imposing rules limiting books, that's idiocy as well.  There is lots of theory even among human players that extends out 20, 25 moves.  God help you if you run out of moves in a Marshall or a Poisoned Pawn, just to name a few.
Parent - - By Nelson Hernandez (Gold) Date 2010-02-22 19:50
If I may argue with myself a bit, in the case in question there can be no doubt that the Sjeng side was clearly the more deficient in terms of book strategy.  What happened may well have been the optimal outcome for the Hiarcs team--a low-risk draw--given the huge disparity in hardware between the two.  The Rybka book team (Jiri, and before him Nick and Jeroen) have got the right idea for tournament play.  You need to let superior hardware bring its weight to bear as early as possible so that it will find superior, novelty lines and grind out the win.  In other words you avoid theory and avoid drawing lines as much as you can without unduly jeopardizing the position.  When you are at a major hardware disadvantage you do the opposite: stay in book and follow the most drawing line you can. 

So in these terms Hiarcs was clever and wise, and Sjeng dropped the ball.
Parent - By BrandonSi (***) Date 2010-02-22 19:56 Edited 2010-02-22 20:00
Absolutely Nelson.. Sometimes a draw is a win. A Pentium III playing against a 200 core cluster resulting in a draw is a "win" in my book for the Pentium III.

This appears to be what happened for the Hiarcs team (though obviously not that exaggerated).

As much as I hate to quote Sun Tzu, this point is very clear.

"If your enemy is secure at all points, be prepared for him. If he is in superior strength, evade him"

In this case, evading means letting Sjeng think as little as possible.
Parent - - By M ANSARI (*****) Date 2010-02-22 22:20
A point to mention is that this book line which has been played out in Playchess quite often is based mostly on Rybka analysis and how Rybka 3 evaluates the position.  When using such books that use only Rybka vs Rybka evaluations to build, you run the risk a non Rybka engine not "understanding" the position when out of book and thus playing inferior to what it would have played with a book optimized for that particular engine.  Harvey is pretty well versed with the latest Playchess theory and so is Suj who was working with GCP on DS, so it is no surprise that they had similar book lines.  The thing is that when Rybka 4 comes out, all those books will become obsolete in one day ... at least that is what happened last time.
Parent - - By Nelson Hernandez (Gold) Date 2010-02-22 23:19
Not quite obsolete, but certainly there will be a revolution of sorts.  That always happens.  This is why I collect games from everywhere.  Rybka has no monopoly on ideas, and what may appear to be a mediocre move to Rybka often turns out to be a winning line for some other engine.
Parent - By Uly (Gold) Date 2010-02-22 23:33
This is specially true on correspondence games.
Parent - - By Uly (Gold) Date 2010-02-22 20:13
Which game? It's really difficult to find it on the provided link.

[Event "CCT12 Blitz Event"]
[Site "FICS"]
[Date "2010.02.21"]
[Round "7"]
[White "Hiarcs"]
[Black "SjengX"]
[Result "1/2-1/2"]
[WhiteElo "2627"]
[BlackElo "2641"]
[ECO "B97"]
[TimeControl "300+3"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Bg5 e6 7.f4 Qb6 8.Qd2
Qxb2 9.Rb1 Qa3 10.e5 h6 11.Bh4 dxe5 12.fxe5 g5 13.exf6 gxh4 14.Ne4 Qxa2
15.Rd1 Bd7 16.Be2 h3 17.O-O Qd5 18.Qe3 Qe5 19.Rf3 Nc6 20.Rxh3 Nxd4 21.Rxd4
Rg8 22.Rh5 Rg5 23.Rh3 Rf5 24.g4 Rf4 25.Rh5 Qc7 26.Rhd5 Bc6 27.Nd6+ Bxd6
28.Rxd6 Rxd4 29.Rxd4 Rd8 30.Rxd8+ Qxd8 31.Qxh6 Qd4+ 32.Kf1 Bd5 33.g5 Qf4+
34.Ke1 Qb4+ 35.Kd1 Qd4+ 36.Kc1 Qe3+ 37.Kd1 Be4 38.h4 Qc3 39.Bd3 Bxd3 40.
cxd3 Qxd3+ 41.Ke1 Qe4+ 42.Kf2 e5 43.h5 Qf4+ 44.Ke2 Qh2+ 45.Kf3 Qh1+ 46.Kg4
Qg1+ 47.Kf5 Qe3 48.Kg4 Qf4+ 49.Kh3 Qe3+ 50.Kg4 Kd7 51.Qh7 Qg1+ 52.Kf3 Qf1+
53.Kg3 Qf4+ 54.Kh3 Ke6 55.g6 Qf5+ 56.Kg3 Qd3+ 57.Kg4 Qe2+ 58.Kg3 Qe1+ 59.
Kg4 Qg1+ 60.Kf3 Qd1+ 61.Kf2 Qd4+ 62.Ke2 Qc4+ 63.Ke3 Qf4+ 64.Ke2 Qe4+ 65.
Kf2 Qc2+ 66.Kf1 Qc4+ 67.Kf2 Qa2+ 68.Ke3 Qa3+ 69.Ke4 Qb4+ 70.Kf3 Qc3+ 71.
Kg4 Qc4+ 72.Kg3 Qb3+ 73.Kg4 Qa4+ 74.Kf3 Qc6+ 75.Ke3 Qc5+ 76.Ke4 Qd4+ 77.
Kf3 e4+ 78.Kf4 Qxf6+ 79.Kxe4 Qh4+ 80.Kd3 Qg3+ 81.Ke4 Qe5+ 82.Kf3 Qf5+ 83.
Ke3 Qg5+ 84.Ke4 Qd5+ 85.Kf4 Qc4+ 86.Kg5 Qc1+ 87.Kg4 Qg1+ 88.Kf3 Qd1+ 89.
Ke3 Qb3+ 90.Kd4 Qd5+ 91.Ke3 Qc5+ 92.Kd2 Qb4+ 93.Kc2 Qa4+ 94.Kc3 Qa5+ 95.
Kc4 Qd5+ 96.Kc3 Qf3+ 97.Kb4 Qf4+ 98.Ka3 Qd6+ 99.Ka2 Qh2+ 100.Kb1 Qg1+ 101.
Kc2 Qf2+ 102.Kd3 Qf3+ 103.Kd2 Qf4+ 104.Kd3 Qf1+ 105.Ke4 Qc4+ 106.Kf3 Qc3+
107.Kg4 Qd4+ 108.Kg5 Qd2+ 109.Kg4 Qd1+ 110.Kf4 Qd6+ 111.Ke4 Qc6+ 112.Kd4
Qb6+ 113.Kc4 Qb5+ 114.Kd4 Qb4+ 115.Kd3 Qb1+ 116.Kc3 Qc1+ 117.Kd4 Qg1+ 118.
Ke4 Qb1+ 119.Kd4 Qb2+ 120.Ke4 Qg2+ 121.Kf4 Qd2+ 122.Kg3 Qe3+ 123.Kg2 Qg5+
124.Kf3 Qf6+ 125.Ke2 Qb2+ 126.Kd3 fxg6 127.hxg6 Qb5+ 128.Ke3 Qf5 129.Qxb7
Qxg6 130.Qxa6+ Kf7 131.Qxg6+ Kxg6 1/2-1/2

Or

[Event "CCT12"]
[Site "FICS"]
[Date "2010.02.20"]
[Round "5"]
[White "Hiarcs"]
[Black "SjengX"]
[Result "1/2-1/2"]
[WhiteElo "2389"]
[BlackElo "2570"]
[ECO "B80"]
[TimeControl "3000+3"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be3 e6 7.f3 b5 8.Qd2 Nbd7
9.g4 b4 10.Na4 h6 11.O-O-O Ne5 12.Qxb4 Bd7 13.Bf4 g5 14.Bd2 Be7 15.h4 Rb8
16.Qa3 gxh4 17.Bb4 Qc7 18.Bxa6 O-O 19.Rxh4 Bxa4 20.Qxa4 Qb6 21.Bb5 Nxe4
22.Rh2 Nxf3 23.Nxf3 Ra8 24.Qb3 Qxb5 25.Re1 Rxa2 26.Rxe4 Ra1+ 27.Kd2 Qf1
28.Rhe2 Rb8 29.Qc3 Rd1+ 30.Ke3 Bd8 31.Qc6 Bb6+ 32.Kf4 Qh3 33.Bxd6 h5 34.
gxh5 Rf1 35.Rb4 Qh4+ 36.Ke5 Qxh5+ 37.Kf4 Qh4+ 38.Ke5 Qg3+ 39.Rf4 Qg7+ 40.
Rf6 Qg3+ 41.Ke4 Qg4+ 42.Rf4 Qg6+ 43.Ke5 Rd1 44.Rxf7 Rd5+ 45.Qxd5 Kxf7 46.
Kf4 Qf6+ 47.Kg4 exd5 48.Bxb8 Bd8 49.Be5 Qa6 50.Rh2 Qg6+ 51.Kf4 Qe4+ 52.Kg3
Ke8 53.Rh6 Bg5 54.Re6+ Be7 55.c3 Kd7 56.Rh6 Qc2 57.b4 Qd3 58.Rb6 Bd8 59.
Rb7+ Kc6 60.Rb8 Bb6 61.b5+ Qxb5 62.c4 Qb1 63.cxd5+ Kxd5 64.Bf4 Qg6+ 65.Ng5
Bd4 66.Rd8+ Kc4 67.Rd6 Qf5 68.Nf3 Bc5 69.Ne5+ Kb5 70.Rd8 Qh7 71.Rb8+ Ka6
72.Ra8+ Kb7 73.Rd8 Be7 74.Rd3 Ka6 75.Rd7 Qh4+ 76.Kf3 Kb5 77.Be3 Bb4 78.
Rd5+ Ka4 79.Rd3 Qf6+ 80.Ke4 Qe7 81.Bd4 Qh4+ 82.Kd5 Qd8+ 83.Ke4 Qg5 84.Bc3
Qh4+ 85.Kd5 Qh1+ 1/2-1/2
Parent - - By Harvey Williamson (*****) Date 2010-02-22 20:15

> [Event "CCT12"]
> [Site "FICS"]
> [Date "2010.02.20"]
> [Round "5"]
> [White "Hiarcs"]
> [Black "SjengX"]
> [Result "1/2-1/2"]
> [WhiteElo "2389"]
> [BlackElo "2570"]
> [ECO "B80"]
> [TimeControl "3000+3"]
>
> 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be3 e6 7.f3 b5 8.Qd2 Nbd7
> 9.g4 b4 10.Na4 h6 11.O-O-O Ne5 12.Qxb4 Bd7 13.Bf4 g5 14.Bd2 Be7 15.h4 Rb8
> 16.Qa3 gxh4 17.Bb4 Qc7 18.Bxa6 O-O 19.Rxh4 Bxa4 20.Qxa4 Qb6 21.Bb5 Nxe4
> 22.Rh2 Nxf3 23.Nxf3 Ra8 24.Qb3 Qxb5 25.Re1 Rxa2 26.Rxe4 Ra1+ 27.Kd2 Qf1
> 28.Rhe2 Rb8 29.Qc3 Rd1+ 30.Ke3 Bd8 31.Qc6 Bb6+ 32.Kf4 Qh3 33.Bxd6 h5 34.
> gxh5 Rf1 35.Rb4 Qh4+ 36.Ke5 Qxh5+ 37.Kf4 Qh4+ 38.Ke5 Qg3+ 39.Rf4 Qg7+ 40.
> Rf6 Qg3+ 41.Ke4 Qg4+ 42.Rf4 Qg6+ 43.Ke5 Rd1 44.Rxf7 Rd5+ 45.Qxd5 Kxf7 46.
> Kf4 Qf6+ 47.Kg4 exd5 48.Bxb8 Bd8 49.Be5 Qa6 50.Rh2 Qg6+ 51.Kf4 Qe4+ 52.Kg3
> Ke8 53.Rh6 Bg5 54.Re6+ Be7 55.c3 Kd7 56.Rh6 Qc2 57.b4 Qd3 58.Rb6 Bd8 59.
> Rb7+ Kc6 60.Rb8 Bb6 61.b5+ Qxb5 62.c4 Qb1 63.cxd5+ Kxd5 64.Bf4 Qg6+ 65.Ng5
> Bd4 66.Rd8+ Kc4 67.Rd6 Qf5 68.Nf3 Bc5 69.Ne5+ Kb5 70.Rd8 Qh7 71.Rb8+ Ka6
> 72.Ra8+ Kb7 73.Rd8 Be7 74.Rd3 Ka6 75.Rd7 Qh4+ 76.Kf3 Kb5 77.Be3 Bb4 78.
> Rd5+ Ka4 79.Rd3 Qf6+ 80.Ke4 Qe7 81.Bd4 Qh4+ 82.Kd5 Qd8+ 83.Ke4 Qg5 84.Bc3
> Qh4+ 85.Kd5 Qh1+ 1/2-1/2

Parent - By Uly (Gold) Date 2010-02-22 20:18
ty
Parent - - By NATIONAL12 (Gold) Date 2010-02-22 21:35
one of the free books gave the moves to move 51.Kf4.lol.My Book 18.3.
Parent - By Nelson Hernandez (Gold) Date 2010-02-23 00:10
I'm in all the way through move 57. 

This game was more an more obviously drawn the further you went in.

Opening position...31.0%
After 5 moves...36.1%
After 10 moves...48.7%
After 15 moves...68.8%
After 20 moves...78.3%
After 25 moves...82.7%
After 30 moves...88.8%
After 40 moves...90.4%
After 50 moves...92.5%

More human lines emanate from 9.O-O-O, which would have been a far superior move for Sjeng in this game.
Parent - By Nelson Hernandez (Gold) Date 2010-02-23 00:28
Novelty game from the position after move 50 is Optimus_Prime-Big Rabbit, 4 Sep 2009, a 16+0 game.  All games I have reaching that position are either Engine Room or opening book test games.

However the true novelty move was black's 57th.  Up to that point we're following Matrix-Suj1312 (!), 24 Jan 2010, a 60+15 game.  Congrats to Suj, that rascal.
Parent - By Octopus (**) Date 2010-02-22 21:10
Indeed, Chess960 or better 10x8 CRC are more interesting and guide
calculating programs to more success by understanding instead of
reproducing moves. Thus say good by to reproducing mega boxes.
Parent - - By h.g.muller (****) Date 2010-02-25 08:20
Indeed, 10x8 Chess would solve this problem in an elegant way. Lower draw rates, more room for imgination, fewer dull end-games... I don't see why people so desparately cling to the Mad Queen variant!
Parent - By Uly (Gold) Date 2010-02-25 17:22
Oh, hi h.g.muller, do you have planned to implement S-Chess into Winboard? I think there is some interest in playing this variant in a server.

S-Chess is just like normal chess, but when developing a piece from the back rank a player can choose to drop a Archbishop or a Chancellor on the new unoccupied square. Each player gets ONE Archbishop and ONE Chancellor, and if they have developed all back rank pieces without dropping their new pieces they can no longer do it. The pieces are called "Hawk" and "Elephant".

A new version of Komodo is being made that supports this variant, so I think it'd be interesting.
Up Topic The Rybka Lounge / Computer Chess / How opening books have destroyed computer chess

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill