Winboard Forum

[Event "ChessWar XIV D 40m/20'"]
[Site "DEVILLE1-5D9C28"]
[Date "2009.06.05"]
[Round "1.30"]
[White "Gaviota 0.33"]
[Black "Gaia 3.5r"]
[Result "0-1"]
[TimeControl "40/1200"]
[Annotator "10. -10.16 9... +9.83"]
[Number "30"]

1. d4 Nf6 2. c4 c6 3. Nc3 d5 4. Nf3 dxc4 5. a4 Bf5 6. e3 e6 7. Bxc4 Bb4 8.
O-O Nbd7 9. Qc2 Bxc2 {+9.83/16 34} 10. Na2 {-10.16/12 1:53} Bd6
{+10.02/16 34} 11. Ne1 {-10.06/12 1:10} Bxa4 {+10.24/16 34} 12. Nc3
{-9.88/12 51} b5 {+10.28/16 34} 13. Nxa4 {-9.80/12 33} bxc4 {+10.45/16 34}
14. Nf3 {-9.96/12 33} Qb8 {+10.43/16 34} 15. Nc3 {-9.94/11 33} c5
{+10.63/16 34} 16. h3 {-10.01/11 33} O-O {+10.82/16 35} 17. Ra4
{-9.78/10 33} Nb6 {+10.88/16 35} 18. Ra2 {-10.01/12 33} cxd4 {+11.08/18 35}
{Gaviota resigns as White} 0-1

Olivier

Oooops,an ugly and nasty bug indeed
Dr.D

Very good/successful continuation for getting the opponent out of book.

Olivier Deville wrote:[Event "ChessWar XIV D 40m/20'"]
[Site "DEVILLE1-5D9C28"]
[Date "2009.06.05"]
[Round "1.30"]
[White "Gaviota 0.33"]
[Black "Gaia 3.5r"]
[Result "0-1"]
[TimeControl "40/1200"]
[Annotator "10. -10.16 9... +9.83"]
[Number "30"]

1. d4 Nf6 2. c4 c6 3. Nc3 d5 4. Nf3 dxc4 5. a4 Bf5 6. e3 e6 7. Bxc4 Bb4 8.
O-O Nbd7 9. Qc2 Bxc2 {+9.83/16 34} 10. Na2 {-10.16/12 1:53} Bd6
{+10.02/16 34} 11. Ne1 {-10.06/12 1:10} Bxa4 {+10.24/16 34} 12. Nc3
{-9.88/12 51} b5 {+10.28/16 34} 13. Nxa4 {-9.80/12 33} bxc4 {+10.45/16 34}
14. Nf3 {-9.96/12 33} Qb8 {+10.43/16 34} 15. Nc3 {-9.94/11 33} c5
{+10.63/16 34} 16. h3 {-10.01/11 33} O-O {+10.82/16 35} 17. Ra4
{-9.78/10 33} Nb6 {+10.88/16 35} 18. Ra2 {-10.01/12 33} cxd4 {+11.08/18 35}
{Gaviota resigns as White} 0-1

Olivier

Ouch!
I have never seen anything like it before... If there was a bug in the evaluation or search, I am sure it is gone from the current (unreleased) version because of the gazillions modifications I made. However, I wonder whether this is related to the opening book. That part of the code has not been touched. Do you have by any chance the log file?

Considering that is move 9 from a very classical opening, it is highly likely that Gaviota was still in the book.

Miguel

Miguel A. Ballicora wrote:

Olivier Deville wrote:[Event "ChessWar XIV D 40m/20'"]
[Site "DEVILLE1-5D9C28"]
[Date "2009.06.05"]
[Round "1.30"]
[White "Gaviota 0.33"]
[Black "Gaia 3.5r"]
[Result "0-1"]
[TimeControl "40/1200"]
[Annotator "10. -10.16 9... +9.83"]
[Number "30"]

1. d4 Nf6 2. c4 c6 3. Nc3 d5 4. Nf3 dxc4 5. a4 Bf5 6. e3 e6 7. Bxc4 Bb4 8.
O-O Nbd7 9. Qc2 Bxc2 {+9.83/16 34} 10. Na2 {-10.16/12 1:53} Bd6
{+10.02/16 34} 11. Ne1 {-10.06/12 1:10} Bxa4 {+10.24/16 34} 12. Nc3
{-9.88/12 51} b5 {+10.28/16 34} 13. Nxa4 {-9.80/12 33} bxc4 {+10.45/16 34}
14. Nf3 {-9.96/12 33} Qb8 {+10.43/16 34} 15. Nc3 {-9.94/11 33} c5
{+10.63/16 34} 16. h3 {-10.01/11 33} O-O {+10.82/16 35} 17. Ra4
{-9.78/10 33} Nb6 {+10.88/16 35} 18. Ra2 {-10.01/12 33} cxd4 {+11.08/18 35}
{Gaviota resigns as White} 0-1

Olivier

Ouch!
I have never seen anything like it before... If there was a bug in the evaluation or search, I am sure it is gone from the current (unreleased) version because of the gazillions modifications I made. However, I wonder whether this is related to the opening book. That part of the code has not been touched. Do you have by any chance the log file?

Considering that is move 9 from a very classical opening, it is highly likely that Gaviota was still in the book.

Miguel

Geeezzz... I found the problem.

I checked the book available for download in Gaviota's site and Qc2 is a BOOK MOVE!!! ARRGGGGGHHHHH.

after typing bk (a winboard command in which the engine returns information about the book) I get:

Code: Select all

+-----------------+
| r . . q k . . r |
| p p . n . p p p |
| . . p . p n . . |
| . . . . . b . . |    Castling: kq
| P b B P . . . . |    ep: -
| . . N . P N . . |
| . P . . . P P P |
| R . B Q . R K . | [White]
+-----------------+

bk
        move    L!?     W     D     L     %
        Qe2     ---     5     6     1   46.3
        Nh4     ---     2     4     1   21.1
        Qb3     ---     1     0     0   16.3
        Qc2     ---     1     0     0   16.3
        Ne2     ---     0     1     0    0.0


So, it is not a bug in the engine. It is book with a wrong move. I created it with GM games, and all the GMs where ELO 2650 or higher. I guess that I took the pgn file from Dann Corbit's site or maybe Crafty's. I do not remember. I know it was free .There is possibly a game in a pgn file that is wrong somewhere. People may want to check out their collections... It will be easy to search for this position (after Qc2):

r2qk2r/pp1n1ppp/2p1pn2/5b2/PbBP4/2N1PN2/1PQ2PPP/R1B2RK1 b kq - 4 18

Most likely, whoever entered the game, mistyped Qc2 instead of Qe2 (the next black book move after Qc2 is Bg6!, typical answer to Qe2).

If book learning was on, it is possible that Gaviota may not play this move anymore. Certainly less likely than 16.3%

Miguel

So, it is not a bug in the engine. It is book with a wrong move. I created it with GM games, and all the GMs where ELO 2650 or higher. I guess that I took the pgn file from Dann Corbit's site or maybe Crafty's. I do not remember. I know it was free .There is possibly a game in a pgn file that is wrong somewhere. People may want to check out their collections... It will be easy to search for this position (after Qc2):

r2qk2r/pp1n1ppp/2p1pn2/5b2/PbBP4/2N1PN2/1PQ2PPP/R1B2RK1 b kq - 4 18

Most likely, whoever entered the game, mistyped Qc2 instead of Qe2 (the next black book move after Qc2 is Bg6!, typical answer to Qe2).

If book learning was on, it is possible that Gaviota may not play this move anymore. Certainly less likely than 16.3%

Miguel

This is interesting, I looked it up in my own opening book database and found it there! However, this move would not have actually made it into Kanguruhs opening book because there is only a sample of one game for Qc2 and therefore would be filtered out.




I double clicked on the to move to see who played it ...




I get most of my openings from a file I downloaded from somewhere called 'elo2600.pgn', and that's where I found the game.

Code: Select all

[Event "Novosibirsk"]
[Site "Novosibirsk RUS"]
[Date "1993.??.??"]
[Round "?"]
[White "Beliavsky, Alexander G"]
[Black "Akopian, Vladimir E"]
[Result "1-0"]
[WhiteElo "2655"]
[BlackElo "2655"]
[ECO "D18"]
[EventDate "1993.??.??"]

1.d4 d5 2.c4 c6 3.Nf3 Nf6 4.Nc3 dxc4 5.a4 Bf5 6.e3 e6 7.Bxc4 Bb4 8.O-O
Nbd7 9.Qc2 Bg6 10.e4 Bxc3 11.bxc3 Nxe4 12.Ba3 Qc7 13.Rfe1 O-O-O 14.Qb2
Rhe8 15.a5 e5 16.Rab1 c5 17.Bf1 f6 18.Ra1 Nd6 19.Nd2 Kb8 20.Rec1 Ka8 21.
Nb3 c4 22.Bxd6 Qxd6 23.Bxc4 a6 24.Bf1 Rc8 25.Nd2 Rc6 26.g3 Bf7 27.Rcb1 Rb8
28.Ra4 exd4 29.Rxd4 Qc7 30.Qb4 Ne5 31.f4 Ng4 32.Re1 Rc5 33.Bg2 Nh6 34.Qb6
Qxb6 35.axb6 Rb5 36.Rd6 Ng8 37.Ra1 Rf8 38.c4 Rb2 39.c5 Ne7 40.Rxa6+ Kb8
41.Ra7 Nc6 42.Rd7 1-0


I agree that this looks like a simple data entry error. I think anybody who used elo2600.pgn to create their opening book could potentially have this move in their engines opening book.

Cheers,
Tom.

Thomas McBurney wrote:

So, it is not a bug in the engine. It is book with a wrong move. I created it with GM games, and all the GMs where ELO 2650 or higher. I guess that I took the pgn file from Dann Corbit's site or maybe Crafty's. I do not remember. I know it was free .There is possibly a game in a pgn file that is wrong somewhere. People may want to check out their collections... It will be easy to search for this position (after Qc2):

r2qk2r/pp1n1ppp/2p1pn2/5b2/PbBP4/2N1PN2/1PQ2PPP/R1B2RK1 b kq - 4 18

Most likely, whoever entered the game, mistyped Qc2 instead of Qe2 (the next black book move after Qc2 is Bg6!, typical answer to Qe2).

If book learning was on, it is possible that Gaviota may not play this move anymore. Certainly less likely than 16.3%

Miguel

This is interesting, I looked it up in my own opening book database and found it there! However, this move would not have actually made it into Kanguruhs opening book because there is only a sample of one game for Qc2 and therefore would be filtered out.

I think this is a good solution (not to include moves from single games) to avoid this type of problems. At least, it should be an option. I will add that to Gaviota's book maker.

It is funny that this wrong move does not break the pgn because the next Queen move is Qb2. Both, Qe2 and Qc2 are compatible with that.

Miguel

I double clicked on the to move to see who played it ...




I get most of my openings from a file I downloaded from somewhere called 'elo2600.pgn', and that's where I found the game.

Code: Select all

[Event "Novosibirsk"]
[Site "Novosibirsk RUS"]
[Date "1993.??.??"]
[Round "?"]
[White "Beliavsky, Alexander G"]
[Black "Akopian, Vladimir E"]
[Result "1-0"]
[WhiteElo "2655"]
[BlackElo "2655"]
[ECO "D18"]
[EventDate "1993.??.??"]

1.d4 d5 2.c4 c6 3.Nf3 Nf6 4.Nc3 dxc4 5.a4 Bf5 6.e3 e6 7.Bxc4 Bb4 8.O-O
Nbd7 9.Qc2 Bg6 10.e4 Bxc3 11.bxc3 Nxe4 12.Ba3 Qc7 13.Rfe1 O-O-O 14.Qb2
Rhe8 15.a5 e5 16.Rab1 c5 17.Bf1 f6 18.Ra1 Nd6 19.Nd2 Kb8 20.Rec1 Ka8 21.
Nb3 c4 22.Bxd6 Qxd6 23.Bxc4 a6 24.Bf1 Rc8 25.Nd2 Rc6 26.g3 Bf7 27.Rcb1 Rb8
28.Ra4 exd4 29.Rxd4 Qc7 30.Qb4 Ne5 31.f4 Ng4 32.Re1 Rc5 33.Bg2 Nh6 34.Qb6
Qxb6 35.axb6 Rb5 36.Rd6 Ng8 37.Ra1 Rf8 38.c4 Rb2 39.c5 Ne7 40.Rxa6+ Kb8
41.Ra7 Nc6 42.Rd7 1-0


I agree that this looks like a simple data entry error. I think anybody who used elo2600.pgn to create their opening book could potentially have this move in their engines opening book.

Cheers,
Tom.