Nunn tournament - Interim Report

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 23 January 2004 22:23:14:

Hi all ,
I am currently running a 13 engines tournament with all 20 Nunn positions from Suite 2.
Conditions:
Athlon 2600+
64MB Hash
4 and 5 men EGTB
4m+2s (Fischer clock)
GUI: I forgot for the moment (please understand that this is my first posting, although I am already testing for years and I want to post here as all are Winboard or UCI engine anyway and I do not want to post in a forum, where people are mainly interested in Shredder, Fritz and Ruffian)
Standings after 4 positions:

1 Movei 00_8_158 ---71.5/96
2 The Baron 1.2.0b8a ---64.5/96
3 Gothmog 0.4.6a ---59.5/96
4 Terra 3.3B3 ---56.0/96
5 Queen 2.42a ---55.0/96
6 Cerebro 1.23 ---50.0/96
7 Dorky 4.0b ---45.0/96
8 Snitch 0.3.5 ---41.5/96
9 Chispa 4.032 ---40.0/96
10 Black Bishop 0.9.2d ---39.0/96
11 Waster 0.11 ---36.5/96
12 Nullmover 0.23b ---36.0/96
13 Chiron 0.36 ---29.5/96
The main reason for me performing this tournament is that I am seeing so many contradictory results, often based upon few games. So Uri always pledged for conditions avoiding as much as possible accidentical results in order to have a better estimation if an improved version was created or not.
With this 480 games per engine tournament we will have a good basis to compare with future versions.
The idea is to build a huge database for rating calculation with Nunn positions exclusively, at 4+2 only and a decent variety of opponents, but not all. I would say between ELO 2550 and 2200. Moreover with same lines it will be possible to see where a new version prefers an alternative move, better or not.
The tornament played afterwards will be a multiple knockout Nunn position tournament with time control of at least 40 moves in 20 minutes + 40 in 20 + 20 minutes for the rest, where each author can choose one position. 16 engines are needed to perform that one, so we need three more who are interested or even more, if some of the prequalified above prefer to withdraw.
Should there be a lot of interest, what I doubt, several knock out tournaments will be held with 32 participants. Authors interested in participation can reply here or send a message to my email-address. Most I can´t ask personally for not having their address. First to come will be in. Rating should fit in the margin 2200-2550.
Best Regards
Heinz

[Peter Fendrich]

Geschrieben von: / Posted by: Peter Fendrich at 24 January 2004 00:10:23:
Als Antwort auf: / In reply to: Nunn tournament - Interim Report geschrieben von: / posted by: Heinz van Kempen at 23 January 2004 22:23:14:

Thank you for running this tournament!
There are two things I like here:
1) I don't know if my new version is improved for blitz games. The results here will give an indication.
2) It's fun when someone whith excellent chess knowledge is running a tourney. I very much apprieciate your detailed analysises, about each Nunn line played, that you send to us (the programmers).
I haven't used the Nunn lines very much as many others seems to do but they sure give a good hint about the engines strength. The book learning will of course not be tested.
Regards
/Peter
PS. Glad to see you posting here. I'm sure people will not kill you just because you're using the Fritz GUI


Terra

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 24 January 2004 00:59:53:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Peter Fendrich at 24 January 2004 00:10:23:

Thank you for running this tournament!
There are two things I like here:
1) I don't know if my new version is improved for blitz games. The results here will give an indication.
2) It's fun when someone whith excellent chess knowledge is running a tourney. I very much apprieciate your detailed analysises, about each Nunn line played, that you send to us (the programmers).
I haven't used the Nunn lines very much as many others seems to do but they sure give a good hint about the engines strength. The book learning will of course not be tested.
Regards
/Peter
PS. Glad to see you posting here. I'm sure people will not kill you just because you're using the Fritz GUI

It is a clear advantage for movei that book learning and books will not be tested.
I guess that the main engine that suffers from not having book is gothmog because I read clearly better result for it when books are used.
Inspite of the 158 number of movei that is clearly bigger than 131 there is no change in the evaluation from 131 to 158 and the only changes are small changes in the search.
Uri

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 02:58:56:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Peter Fendrich at 24 January 2004 00:10:23:

Thank you for running this tournament!
There are two things I like here:
1) I don't know if my new version is improved for blitz games. The results here will give an indication.
2) It's fun when someone whith excellent chess knowledge is running a tourney. I very much apprieciate your detailed analysises, about each Nunn line played, that you send to us (the programmers).
I haven't used the Nunn lines very much as many others seems to do but they sure give a good hint about the engines strength. The book learning will of course not be tested.
Regards
/Peter
PS. Glad to see you posting here. I'm sure people will not kill you just because you're using the Fritz GUI

Hello Peter,
thanks for your support, I was really afraid for being killed . Maybe I should repeat the same offer using Winboard or Arena instead.
Another thing: I have a huge database with more than 68000 Blitz games for more than 200 engines and different versions of them down to the weakest ones, for most a few hundred. I have also a decent database with a lot of games played with a time control of 25 minutes each engine for engines above 2400 ELO. And astonishingly the results under Fritz do not differ a lot from those others have with Arena or Winboard.It makes no sense to keep all this secretly and for my own, as I am doing for years, only for being afraid to get some nasty comments for using the "wrong" GUI.
Anyone interested in those games and EloStat calculation or complete rating list can also write to me and he will be given those games, indepent of wanting to participate in my tournaments or not.
Best Regards
Heinz

[Olivier Deville]

Geschrieben von: / Posted by: Olivier Deville at 24 January 2004 04:07:01:
Als Antwort auf: / In reply to: For those interested in games of their engine geschrieben von: / posted by: Heinz van Kempen at 24 January 2004 02:58:56:

Thank you for running this tournament!
There are two things I like here:
1) I don't know if my new version is improved for blitz games. The results here will give an indication.
2) It's fun when someone whith excellent chess knowledge is running a tourney. I very much apprieciate your detailed analysises, about each Nunn line played, that you send to us (the programmers).
I haven't used the Nunn lines very much as many others seems to do but they sure give a good hint about the engines strength. The book learning will of course not be tested.
Regards
/Peter
PS. Glad to see you posting here. I'm sure people will not kill you just because you're using the Fritz GUI

Hello Peter,
thanks for your support, I was really afraid for being killed . Maybe I should repeat the same offer using Winboard or Arena instead.
Another thing: I have a huge database with more than 68000 Blitz games for more than 200 engines and different versions of them down to the weakest ones, for most a few hundred. I have also a decent database with a lot of games played with a time control of 25 minutes each engine for engines above 2400 ELO. And astonishingly the results under Fritz do not differ a lot from those others have with Arena or Winboard.It makes no sense to keep all this secretly and for my own, as I am doing for years, only for being afraid to get some nasty comments for using the "wrong" GUI.
Anyone interested in those games and EloStat calculation or complete rating list can also write to me and he will be given those games, indepent of wanting to participate in my tournaments or not.
Best Regards
Heinz

Hi Heinz ! I'm using the Fritz GUI for my tournaments and authors still find useful to send me versions of their engines . Best engines are still on top, and weak ones at the bottom.
Olivier



ChessWar

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 04:17:25:
Als Antwort auf: / In reply to: Re: For those interested in games of their engine geschrieben von: / posted by: Olivier Deville at 24 January 2004 04:07:01:

Hi Heinz ! I'm using the Fritz GUI for my tournaments and authors still find useful to send me versions of their engines . Best engines are still on top, and weak ones at the bottom.
Olivier

Hello Olivier,
good to know that I am not a complete outsider. Note that I am following your interesting Chess War tournaments and others like WBEC, RWBC, WCCL, Le Fou etc. with suspense.
It is not that I do not like Winboard and fast improving Arena. I am also using that ones in the last months. Depends a lot on the engines I want to have in.
Have fun with your tournaments
Heinz

[Roger Brown]

Geschrieben von: / Posted by: Roger Brown at 24 January 2004 06:50:58:
Als Antwort auf: / In reply to: Nunn tournament - Interim Report geschrieben von: / posted by: Heinz van Kempen at 23 January 2004 22:23:14:

Hi all ,
Best Regards
Heinz

Welcome Heinz!
Later.

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 24 January 2004 09:26:53:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Uri Blass at 24 January 2004 00:59:53:

PS. Glad to see you posting here. I'm sure people will not kill you just because you're using the Fritz GUI

I guess that the main engine that suffers from not having book is gothmog because I read clearly better result for it when books are used.

Yes, I've found this group to be very friendly. You won't easily get flamed or killed here.
I don't think playing without book is a bigger disadvantage for Gothmog than for the other
engines. You are right that there has been a few tournaments lately where Gothmog has
performed much better, but I think this is maily due to luck.
I am satisfied with Gothmog's results so far in the tournament, especially considering that the
first three Nunn positions were probably badly suited to its style of play. I think it should
generally score better in the sharper opening lines, but I could be wrong.
Tord

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 09:38:41:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Tord Romstad at 24 January 2004 09:26:53:

I am satisfied with Gothmog's results so far in the tournament, especially considering that the
first three Nunn positions were probably badly suited to its style of play. I think it should
generally score better in the sharper opening lines, but I could be wrong.
Tord

Hello Tord,
your presumption seems right so far. Gothmog is just winning Nunn2 position 5 (Pirc-Ufimzev) with a big edge and some gorgeous games.
More this evening when I will keep you informed with crosstables and games.
Have fun
Heinz

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 16:08:57:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Uri Blass at 24 January 2004 00:59:53:

To the participants in Nunn tournament:
Some notes on the tablebases. There are 5 men on one machine and 4 men on the other one. From tomorrow onwards there will be present 5 men EGTB on both.
There are positions K+R+N vs. K+R. In GM games here is usually agreed a draw, if there is no severe time-trouble or the King of the weaker party restricted in a corner. With the King of the weaker side in or near the centre there is practiaclly no win. So "tell" your engines if you havn´t already to avoid such a material balance when having a promising position and to try to get this final to rescue a draw. (I suppose that I do not have to tell you in what kind of positions it is better to exchange pieces when going for a win, or better to exchange pawns, when inferior, as this is well-known chessknowledge you gave to your mule years ago ).
Another endgame is K+R+B against K+R. One guy statistically examined by checking databases that in GM games only 20% of all positions with this were won for white. But in games confronting weaker players there are more wins. This now was meaningful in a game The Baron vs. Movei, where Movei had the additional bishop. But as this was on the 4 men machine where the Pirc was held, except for games of Black Bishop, that is more stable on the other machine. So Movei might have won with 5 men present (I did not check this so far). But I also understood that most engines already are checking the tablebases for positions that might arise with 5 men, when there are still more pieces present on the board. I would like to have your opinions here. If I do not receive any comments, this will be an indication for me, that we will take the Blitz contest not that seriously. Anyway do me one favour. Do not check all endgames arisen so far, for what would have happened with 5 men, or viceversa, what would have happened if there were only 4 men instead of 5.

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 16:30:19:
Als Antwort auf: / In reply to: table bases and critical game geschrieben von: / posted by: Heinz van Kempen at 24 January 2004 16:08:57:

Correction:
One guy statistically examined by checking databases that in GM games only 20% of all positions with this were won for white.
Correction: "were won for the party with R+B of course."
Bye
Heinz

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 24 January 2004 16:53:52:
Als Antwort auf: / In reply to: table bases and critical game geschrieben von: / posted by: Heinz van Kempen at 24 January 2004 16:08:57:

To the participants in Nunn tournament:
Some notes on the tablebases. There are 5 men on one machine and 4 men on the other one. From tomorrow onwards there will be present 5 men EGTB on both.
There are positions K+R+N vs. K+R. In GM games here is usually agreed a draw, if there is no severe time-trouble or the King of the weaker party restricted in a corner. With the King of the weaker side in or near the centre there is practiaclly no win. So "tell" your engines if you havn´t already to avoid such a material balance when having a promising position and to try to get this final to rescue a draw. (I suppose that I do not have to tell you in what kind of positions it is better to exchange pieces when going for a win, or better to exchange pawns, when inferior, as this is well-known chessknowledge you gave to your mule years ago ).
Another endgame is K+R+B against K+R. One guy statistically examined by checking databases that in GM games only 20% of all positions with this were won for white. But in games confronting weaker players there are more wins. This now was meaningful in a game The Baron vs. Movei, where Movei had the additional bishop. But as this was on the 4 men machine where the Pirc was held, except for games of Black Bishop, that is more stable on the other machine. So Movei might have won with 5 men present (I did not check this so far). But I also understood that most engines already are checking the tablebases for positions that might arise with 5 men, when there are still more pieces present on the board.

The problem for me is not chess knowledge but implementing things without bugs so I did not implement most of the things that I know about chess.
I also consider tactics as more important than evaluation at the stage of movei and I believe that better pruning and extensions can give more to movei.
Movei lost significant part of the points because of tactical mistakes(for example it lost a pawn in the opening in 2 of it's losses in nunn position number 1).

Not movei so the draw is clearly a fair result for the baron.
Not using tablebases is a weakness of movei and I plan to continue to have this weakness in the near future.
Uri

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 24 January 2004 17:23:51:
Als Antwort auf: / In reply to: Re: table bases and critical game geschrieben von: / posted by: Uri Blass at 24 January 2004 16:53:52:

To the participants in Nunn tournament:
Some notes on the tablebases. There are 5 men on one machine and 4 men on the other one. From tomorrow onwards there will be present 5 men EGTB on both.
There are positions K+R+N vs. K+R. In GM games here is usually agreed a draw, if there is no severe time-trouble or the King of the weaker party restricted in a corner. With the King of the weaker side in or near the centre there is practiaclly no win. So "tell" your engines if you havn´t already to avoid such a material balance when having a promising position and to try to get this final to rescue a draw. (I suppose that I do not have to tell you in what kind of positions it is better to exchange pieces when going for a win, or better to exchange pawns, when inferior, as this is well-known chessknowledge you gave to your mule years ago ).
Another endgame is K+R+B against K+R. One guy statistically examined by checking databases that in GM games only 20% of all positions with this were won for white. But in games confronting weaker players there are more wins. This now was meaningful in a game The Baron vs. Movei, where Movei had the additional bishop. But as this was on the 4 men machine where the Pirc was held, except for games of Black Bishop, that is more stable on the other machine. So Movei might have won with 5 men present (I did not check this so far). But I also understood that most engines already are checking the tablebases for positions that might arise with 5 men, when there are still more pieces present on the board.

The problem for me is not chess knowledge but implementing things without bugs so I did not implement most of the things that I know about chess.
I also consider tactics as more important than evaluation at the stage of movei and I believe that better pruning and extensions can give more to movei.
Movei lost significant part of the points because of tactical mistakes(for example it lost a pawn in the opening in 2 of it's losses in nunn position number 1).
Not movei so the draw is clearly a fair result for the baron.
Not using tablebases is a weakness of movei and I plan to continue to have this weakness in the near future.
Uri

In a second thought I think that it is possible that with 5 piece tablebases baron could lose because of prefering not to go to the losing KRB vs KR.
I find it as an unfair advantage for movei relative to programs that know to use tablebases if Fritz is playing tablebases positions.
Uri
Uri

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 24 January 2004 20:22:45:
Als Antwort auf: / In reply to: Re: table bases and critical game geschrieben von: / posted by: Uri Blass at 24 January 2004 17:23:51:

To the participants in Nunn tournament:
Some notes on the tablebases. There are 5 men on one machine and 4 men on the other one. From tomorrow onwards there will be present 5 men EGTB on both.
There are positions K+R+N vs. K+R. In GM games here is usually agreed a draw, if there is no severe time-trouble or the King of the weaker party restricted in a corner. With the King of the weaker side in or near the centre there is practiaclly no win. So "tell" your engines if you havn´t already to avoid such a material balance when having a promising position and to try to get this final to rescue a draw. (I suppose that I do not have to tell you in what kind of positions it is better to exchange pieces when going for a win, or better to exchange pawns, when inferior, as this is well-known chessknowledge you gave to your mule years ago ).
Another endgame is K+R+B against K+R. One guy statistically examined by checking databases that in GM games only 20% of all positions with this were won for white. But in games confronting weaker players there are more wins. This now was meaningful in a game The Baron vs. Movei, where Movei had the additional bishop. But as this was on the 4 men machine where the Pirc was held, except for games of Black Bishop, that is more stable on the other machine. So Movei might have won with 5 men present (I did not check this so far). But I also understood that most engines already are checking the tablebases for positions that might arise with 5 men, when there are still more pieces present on the board.

The problem for me is not chess knowledge but implementing things without bugs so I did not implement most of the things that I know about chess.
I also consider tactics as more important than evaluation at the stage of movei and I believe that better pruning and extensions can give more to movei.
Movei lost significant part of the points because of tactical mistakes(for example it lost a pawn in the opening in 2 of it's losses in nunn position number 1).
Not movei so the draw is clearly a fair result for the baron.
Not using tablebases is a weakness of movei and I plan to continue to have this weakness in the near future.
Uri

In a second thought I think that it is possible that with 5 piece tablebases baron could lose because of prefering not to go to the losing KRB vs KR.
I find it as an unfair advantage for movei relative to programs that know to use tablebases if Fritz is playing tablebases positions.
Uri

In third thought after looking at the game the position was clear draw and movei could not win.
Uri

[Heinz van Kempen]

Geschrieben von: / Posted by: Heinz van Kempen at 24 January 2004 23:54:29:
Als Antwort auf: / In reply to: Re: Nunn tournament - Interim Report geschrieben von: / posted by: Roger Brown at 24 January 2004 06:50:58:

Hi all ,
Best Regards
Heinz

Welcome Heinz!
Later.

Hello Roger,
thanks for your welcome. I already liked your fine sense of humour, although staying mainly passively here.
All the best to you
Heinz