WBEC Ridderkerk new results.

[Leo Dijksman]

Geschrieben von: / Posted by: Leo Dijksman at 06 March 2004 16:54:29:

1st Division:
Round 8:
Abrok 5.0 0½½1 2.0/4
King Of Kings 2.52 1½½0 2.0/4
------------------------------------
El Chinito 3.25 0½11 2.5/4
SOS 99.11.03 1½00 1.5/4
------------------------------------
PostModernist 1008 1110 3.0/4
XiniX 04.02.09 0001 1.0/4
------------------------------------
Round 9:
Patzer 3.61 11 2.0/2 = 2 games to go!
FrancescaMAD 0.0.9 00 0.0/2
------------------------------------

3rd Division:
Round 10:
NullMover 0.24d 111½ 3.5/4
TRACE 1.25 000½ 0.5/4
------------------------------------
Dorky 3.48 ½11½ 3.0/4
BlackBishop 0.9.7g ½00½ 1.0/4
------------------------------------
Round 11:
Maestro Exp. ½½ 1.0/2 = 2 games to go!
Gaviota 0.33 ½½ 1.0/2
------------------------------------

5th Division:
Round 16:
Chiron 0.38 - DrunkenMaster 0.9 1-0 61
DrunkenMaster 0.9 - Chiron 0.38 0-1 40
BigLion 2.23i - RDChess 3.23 1-0 43
RDChess 3.23 - BigLion 2.23i 0-1 67 The Lion at 1! (but Waster have still 2 games to go!)
LaDameBlanche 2.0c - Ranita 2.4 =-= 126
Ranita 2.4 - LaDameBlanche 2.0c 0-1 53
Simontacchi 1.8a - Grizzly 1.40.1b 1-0 61
Grizzly 1.40.1b - Simontacchi 1.8a 1-0 31
Cilian 4.13 - Rinko 1.15 1-0 53
Rinko 1.15 - Cilian 4.13 0-1 89
EnginMax 5.11c - Bodo 0.2e =-= 107
Bodo 0.2e - EnginMax 5.11c 0-1 67
-------------------------------------------------------------

6th Division:
Round 14:
BSC 2.8 - PolarChess 1.3 =-= 183
Belzebub 0.64 - Adam 1.6 1-0 33
Embracer 1.12 - Enigma 1.1.3 1-0 35
Gargamella 0.7.0 - Zotron 2003 R7.5 1-0 63
Storm 0.6 - WJChess 1.52 =-= 82
BremboCE 0.31 - ChessRikus 1.4.60 1-0 41
Golem 0.4 - Hoplite 1.4.1 0-1 53
Awesome 1.45 - Booot 2.4 0-1 55
ApiChess 1.29 - Replicant 1.2e 0-1 61
Robin 0.9.86 - Smash 0.8d =-= 58
SdBC 0.4.13.0 - Parrot 031231 0-1 78
WESP 0.8 - Matacz 0.93 1-0 92
-------------------------------------------------------------

Crosstables and pgn on the WBEC homepage.
Enginelist, latest updates/new engines:
06/03/2004:
Updated: Amy v0.8.7b (bugfixed Windows exe by Dann Corbit)
Updated: Tytan v3.6
Leo.


WBEC Ridderkerk homepage.

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 06 March 2004 17:40:09:
Als Antwort auf: / In reply to: WBEC Ridderkerk new results. geschrieben von: / posted by: Leo Dijksman at 06 March 2004 16:54:29:

BigLion 2.23i - RDChess 3.23 1-0 43
RDChess 3.23 - BigLion 2.23i 0-1 67 The Lion at 1! (but Waster have still 2 games to go!)

Hi Leo,
I have been silently watching the performance of BigLion and asking myself
when he intends to wake up. I have been however happy about his stability.
Thanks for the daily tension.
/Matthias.


BigLion + Taktix

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 06 March 2004 20:11:52:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Matthias Gemuh at 06 March 2004 19:49:27:

BigLion 2.23i - RDChess 3.23 1-0 43
RDChess 3.23 - BigLion 2.23i 0-1 67 The Lion at 1! (but Waster have still 2 games to go!)

Hi Matthias,
congratulations to your wins and your strong engine!
RDChess is too weak and still plays horrible blunder moves.
I finally decided to stop work on Delphi RDChess V3.xx and have already begun work on a new engine, built up from the scratch. The GUI is written in Visual Studio C#, the search engine will be a C++ DLL.
I am convinced my second chess program will be stronger than my first one, how much I do not know.
I think I will need at least half a year for programming until releasing a first beta version and one more half year until relasing it under the name RDChess V4.0.
Rudolf

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.
The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

[Marcus Prewarski]

Geschrieben von: / Posted by: Marcus Prewarski at 06 March 2004 20:24:50:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Uri Blass at 06 March 2004 20:11:52:

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.
The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

It really depends on how your move ordering works. In previous versions of my engine at longer time controls I've seen it score some history moves above winning captures, this can really screw up your move ordering and branch factor. It wouldn't cause a blunder so maybe it isn't so bad but it can be a problem. I might still have some problems and one of the things I plan to do is re-write my move ordering and move selections routines to completely prevent anything like this. My engine has some other problems as well like a hash-bug maybe, and my plan is to debug what I have before changing much.
I haven't seen too many games of RDChess but given how much stronger it is at blitz compared to longer games it must have some bugs. I was a little suprised how well it did in Leo's test tournament then how poorly it has done in the longer one.
-Marcus

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 06 March 2004 20:33:36:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Uri Blass at 06 March 2004 20:11:52:

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.
The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

In the case I explained, BigLion's move ordering used to drop to below 50% as captures were considered bad everywhere in the tree and my PVS would sometimes
need several minutes to see good moves in the virtually unsorted tree.
So within a reasonably long time control, BigLion would blunder.
At Blitz, the problem disappeared as history heuristics values remained small.
/Matthias.


BigLion + Taktix

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 06 March 2004 21:24:01:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Matthias Gemuh at 06 March 2004 20:33:36:

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.
The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

In the case I explained, BigLion's move ordering used to drop to below 50% as captures were considered bad everywhere in the tree and my PVS would sometimes
need several minutes to see good moves in the virtually unsorted tree.

My point is that the problem of RDchess is stupid blunders and not bad order of moves at long time control.
Even if your order starts to be bad at long time control you still play better at longer time control but only slightly better unless your bad move ordering also influence your move ordering in the next moves(I set my history to 0 after every move and I also do not have a single score for every move for order of moves and only when phase[ply]=historyphase I use the score of the history tables).
Defining phase[ply] is an idea that I learned from Crafty
Uri

[Roger Brown]

Geschrieben von: / Posted by: Roger Brown at 06 March 2004 22:49:06:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Uri Blass at 06 March 2004 21:24:01:

The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

My point is that the problem of RDchess is stupid blunders and not bad order of moves at long time control.

Hello Uri,
I am going to assume that you meant no harm but your particular method of expression could be mistaken for contempt. I do not believe that you intend to do so BUT you seem to convey the idea in your two posts.
The mere repitition of the same concept in the later post does not in my opinion, yield any more useful data. I suggest that you take a moment and consider the effect that your posts may have on the casual reader.
Of course there is always the possibility that no-one minds but I am just expressing my opinion as a member of this Forum.
That the program may or not make elementary errors is surely subject to rational analysis. The repeated use of the word stupid - particularly when blunder adequately conveys the idea - adds an unwelcome dimension to the post.
Feel free to disagree of course.
Later.

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 06 March 2004 23:13:20:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Roger Brown at 06 March 2004 22:49:06:

The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

My point is that the problem of RDchess is stupid blunders and not bad order of moves at long time control.

Hello Uri,
I am going to assume that you meant no harm but your particular method of expression could be mistaken for contempt. I do not believe that you intend to do so BUT you seem to convey the idea in your two posts.
The mere repitition of the same concept in the later post does not in my opinion, yield any more useful data. I suggest that you take a moment and consider the effect that your posts may have on the casual reader.
Of course there is always the possibility that no-one minds but I am just expressing my opinion as a member of this Forum.
That the program may or not make elementary errors is surely subject to rational analysis. The repeated use of the word stupid - particularly when blunder adequately conveys the idea - adds an unwelcome dimension to the post.
Feel free to disagree of course.
Later.

It probably would have been a lot more gentle if he had said:
"The problem is in the evaluation, and not in the search or move ordering."
Conveys the same information but does not taste as salty.
IMO-YMMV.



my ftp site {remove http:// unless you like error messages}

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 06 March 2004 23:35:22:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Dann Corbit at 06 March 2004 23:13:20:

The real big problem is when the engine make stupid blunders at long time control and based on my understanding it is exactly the case with Rdchess.
Uri

My point is that the problem of RDchess is stupid blunders and not bad order of moves at long time control.

Hello Uri,
I am going to assume that you meant no harm but your particular method of expression could be mistaken for contempt. I do not believe that you intend to do so BUT you seem to convey the idea in your two posts.
The mere repitition of the same concept in the later post does not in my opinion, yield any more useful data. I suggest that you take a moment and consider the effect that your posts may have on the casual reader.
Of course there is always the possibility that no-one minds but I am just expressing my opinion as a member of this Forum.
That the program may or not make elementary errors is surely subject to rational analysis. The repeated use of the word stupid - particularly when blunder adequately conveys the idea - adds an unwelcome dimension to the post.
Feel free to disagree of course.
Later.

It probably would have been a lot more gentle if he had said:
"The problem is in the evaluation, and not in the search or move ordering."
Conveys the same information but does not taste as salty.
IMO-YMMV.

I accept the suggestion not to use the word stupid but my guess is that the problem is not in the evaluation but in bugs that bound checkers can find.
Based on experience these bugs do not have to cause the program to crush and I suspect that they may cause 1 ply blunders only at long time control.
1 ply blunders at long time control was a problem of Terra and Horizon so this problem is not unique to RDchess.
Uri

[Albert Bertilsson]

Geschrieben von: / Posted by: Albert Bertilsson at 07 March 2004 08:42:35:
Als Antwort auf: / In reply to: WBEC Ridderkerk new results. geschrieben von: / posted by: Leo Dijksman at 06 March 2004 16:54:29:

Hi!
As usual it is getting exciting when the end of the division is getting closer. I'm following the former main opponents for Sharper, to see which one is also going to step up to the 4th division. I can't wait for Sharper to start playing again.
/Albert

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 07 March 2004 10:00:08:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Marcus Prewarski at 06 March 2004 20:24:50:

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.

It really depends on how your move ordering works. In previous versions of my engine at longer time controls I've seen it score some history moves above winning captures, this can really screw up your move ordering and branch factor. It wouldn't cause a blunder so maybe it isn't so bad but it can be a problem.

This is true, but the problem is not very difficult to fix. Whenever you increase an entry
in your history table, you check whether the new value of the entry exceeds some
threshold. If it does, loop through your history table and divide all entries by 2.
Tord

[Leo Dijksman]

Geschrieben von: / Posted by: Leo Dijksman at 07 March 2004 11:39:23:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk, exciting... geschrieben von: / posted by: Albert Bertilsson at 07 March 2004 08:42:35:

Hi!
As usual it is getting exciting when the end of the division is getting closer. I'm following the former main opponents for Sharper, to see which one is also going to step up to the 4th division. I can't wait for Sharper to start playing again.
/Albert

Hi Albert,
No need to wait very long, Sharper will start in arond one week
Leo.




WBEC Ridderkerk homepage.

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 07 March 2004 12:58:08:
Als Antwort auf: / In reply to: Waster shall be Champ geschrieben von: / posted by: Matthias Gemuh at 07 March 2004 00:34:15:

Waster needs to win both the games in hand to sneak ahead again,

No engines can stop Waster from being the Champ down here in Division 5.

I hope that your engine is strong enough to prevent some modified tscp from being the champion.
see http://www.talkchess.com/forums/1/message.html?353276
Uri

[Alejandro Dubrovsky Aleja]

Geschrieben von: / Posted by: Alejandro Dubrovsky Alejandro Dubrovsky at 07 March 2004 13:30:11:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk new results. geschrieben von: / posted by: Tord Romstad at 07 March 2004 10:00:08:

Hi Rudolf,
RDChess is stronger than BigLion at Blitz.
When the time control gets long, RDChess indeed messes a bit.
A BigLion beta once behaved similarly and I traced it to history heuristics
values growing large enough at long time controls to even override good
captures in move ordering.

This is not a real big problem.

It really depends on how your move ordering works. In previous versions of my engine at longer time controls I've seen it score some history moves above winning captures, this can really screw up your move ordering and branch factor. It wouldn't cause a blunder so maybe it isn't so bad but it can be a problem.

This is true, but the problem is not very difficult to fix. Whenever you increase an entry
in your history table, you check whether the new value of the entry exceeds some
threshold. If it does, loop through your history table and divide all entries by 2.

Ahhh, that's a good one Tord. I've clamped mine to the limit, but is an obviously better solution. Thanks!
alejandro



SmallPotato homepage

[Günther Simon]

Geschrieben von: / Posted by: Günther Simon at 07 March 2004 14:53:35:
Als Antwort auf: / In reply to: Re: Waster shall be Champ geschrieben von: / posted by: Uri Blass at 07 March 2004 12:58:08:

Waster needs to win both the games in hand to sneak ahead again,

No engines can stop Waster from being the Champ down here in Division 5.

I hope that your engine is strong enough to prevent some modified tscp from being the champion.
see http://www.talkchess.com/forums/1/message.html?353276
Uri

Doesnt this mean Waster is almost as near to TSCP than DanChess
was to Crafty according to Bob Hyatt? (not Dann)
This is not meant as an offense, but I think it would be good
if that issue could be enlightened a bit for non programmers.
Regards,
Güntehr

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 07 March 2004 16:18:27:
Als Antwort auf: / In reply to: Re: Waster shall be Champ geschrieben von: / posted by: Günther Simon at 07 March 2004 14:53:35:

Waster needs to win both the games in hand to sneak ahead again,

No engines can stop Waster from being the Champ down here in Division 5.

I hope that your engine is strong enough to prevent some modified tscp from being the champion.
see http://www.talkchess.com/forums/1/message.html?353276
Uri

Doesnt this mean Waster is almost as near to TSCP than DanChess
was to Crafty according to Bob Hyatt? (not Dann)
This is not meant as an offense, but I think it would be good
if that issue could be enlightened a bit for non programmers.
Regards,
Güntehr

I got the impression based on the post that the similiarity between tscp and Waster is bigger than the similiarity between Danchess and Crafty.
I think that it may better if Tom Kerrigen give his opinion about it.
My program also has some similiarity to tscp but it is not modified version of tscp and Tom told me in CCC that there is no problem even without seeing my source code.
I can say for example that nothing was copied from tscp's evaluation unlike the cases of Waster and Danchess.
Uri

[Marcus Prewarski]

Geschrieben von: / Posted by: Marcus Prewarski at 07 March 2004 17:29:57:
Als Antwort auf: / In reply to: Re: Waster shall be Champ geschrieben von: / posted by: Uri Blass at 07 March 2004 12:58:08:

Waster needs to win both the games in hand to sneak ahead again,

No engines can stop Waster from being the Champ down here in Division 5.

I hope that your engine is strong enough to prevent some modified tscp from being the champion.
see http://www.talkchess.com/forums/1/message.html?353276
Uri

Hmmm, thanks for the link Uri. I don't read CCC enough to see everything there. Time to start working on my engine again since it is still getting kicked around by TSCP.
Go Big Lion.
-Marcus