Stars Only-4d (at tournament time control, Aristarch 4.41)

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 16 June 2004 15:05:00:

Stars Only-4d (at tournament time control, Aristarch 4.41)
================================================================
I continue to run a tournament (actually 10-game matches) named 'Stars
Only' for the very best free engines at 7-hour tournament time control.
Time control: 40/7200:20/3600:1800
Celeron 2.6GHZ 256MB; hash settings is up to 38MB in sum for each engine;
resign is set to 6; few 5-men Nalimov's TBs.
Arena is used to arrange the event.
The last addition is Aristarch 4.41 (it takes 8-9th place).

Here are the match results of Aristarch 4.41
--------------------------------------------
vs Delfi 4.4 1.5:8.5
vs Ruffian 1.0.1 4.5:5.5
vs SOS 4 for Arena 4.0:6.0
vs AnMon 5.30 6.0:4.0
vs Chinito 3.25 6.5:3.5
vs Ktulu 4.1 4.5:5.5
vs Crafty 16.19 4.0:6.0
vs SOS.3 for Arena 3.5:6.5
vs SOS 11.99 6.0:4.0
vs Chinito 3.1 5.5:4.5
Assembled cross-table of the event
===================================
Stars Only
Celeron 2.6GHZ 128MB, 2003.10.25 - 2004.05.14
S R B 1 2 3 4 5 6 7 8 9 0 1
----------------------------------------------------------------------------------------------------------------------------------------------------------------
1 Delfi 4.4 68 1 3213.00 XXXXXXXXXX 00=00==110 0=1=110=1= 1=11===1=1 1==1==1=== 1=11=0=11= 0=111111=1 0111=111=1 11101111=1 1110==0110 =01=110111
2 Ruffian 1.0.1 63= 2 3086.00 11=11==001 XXXXXXXXXX 11=110=101 =1=101==1= =0==101=10 0=0=111=01 010=0===11 1011=101=1 010==1=101 10=11=1111 1==1==111=
3 SOS 4 for Arena 56= 3 2700.75 1=0=001=0= 00=001=010 XXXXXXXXXX 1=10=1101= ==1===0100 1==11=1=01 ==0=11==11 0=1==101=0 101111=01= =11=1===10 =1011=1===
4 AnMon 5.30 50 4 2398.75 0=00===0=0 =0=010==0= 0=01=0010= XXXXXXXXXX 01110011=1 111001011= 1110101011 1=1=0====1 =01010100= 100=0011=0 111==0011=
5 Crafty 16.19 46= 5-7 2322.50 0==0==0=== =1==010=01 ==0===1011 10001100=0 XXXXXXXXXX 1100=0==01 =0001=0==1 =010110=== =101111=00 101001010= 10==1===0=
6 SOS.3 for Arena 46= 5-7 2270.25 0=00=1=00= 1=1=000=10 0==00=0=10 000110100= 0011=1==10 XXXXXXXXXX 11000101=1 0101=00=10 =1=01=1101 =1===0=0=1 01101=1=10
7 Ktulu 4.1 46= 5-7 2253.50 1=000000=0 101=1===00 ==1=00==00 0001010100 =1110=1==0 00111010=0 XXXXXXXXXX 1=000=1011 =1====1010 100110==10 =1101=1==1
8 Chinito 3.25 45 8-9 2185.75 1000=000=0 0100=010=0 1=0==010=1 0=0=1====0 =101001=== 1010=11=01 0=111=0100 XXXXXXXXXX 1=000=0=10 1=111=1000 =======110
9 Aristarch 4.41 45 8-9 2183.00 00010000=0 101==0=010 010000=10= =10101011= =010000=11 =0=10=0010 =0====0101 0=111=1=01 XXXXXXXXXX 01101==110 101=101010
0 SOS 11.99 43 10 2135.50 0001==1001 01=00=0000 =00=0===01 011=1100=1 010110101= =0===1=1=0 011001==01 0=000=0111 10010==001 XXXXXXXXXX 10=000=110
1 Chinito 3.1 39= 11 1956.50 =10=001000 0==0==000= =0100=0=== 000==1100= 01==0===1= 10010=0=01 =0010=0==0 =======001 010=010101 01=111=001 XXXXXXXXXX
----------------------------------------------------------------------------------------------------------------------------------------------------------------
Games 550/550, +217 -159 =174


The next events are qualifiers for new versions and engines.
All the games can be found at
http://www.digichess.gr/games/so4d.zip
====================================================
For more info on the Stars Only event apply to
http://www.digichess.gr/infiniteloop/home.php
Igor Gorelikov

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 16 June 2004 15:13:03:
Als Antwort auf:/In reply to: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Igor Gorelikov at 16 June 2004 15:05:00:


Hi Igor,
Why do you include different versions of the same engine? Wouldn't letting other engines qualify instead be more interesting?
Fabien.

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 16 June 2004 15:47:10:
Als Antwort auf:/In reply to: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Igor Gorelikov at 16 June 2004 15:05:00:

Stars Only-4d (at tournament time control, Aristarch 4.41)
================================================================
I continue to run a tournament (actually 10-game matches) named 'Stars
Only' for the very best free engines at 7-hour tournament time control.
Time control: 40/7200:20/3600:1800

I wonder if all the engines know about this time control.
Winboard does not support this time control and it is possible that one of the reasons that delfi performs better in your tournament relative to other tournaments is that delfi support this time control correctly unlike part of the other programs(other programs may for example believe after move 40 that they have an hour for moves 41-80 and play too quickly at moves 41-60 so they have an hour to finish the game at move 60 and later they may believe that they have an hour for moves 61-80 and play too slowly getting to serious time trouble at move 80 when they have few seconds to finish the game).
Uri

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 16 June 2004 15:53:13:
Als Antwort auf:/In reply to: Re: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Fabien Letouzey at 16 June 2004 15:13:03:

Hi Igor,
Why do you include different versions of the same engine? Wouldn't letting other engines qualify instead be more interesting?
Fabien.

New versions of the stars are included only via qualifiers so that they should beat the previous ones.
And I let any engine suspected to be a star participate in the qualifiers.
Rules of SO
============
"1. First included into tournament are only three top engines by rating list
of New IL.
2. New additions are made by qualifiers.
3. Participants in the qualifiers can be
a) any new versions of the stars.
b) any new versions of the top engines (according to rating list of
New IL) suspected to be stronger.
c) any new engine suspected to be a star.
4. If no promotion is made, new matches are running between stars.
5. Qualifiers are arranged in a way that any engine get 18 games at least."
Igor

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 16 June 2004 15:56:37:
Als Antwort auf:/In reply to: Re: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Uri Blass at 16 June 2004 15:47:10:

Stars Only-4d (at tournament time control, Aristarch 4.41)
================================================================
I continue to run a tournament (actually 10-game matches) named 'Stars
Only' for the very best free engines at 7-hour tournament time control.
Time control: 40/7200:20/3600:1800

I wonder if all the engines know about this time control.
Winboard does not support this time control and it is possible that one of the reasons that delfi performs better in your tournament relative to other tournaments is that delfi support this time control correctly unlike part of the other programs(other programs may for example believe after move 40 that they have an hour for moves 41-80 and play too quickly at moves 41-60 so they have an hour to finish the game at move 60 and later they may believe that they have an hour for moves 61-80 and play too slowly getting to serious time trouble at move 80 when they have few seconds to finish the game).
Uri

I watch many games live and cannot confirm your guess.
Sorry,
Igor

[Niyaz Khasanov]

Geschrieben von:/Posted by: Niyaz Khasanov at 16 June 2004 16:34:51:
Als Antwort auf:/In reply to: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Igor Gorelikov at 16 June 2004 15:05:00:

Stars Only-4d (at tournament time control, Aristarch 4.41)
================================================================
I continue to run a tournament (actually 10-game matches) named 'Stars
Only' for the very best free engines at 7-hour tournament time control.
Time control: 40/7200:20/3600:1800
Celeron 2.6GHZ 256MB; hash settings is up to 38MB in sum for each engine;
resign is set to 6; few 5-men Nalimov's TBs.
1 Delfi 4.4 68 1
2 Ruffian 1.0.1 63= 2
..................................(snip).............................
====================================================

There are five possibilities that can explain the difference between Delfi results in Stars Only and WBEC tourneys:
1. The hardware difference.
2. The time control difference.
3. Delfi has a weak pondering function. (WBEC is on a dual computer).
4. Delphi may use a part of processor's time when the opponent's time.
5. Too good settings in Igor's delfi.ini file
One can prove it by two gauntlets on the same DUAL machine: one with ponder=on and another with ponder=off. The time control can be blitz.
I have no dual computers

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 16 June 2004 16:57:15:
Als Antwort auf:/In reply to: Re: Stars Only-4d (at tournament time control, Aristarch 4.41) geschrieben von:/posted by: Niyaz Khasanov at 16 June 2004 16:34:51:

Stars Only-4d (at tournament time control, Aristarch 4.41)
================================================================
I continue to run a tournament (actually 10-game matches) named 'Stars
Only' for the very best free engines at 7-hour tournament time control.
Time control: 40/7200:20/3600:1800
Celeron 2.6GHZ 256MB; hash settings is up to 38MB in sum for each engine;
resign is set to 6; few 5-men Nalimov's TBs.
1 Delfi 4.4 68 1
2 Ruffian 1.0.1 63= 2
..................................(snip).............................
====================================================

There are five possibilities that can explain the difference between Delfi results in Stars Only and WBEC tourneys:
1. The hardware difference.
2. The time control difference.
3. Delfi has a weak pondering function. (WBEC is on a dual computer).
4. Delphi may use a part of processor's time when the opponent's time.
5. Too good settings in Igor's delfi.ini file
One can prove it by two gauntlets on the same DUAL machine: one with ponder=on and another with ponder=off. The time control can be blitz.
I have no dual computers

I can add to it one more guess:
6. Delfi 4.4 likes more Celeron than Athlon.
So we need more tests on other computers to draw conclusions.
Igor