KUT-O2 WB tournament new started / Standing after 4 rounds

[Kurt Utzinger]

Geschrieben von: / Posted by: Kurt Utzinger at 16 July 2000 19:23:13:
WinBoard tournament KUT-O2
Pentium III 650/128
ponder=off
hashtables 32 MB
Nalimov tablebases
time control: 60/60'
I have in the meantime withdrawn Little Goliath 2000 v2.8 from my tournament (does not support Nalimov tablebases = great disadvantage)and replaced it by ZChess 1.61x (author Franck Zibi). And so far ZChess did very well whereas Comet B23 lost the first four games.
The results so far:
Round 1:
Amy07 - Yace020 (1-0)
AnMon509 - ZChess161x (1/2)
CometB23 - TCB0045 (0-1)
Crafty1711 - Gromit30 (0-1)
Round 2:
Yace020 - Gromit30 (1-0)
TCB0045 - Crafty1711 (1/2)
ZChess161x - CometB23 (1-0)
Amy07 - AnMon509 (1-0)
Round 3:
AnMon509 - Yace020 (1-0)
CometB23 - Amy07 (0-1)
Crafty1711 - ZChess161x (1-0)
Gromit30 - TCB0045 (0-1)
Round 4:
Yace020 - TCB0045 (1-0)
ZChess161x - Gromit30 (1/2)
Amy07 - Crafty1711 (1/2)
AnMon509 - CometB23 (1-0)
Standing after 4th round:
1) Amy 0.7 (3.5/4)
2) AnMon 5.06 (2.5/4)
3) TCB 0045 (2.5/4)
4) Crafty 17.11 (2.0/4)
5) Yace 0.20 (2.0/4)
6) ZChess x1.61 (2.0/4)
7) Gromit 3.0 (1.5/4)
8) Comet B23 (0.0/4)

[pete]

Geschrieben von: / Posted by: pete at 16 July 2000 19:40:55:
Als Antwort auf: / As an answer to: KUT-O2 WB tournament new started / Standing after 4 rounds geschrieben von: / posted by: Kurt Utzinger at 16 July 2000 19:23:13:

WinBoard tournament KUT-O2
Pentium III 650/128
ponder=off
hashtables 32 MB
Nalimov tablebases
time control: 60/60'
I have in the meantime withdrawn Little Goliath 2000 v2.8 from my tournament (does not support Nalimov tablebases = great disadvantage)and replaced it by ZChess 1.61x (author Franck Zibi).

I really have no problem with your decision , only with your reason .
Programs that may very well be the best in the endgame though not supporting Nalimov tablebases include :
Chessmaster 6000 , Genius , Rebel Century , Hiarcs 7.0 , Rebel Tiger
Only for engines that understand nothing about endgame tablebases should make a major difference.
I suspect for most engines a perfomance win of 10ELO when using tablebases against not using them at all is already a high estimate.

[Kurt Utzinger]

Geschrieben von: / Posted by: Kurt Utzinger at 16 July 2000 20:03:39:
Als Antwort auf: / As an answer to: Re: KUT-O2 WB tournament new started / Standing after 4 rounds geschrieben von: / posted by: pete at 16 July 2000 19:40:55:

WinBoard tournament KUT-O2
Pentium III 650/128
ponder=off
hashtables 32 MB
Nalimov tablebases
time control: 60/60'
I have in the meantime withdrawn Little Goliath 2000 v2.8 from my tournament (does not support Nalimov tablebases = great disadvantage)and replaced it by ZChess 1.61x (author Franck Zibi).

I really have no problem with your decision , only with your reason .
Programs that may very well be the best in the endgame though not supporting Nalimov tablebases include :
Chessmaster 6000 , Genius , Rebel Century , Hiarcs 7.0 , Rebel Tiger
Only for engines that understand nothing about endgame tablebases should make a major difference.
I suspect for most engines a perfomance win of 10ELO when using tablebases against not using them at all is already a high estimate.

Answer Kurt Utzinger:
Your comments are surely not wrong. A good example is Rebel Tiger with an excellent end game. Nevertheless you can sometimes observe that this strong program estimates an advantage of +3.00 and more before transposing into an drawn ending. If Rebel Tiger would use tablebases it could have done much better. For this reason I really doubt if your suspicion relating to ELO difference is correct.

[pete]

Geschrieben von: / Posted by: pete at 17 July 2000 01:38:57:
Als Antwort auf: / As an answer to: Re: KUT-O2 WB tournament new started / Standing after 4 rounds geschrieben von: / posted by: Kurt Utzinger at 16 July 2000 20:03:39:

Answer Kurt Utzinger:
Your comments are surely not wrong. A good example is Rebel Tiger with an excellent end game. Nevertheless you can sometimes observe that this strong program estimates an advantage of +3.00 and more before transposing into an drawn ending. If Rebel Tiger would use tablebases it could have done much better. For this reason I really doubt if your suspicion relating to ELO difference is correct.

An advantage of 3.00 heading to a drawn endgame is no problem ; evaluation scores are only good for deciding which move to make .
So you have to search for those games where the engine would have won with tablebases ( means found a better move ) and not won without them .
To my experience the number of these games is little but I don't have enough data to prove .
The experiment to test is clear .
Run two tournaments :
a.) 1st one with half of the programs not able to use TBs , 2nd half of the tablebase users using all tablebases they can eat .
b.) 1st half same as above , 2nd half the tablebase users but don't give them any TB .
This tournament could be played similar to the "Ponder-Noponder-tournament" and we would know positively .
So far we can only guess ; and my guess is 10 points ELO for tablebase usage based on yet unsufficient experience.
Btw , my guess for pondering is still 50 ELO as this is another untested thing still ; one half running with pondering , the other half without .
pete