Persistent hashes - performance

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Post Reply
Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Persistent hashes - performance

Post by Rodolfo Leoni » Wed Dec 06, 2017 1:03 pm

Hi everybody. I've been away these last months but I'm here again and I still burden you with this subject. :)

I tested persistent hashes at LSS server so far. It's a great correspondence games server and I recommend it to anybody who's interested to play correspondence chess for free.

I'm using the persistent hashes feature of SugaR (original code by Daniel Jose, based on an idea by HGM) and the performance is:

+19 =57 -0

I played 4 games against players who don't like engines, but all other wins were against 2000+ ELO players, so I consider this result really significant. The best opponent has been FIDE GM Igor Kovalenko, and I got +1 =2 -0 against him, but to be honest the win was due to a blunder in opening so I'd say I drew against Kovalenko.

I invite, again, the top engine programmers to reconsider the idea to include Daniel's code as an UCI option. It doesn't hurt anybody if it's an option...

Bye,
Rodolfo
F.S.I. Chess Teacher

User avatar
kranium
Posts: 1767
Joined: Thu May 29, 2008 8:43 am

Re: Persistent hashes - performance

Post by kranium » Wed Dec 06, 2017 2:46 pm

Rodolfo Leoni wrote:Hi everybody. I've been away these last months but I'm here again and I still burden you with this subject. :)

I tested persistent hashes at LSS server so far. It's a great correspondence games server and I recommend it to anybody who's interested to play correspondence chess for free.

I'm using the persistent hashes feature of SugaR (original code by Daniel Jose, based on an idea by HGM) and the performance is:

+19 =57 -0

I played 4 games against players who don't like engines, but all other wins were against 2000+ ELO players, so I consider this result really significant. The best opponent has been FIDE GM Igor Kovalenko, and I got +1 =2 -0 against him, but to be honest the win was due to a blunder in opening so I'd say I drew against Kovalenko.

I invite, again, the top engine programmers to reconsider the idea to include Daniel's code as an UCI option. It doesn't hurt anybody if it's an option...

Bye,
Rodolfo

With or without persistent hash...
is this an entirely unexpected result against human opponents?

Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Re: Persistent hashes - performance

Post by Rodolfo Leoni » Wed Dec 06, 2017 4:16 pm

kranium wrote:
Rodolfo Leoni wrote:Hi everybody. I've been away these last months but I'm here again and I still burden you with this subject. :)

I tested persistent hashes at LSS server so far. It's a great correspondence games server and I recommend it to anybody who's interested to play correspondence chess for free.

I'm using the persistent hashes feature of SugaR (original code by Daniel Jose, based on an idea by HGM) and the performance is:

+19 =57 -0

I played 4 games against players who don't like engines, but all other wins were against 2000+ ELO players, so I consider this result really significant. The best opponent has been FIDE GM Igor Kovalenko, and I got +1 =2 -0 against him, but to be honest the win was due to a blunder in opening so I'd say I drew against Kovalenko.

I invite, again, the top engine programmers to reconsider the idea to include Daniel's code as an UCI option. It doesn't hurt anybody if it's an option...

Bye,
Rodolfo

With or without persistent hash...
is this an entirely unexpected result against human opponents?
Hi Norman. :)
Almost everybody uses chess engines there. So it's almost impossible to win a game if they use a good opening book too. Without PHs, that'd be quite surprising. With PHs, I use them in infinite analysis trying variations on the GUI. The score propagation down to the root gave me a lot of confidence, so it was decisive.
F.S.I. Chess Teacher

User avatar
kranium
Posts: 1767
Joined: Thu May 29, 2008 8:43 am

Re: Persistent hashes - performance

Post by kranium » Wed Dec 06, 2017 4:19 pm

Rodolfo Leoni wrote:
kranium wrote:
Rodolfo Leoni wrote:Hi everybody. I've been away these last months but I'm here again and I still burden you with this subject. :)

I tested persistent hashes at LSS server so far. It's a great correspondence games server and I recommend it to anybody who's interested to play correspondence chess for free.

I'm using the persistent hashes feature of SugaR (original code by Daniel Jose, based on an idea by HGM) and the performance is:

+19 =57 -0

I played 4 games against players who don't like engines, but all other wins were against 2000+ ELO players, so I consider this result really significant. The best opponent has been FIDE GM Igor Kovalenko, and I got +1 =2 -0 against him, but to be honest the win was due to a blunder in opening so I'd say I drew against Kovalenko.

I invite, again, the top engine programmers to reconsider the idea to include Daniel's code as an UCI option. It doesn't hurt anybody if it's an option...

Bye,
Rodolfo

With or without persistent hash...
is this an entirely unexpected result against human opponents?
Hi Norman. :)
Almost everybody uses chess engines there. So it's almost impossible to win a game if they use a good opening book too. Without PHs, that'd be quite surprising. With PHs, I use them in infinite analysis trying variations on the GUI. The score propagation down to the root gave me a lot of confidence, so it was decisive.
Ah understood now...
I missed that it's a correspondence chess site and everyone is using engines, my bad.
Congratulations on a great result!

Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Re: Persistent hashes - performance

Post by Rodolfo Leoni » Wed Dec 06, 2017 4:25 pm

kranium wrote:
Rodolfo Leoni wrote:
kranium wrote:
Rodolfo Leoni wrote:Hi everybody. I've been away these last months but I'm here again and I still burden you with this subject. :)

I tested persistent hashes at LSS server so far. It's a great correspondence games server and I recommend it to anybody who's interested to play correspondence chess for free.

I'm using the persistent hashes feature of SugaR (original code by Daniel Jose, based on an idea by HGM) and the performance is:

+19 =57 -0

I played 4 games against players who don't like engines, but all other wins were against 2000+ ELO players, so I consider this result really significant. The best opponent has been FIDE GM Igor Kovalenko, and I got +1 =2 -0 against him, but to be honest the win was due to a blunder in opening so I'd say I drew against Kovalenko.

I invite, again, the top engine programmers to reconsider the idea to include Daniel's code as an UCI option. It doesn't hurt anybody if it's an option...

Bye,
Rodolfo

With or without persistent hash...
is this an entirely unexpected result against human opponents?
Hi Norman. :)
Almost everybody uses chess engines there. So it's almost impossible to win a game if they use a good opening book too. Without PHs, that'd be quite surprising. With PHs, I use them in infinite analysis trying variations on the GUI. The score propagation down to the root gave me a lot of confidence, so it was decisive.
Ah understood now...
I missed that it's a correspondence chess site and everyone is using engines, my bad.
Congratulations on a great result!
Thanks... But I'm only a good player, not a GM at all. ;)
I got it thanks to the "Never clear hash" UCI option, which works with infinite mode only.
F.S.I. Chess Teacher

Post Reply