I was wondering if I could get a few authors to run a round robin with the following versions of their programs, so I could see how their strength scales with/without hashing and null move reductions. I am getting some improvement with each, but not nearly what I think I should be getting, so I was hoping for a few data points to compare against.
1. Hash table and null move enabled.
2. Hash table disabled.
3. Null move disabled.
4. No hash table or null move.
Tournament request
Moderators: hgm, Rebel, chrisw
-
- Posts: 686
- Joined: Thu Mar 03, 2011 4:57 pm
- Location: Germany
Re: Tournament request
Hi Robert,
you will certainly get a few authors to release the source code of their engine, in fact a lot already did that.
This should enable you to run this type of tournament but it of course involves some effort on your side. Don't wait until someone just does it for you. This is not how it works.
you will certainly get a few authors to release the source code of their engine, in fact a lot already did that.
This should enable you to run this type of tournament but it of course involves some effort on your side. Don't wait until someone just does it for you. This is not how it works.
-
- Posts: 855
- Joined: Sun May 23, 2010 1:32 pm
Re: Tournament request
It's not so easy to remove hash, many engine use it to follow principal variation and for move ordering
-
- Posts: 893
- Joined: Mon Jan 15, 2007 11:23 am
- Location: Warsza
Re: Tournament request
What does it mean to "remove hash"?
Not doing hash cutoffs is trivial. not using hash for move sorting requires implementing a replacement method, like sorting PV moves upfront the way TSCP does it.
Not doing hash cutoffs is trivial. not using hash for move sorting requires implementing a replacement method, like sorting PV moves upfront the way TSCP does it.
Pawel Koziol
http://www.pkoziol.cal24.pl/rodent/rodent.htm
http://www.pkoziol.cal24.pl/rodent/rodent.htm
-
- Posts: 855
- Joined: Sun May 23, 2010 1:32 pm
Re: Tournament request
Yes you are right.
I intended to say: removing nill move is very easy. Removing the hash is not so simple because it's used for more than one thing. Cutoff in the search, following principal variation and move ordering.
I intended to say: removing nill move is very easy. Removing the hash is not so simple because it's used for more than one thing. Cutoff in the search, following principal variation and move ordering.
-
- Posts: 558
- Joined: Sat Mar 25, 2006 8:27 pm
Re: Tournament request
Yes, of course I could do it myself with open source engines. The down side of that is I have to invest a disproportionate amount of time to get familiar with their engine to figure out exactly where/how to make those changes without accidentally breaking something I didn't intend to, while for the author, it's an obvious one-line change. Even compiling an unfamiliar engine can be time consuming, as you have to figure out the correct make files, and they may be using libraries specific to their OS/compiler.tpetzke wrote:Hi Robert,
you will certainly get a few authors to release the source code of their engine, in fact a lot already did that.
This should enable you to run this type of tournament but it of course involves some effort on your side. Don't wait until someone just does it for you. This is not how it works.
I'm not demanding anything from anyone. But people do simple requests from others all the time, whether testing an idea or seeing what their engine returns for a given FEN.
-
- Posts: 558
- Joined: Sat Mar 25, 2006 8:27 pm
Re: Tournament request
In my test code, I simply added a "return valueUnknown" to the beginning of my ProbeHash routine, so I get NULL results no matter how full the table gets. And yes, this may also hamper other aspects like move ordering.elcabesa wrote:Yes you are right.
I intended to say: removing nill move is very easy. Removing the hash is not so simple because it's used for more than one thing. Cutoff in the search, following principal variation and move ordering.
-
- Posts: 859
- Joined: Mon Aug 10, 2009 10:05 pm
- Location: Italy
- Full name: Stefano Gemma
Re: Tournament request
This thread maybe can help:
http://stackoverflow.com/questions/1650 ... ing-factor
look at the response with the 8 steps but still you should adapt it at your engine.
http://stackoverflow.com/questions/1650 ... ing-factor
look at the response with the 8 steps but still you should adapt it at your engine.
Author of Drago, Raffaela, Freccia, Satana, Sabrina.
http://www.linformatica.com
http://www.linformatica.com