Worthless extension ideas...

Programming Topics (Computer Chess) and technical aspects as test techniques, book building, program tuning etc

Moderator: Andres Valverde

Worthless extension ideas...

Postby mjlef » 06 Jun 2006, 06:52

I thought it might be useful to pass on my results in ideas that did not work (to save you the bother of trying them). Last night, the test was for extending any safe move by a passed pawn whenever total matrial on the board was under 2000 points (where a pawn=100, Knight-300, etc). Extensions were allowed even in null window searches. "Safe" means to any un-attacked sqaure, or if attacked, at least defended (I did not do a full SEE). Against another program, my program dropped about 60 rating points. Perhaps more limits might help (like only extending if the 5th or greater rank, or extending only when material is even less). If you have a better passed pawn extension rule, let me know.

Since most all of my extension ideas prove bad, expect more worthless results here soon!

Mark
mjlef
 
Posts: 64
Joined: 29 Mar 2006, 22:04

Re: Worthless extension ideas...

Postby Daniel Shawul » 06 Jun 2006, 08:39

mjlef wrote:I thought it might be useful to pass on my results in ideas that did not work (to save you the bother of trying them). Last night, the test was for extending any safe move by a passed pawn whenever total matrial on the board was under 2000 points (where a pawn=100, Knight-300, etc). Extensions were allowed even in null window searches. "Safe" means to any un-attacked sqaure, or if attacked, at least defended (I did not do a full SEE). Against another program, my program dropped about 60 rating points. Perhaps more limits might help (like only extending if the 5th or greater rank, or extending only when material is even less). If you have a better passed pawn extension rule, let me know.

Since most all of my extension ideas prove bad, expect more worthless results here soon!

Mark


What i use in combination with reductions , is not to extend anything but also not to reduce any move that you think is interesting enough to deserve an extension. This way i was able to throw out mate_threat, passed pawn extensions.
You can try not to reduce any passed pawn move to ranks >=5 without blowing up your search tree.

Daniel
User avatar
Daniel Shawul
 
Posts: 366
Joined: 28 Sep 2004, 09:33
Location: Ethiopia

Re: Worthless extension ideas...

Postby mjlef » 06 Jun 2006, 10:35

Daniel Shawul wrote:
mjlef wrote:I thought it might be useful to pass on my results in ideas that did not work (to save you the bother of trying them). Last night, the test was for extending any safe move by a passed pawn whenever total matrial on the board was under 2000 points (where a pawn=100, Knight-300, etc). Extensions were allowed even in null window searches. "Safe" means to any un-attacked sqaure, or if attacked, at least defended (I did not do a full SEE). Against another program, my program dropped about 60 rating points. Perhaps more limits might help (like only extending if the 5th or greater rank, or extending only when material is even less). If you have a better passed pawn extension rule, let me know.

Since most all of my extension ideas prove bad, expect more worthless results here soon!

Mark


What i use in combination with reductions , is not to extend anything but also not to reduce any move that you think is interesting enough to deserve an extension. This way i was able to throw out mate_threat, passed pawn extensions.
You can try not to reduce any passed pawn move to ranks >=5 without blowing up your search tree.

Daniel

I agree completely.

I do have rules in the program along these lines (not reducing moves that change the King Safety score a lot, Passed Pawn scores a lot, pushing passed pawns, threaten a bigger valued piece...). They definitely help. I find that any reduction technique has to have some rules to cover these things, or risk overlooking too much. Of course, tuning the rules is a bit tricky (reaching the correct point of still having enough reductions to increase overall search depth without tossing critical lines). I think my current rules in the version without the PP extension is to not reduce any passed pawn move. Maybe your rank>=5 is a good refinement. I will give it a try.
mjlef
 
Posts: 64
Joined: 29 Mar 2006, 22:04


Return to Programming and Technical Discussions

Who is online

Users browsing this forum: No registered users and 23 guests