Bad Pruning

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: Bad Pruning

Post by Onno Garms »

Ferdy wrote: To improve the idea, and if SEE is available at that point, then try a combination of SEE and search with reduced depth. For example if SEE is bad then search this with a reduced depth as a verification, if it losses material or losses position value then prune/reduce.
Maybe worth a try, but I'm not sure if that would be an improvement. For example, moves leaving your queen in gardez have good see but should nevertheless be pruned.
User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: Bad Pruning

Post by Onno Garms »

mcostalba wrote: If I have understood correctly this is a kind of LMR launched on (beta - margin) instead of just beta....
This is not the way I viewed it, but maybe.

Note that I try bad pruning on all moves, not just late ones. Also depth reduction is much heavier and as you mentioned there is a margin.
User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: Bad Pruning

Post by Onno Garms »

Teemu Pudas wrote:This is Multi-ProbCut. http://www.cs.ubc.ca/~jiang/papers/mpc_main.pdf
In deed. Thank you for pointing out.

Actually "bad pruning" is a special parametrization of Multi-ProbCut. I found my values by trial and error, without the statistical background of Multi-PropCut.
uaf
Posts: 98
Joined: Sat Jul 31, 2010 8:48 pm
Full name: Ubaldo Andrea Farina

Re: Bad Pruning

Post by uaf »

Interesting. I tried MPC several years ago in my engine Chiron but it never worked. I'll soon give your "bad pruning" algorithm a try.

How many points did Onno gain?
User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: Bad Pruning

Post by Onno Garms »

uaf wrote:How many points did Onno gain?

Code: Select all

Rank Name                     Elo    +    - games score oppo. draws 
   5 Onno 0-10-9-w             60   18   18  1040   59%    -6   31% 
   8 Onno 0-10-9               15   18   18  1040   53%    -6   33%
Ferdy
Posts: 4840
Joined: Sun Aug 10, 2008 3:15 pm
Location: Philippines

Re: Bad Pruning

Post by Ferdy »

Onno Garms wrote:
Ferdy wrote: To improve the idea, and if SEE is available at that point, then try a combination of SEE and search with reduced depth. For example if SEE is bad then search this with a reduced depth as a verification, if it losses material or losses position value then prune/reduce.
Maybe worth a try, but I'm not sure if that would be an improvement. For example, moves leaving your queen in gardez have good see but should nevertheless be pruned.
Can you elaborate further on this,
"moves leaving your queen in gardez have good see"
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Bad Pruning

Post by mcostalba »

Ferdy wrote: Can you elaborate further on this,
"moves leaving your queen in gardez have good see"
It means moves that pin the queen.
Ferdy
Posts: 4840
Joined: Sun Aug 10, 2008 3:15 pm
Location: Philippines

Re: Bad Pruning

Post by Ferdy »

mcostalba wrote:
Ferdy wrote: Can you elaborate further on this,
"moves leaving your queen in gardez have good see"
It means moves that pin the queen.
Thanks, very aggressive prunning indeed, even a move with good see is pruned because of a bad result of a search to a reduced depth :)
User avatar
Onno Garms
Posts: 224
Joined: Mon Mar 12, 2007 7:31 pm
Location: Bonn, Germany

Re: Bad Pruning

Post by Onno Garms »

mcostalba wrote: It means moves that pin the queen.
Not really. It means that when you opponent attacks your queen with say a rook, you should do something to escape that attack. ("Gardez" means that your queen is under attack as "check" means that your king is under attack.)

Take 1. e4 d5 2. ed Qd5 3. Nc3. SEE only considers the moving piece, so 3... Nf6 seems OK. In spite of good SEE 3...Nf6 is a very bad move because it leaves your queen in gardez. "bad pruning" will prune 3...Nf6
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: Bad Pruning

Post by mcostalba »

Onno Garms wrote:
mcostalba wrote: It means moves that pin the queen.
Not really. It means that when you opponent attacks your queen with say a rook, you should do something to escape that attack. ("Gardez" means that your queen is under attack as "check" means that your king is under attack.)

Take 1. e4 d5 2. ed Qd5 3. Nc3. SEE only considers the moving piece, so 3... Nf6 seems OK. In spite of good SEE 3...Nf6 is a very bad move because it leaves your queen in gardez. "bad pruning" will prune 3...Nf6
Yes, I meant this but I expressed with bad wording: pinning the queen is always SEE negative move if the attacking piece is not the opponent queen.