Discussion of chess software programming and technical issues.
Moderators: hgm , Rebel , chrisw
xr_a_y
Posts: 1871 Joined: Sat Nov 25, 2017 2:28 pm
Location: France
Post
by xr_a_y » Tue Jul 06, 2021 8:04 pm
Have you ever consider using a negative (little !) margin for static null move pruning at depth 1 ?
We often seek for parameters in a formula like this one
Code: Select all
if ( staticScore > beta + depth * slope + offset ) return staticScore
And often "offset" is 0 and is omitted and coeff is something around 100cp.
I am wondering if (and testing this currently) using offset around -100 or -150 and tune slope might be good, even such that slope + offset is a little negative.
hgm
Posts: 27796 Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller
Post
by hgm » Wed Jul 07, 2021 7:43 pm
This is just redefining d=1 as a QS level, right? So it will not hurt, but just lead to a funny redefinition of 'depth', exaggerating all depths by 1, and otherwise doing the same search.
Koivisto
Posts: 50 Joined: Fri Sep 04, 2020 10:30 pm
Full name: Kim Kahre
Post
by Koivisto » Thu Jul 08, 2021 11:22 am
we have this in Koivisto search
Code: Select all
if (depth == 1 && staticEval > beta && ownThreats && !enemyThreats)
return beta;
Threats are defined as smaller piece attacking larger.
hgm
Posts: 27796 Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller
Post
by hgm » Thu Jul 08, 2021 11:29 am
So what are 'threats'?
Koivisto
Posts: 50 Joined: Fri Sep 04, 2020 10:30 pm
Full name: Kim Kahre
Post
by Koivisto » Thu Jul 08, 2021 3:54 pm
hgm wrote: ↑ Thu Jul 08, 2021 11:29 am
So what are 'threats'?
Threats are defined as a smaller piece attacking a larger piece. Pawn attacking knight for example.