search statistics

Discussion of chess software programming and technical issues.

Moderator: Ras

tcusr
Posts: 325
Joined: Tue Aug 31, 2021 10:32 pm
Full name: tcusr

Re: search statistics

Post by tcusr »

likeawizard wrote: Sun Sep 25, 2022 1:18 pm
tcusr wrote: Sun Sep 25, 2022 12:13 pm i count AB nodes only if depth > 0 because otherwise i would count nodes twice wouldn't i? (when reporting them to the GUI, not specifically here).
to get the QS nodes i just changed the counter that is increased inside QS, this way AB + QS is the same number of nodes that my engine reports to the GUI when normally searching.
Yea, that all makes sense to me. But you report hitting 1 AB (initial position depth=1) and then 21 QS... Since there are only 20 legal moves in the starting position where does that one extra QS come from? Is it a miscalculation of QS nodes, like calculating a QS twice for some reason or including the 1 AB to the count of QS? Or does your QS actually looks at a reply of black in one of the 20 moves? The percentage you report implies 21/22 = 95.45% (QS/(AB+QS))
that's because it found a better move and started a new search without a null window but because it dives immediately to QS it does not count the AB node in which QS is called.