Author: Dieter Buerssner
Date: 12:22:20 05/10/03
Go up one level in this thread
On May 09, 2003 at 19:27:54, Ryan B. wrote: >My quiescences search is in a function I call just before the alpha beta search >if depth <= 0. I fear, I don't understand this sentence. >I changed it to if depth -3 <= 0 do quiescences else do null >move. Also, what exactly do you mean by "it" here. My suggestion was different, it seems (or my English comprehension is really bad ...). The idea of null move is: Even if I don't do anything, and I still "win" (=score > beta here), my position must be good (good enough for refuting my opponents last move). That alone would not help too much. The other idea is: How to detect if I would still win. It turns out, that a shallower search, than in normal alpha-beta is enough, to detect this rather reliably. In the discussed code snippet, depth -3 was used (often called R=2, another ply goes for the null move done). So, you should not do: "depth -3 <= 0 do quiescences else do null move", but rather do: subtract 3 from depth, if > 0 call "normal" alphabeta, else call the quiescence search. >I also added a doNull flag so it can be false if pieces >8 || inCheck. You probably meant "< 8" instead of "> 8". I don't know, how you define "pieces". If it is the number of pieces and pawns on both sides, this might be not the best choice, but still debatable. In a pawn endgame with 4 pawns on both sides, Zugzwang might be a very likely theme. This will make null move wrong, because "Even if I don't do anything, and I still "win" (=score > beta here), my position must be good" is not true anymore. Don't forget to set doNull to false, when the last move done was a null move. >Now I need to do more research as to when doNull should be false. Would a good >next step be to add transposition hash tables? From what I read it looks like >transposition hash tables should be fairly easy to add. It is certainly a good idea. The implementation is in principle rather easy, but subtle bugs might be difficult to find. Regards, Dieter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.