Author: Andreas Guettinger
Date: 10:22:41 03/04/04
Go up one level in this thread
On March 04, 2004 at 08:41:38, Uri Blass wrote: >On March 04, 2004 at 08:35:04, Andrew Wagner wrote: > >>Once again, you guys have provided a ton of useful info. Thanks so much! >> >>First of all, I have to apologize - I forgot to mention that I do use a history >>heuristic table. Just a very simple one...if a move fails high, I increment >>history(square.source, square.dest). Then I add the MVV/LA score if it's a >>capturing move, and sort by the sum of those two. I had hash tables, aspiration >>windows, and null-move, but I took them out to seek if I could get my move order >>% up without them. >> >>As far as killer moves, what usually defines them? I haven't read much about >>them. >> >>And with reference to Andreas' comment, do most of you agree? It makes sense to >>me that if both move order % and nps are low, it could be something in >>quiescence. What do you guys suggest I look at first? >> >>Thanks a ton for your help! Andrew > >I suggest that you do not care about nps now. What I wanted to say is that better move ordering will not fix low NPS. If you are sure you do it reasonably, I agree with you. I meant if the NPS are _very_ low, you might have the suspicion that something is wrong. You should at least receive a NPS similar to crafty, because not much eval is slowing you down etc. Let me give an example: In my very first search I got around 35KNPS (simple material counting eval), and I new something was wrong. It was just a first try, and for qsearch I generated all moves and then threw everything out that was not a capture. A very inefficient way to do it, but it was just a test. :) I then wrote a routine that only generates captures and got 300 kNPS. My qsearch was slowing me down terribly, and when I fixed that I was happy. > >order of moves seems to be your main problem. > Order of moves is the main bottleneck for the search going deep quickly, I agree. But better move ordering will reduce the NPS, unlikely increase it, because sophistic ordering hast a cost. regards Andy >from more nps you get only linear improvement but from better order of moves you >can get exponential improvement. > >Uri >Uri
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.