Author: Omid David Tabibi
Date: 09:03:37 10/12/03
Go up one level in this thread
On October 12, 2003 at 11:45:07, Tord Romstad wrote: >On October 12, 2003 at 10:23:35, Omid David Tabibi wrote: > >>[D]6k1/5p2/3P2p1/7n/3QPP2/7q/r2N3P/6RK b - - 0 1 >> >>If you do checks everywhere in quiescence, you should see this immediately. > >If I did *all* checks everywhere in qsearch, I should see it instantly, yes. >But as you >remarked in the article at the beginning of this thread, this is too expensive. >Beyond >the first ply of qsearch, I have very strong restrictions about when and which >checks >to generate. > >>After 1...Rxd2 2.Qxd2 all the rest of the moves are checks until you detect draw >>by threefold repetition (maybe you've turned off repetition detection in >>quiescence? or your max extensions limit is too shallow...). HIARCS finds the >>move at the first iteration! > >I do repetition detection in quiescence, and I have no max extension limit. >Looking >closer on the position in question, it does seem a bit strange that I need such >a long >time to find the solution. The combination is not very deep. Perhaps I have a >bug >somewhere -- it's worth a closer look. > >>Falcon doesn't manage to solve number 12 either. > >Number 12 is very hard. But even solving number 10 and 11 in less than a >second >is very impressive, IMHO. > >>>You must have a very inefficient way of generating checks, I think. >> >>That's true. Only recently I added checks in quiescence to the engine, and so >>still haven't written a gen_checks() functions. However, the kind of attack >>tables I use result in a very speedy generation of captures, which results in a >>very optimized captures only quiescence. Adding checking moves will slow down >>the engine considerably anyway, even if I write a good gen_checks()... > >I am not so sure about that. Most of what you write above applies to my engine, >too. The attack tables are useful when generating checks, too. And of course >you >do not generate checks before you have generated and searched all captures and >they all failed low. > >>One thing I have to mention is that in the normal version I never check for >>check evasions in quiescence. If the side to move is in check and doesn't have >>any legal non-losing capture, I just return eval(). That's another reason why >>the normal quiescence is so fast. > >This sounds extremely dangerous to me -- doesn't this imply that you will not >always detect mates in qsearch? No checkmate can possibly take place at the first ply of quiescence, since I do the following in the main search: ... makemove(move); if (other side is in check) extension += 1; call_depth = depth - 1 + extension; if (call_depth > 0) score = -search(..., call_depth); else score = -quiescence(...); ... So if the other side is in check the depth will be extended instead of calling quiescence. But within the quiescence no checkmate can be detected in the normal version. >And doesn't this cause too many tactical mistakes? It to causes problems mainly in null-move pruning. Assume you are at depth = 3 and use R = 2. Your calling depth is 3-2-1=0, i.e., you directly call quiescence (after doing a null-move). Now it is the opponent's turn who checkmates you in the first ply of quiescence. Using checks in quiescence the checkmate will be detected, which will trigger a mate threat extension in the main search. Otherwise (in the normal version) just eval() is returned and assuming it is above beta we have a fail-high. All good and nice we are sure that our position is good enough to justify a cutoff, while in fact we are mate in 1! That's the main reason why checks in the first ply of quiescence contribute so much to tactical strength. > >>>It seems like checks in the qsearch is one of those things that works well in >>>some >>>programs, and not in others. Crafty, for instance, seems to do very well >>>without >>>any checks whatsoever, >> >>I wouldn't say so from a tactical point of view. Whenever the game turned >>tactical, Crafty didn't have any chance against Falcon with checks in >>quiescence. But Crafty did search deeper and played a better positional game. I >>must also add that Falcon uses a huge number of different extensions (I think >>only HIARCS has more extensions), and so maybe adding checks in quiescence on >>top of them all isn't such a good idea... > >Very interesting. > >>>but for me the results without checks are clearly worse. >>> >>>Other ideas that I have never been able to make work are recapture extensions >>>and >>>all sorts of nullmove pruning except plain R=3 (R=2, R=2.5, adaptive pruning and >>>verified >>>nullmove pruning are all clearly worse for me). >> >>In Falcon I conducted all the experiments I conducted on Genesis for the paper >>verified null-move pruning, and got the same results. Plain R=3 was too risky >>neglecting many tactical shots. I now use a modified version of verified >>null-move pruning. >> >>But maybe plain R=3 didn't work for me because I didn't have checks in >>quiescence, and so it resulted in a very inaccurate search. The only program >>I've heard which uses plain R=3 is DIEP, which does conduct checks everywhere in >>quiescence. > >This is very possible. I have experimented with values of R below 3, and with a >minimalistic qsearch without checks, but never in combination. Probably yet >another thing I should try ... > >Tord
This page took 0.01 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.