Author: Robert Hyatt
Date: 20:19:20 02/26/04
Go up one level in this thread
On February 26, 2004 at 15:19:56, Gerd Isenberg wrote: >On February 26, 2004 at 13:28:48, Robert Hyatt wrote: > >>On February 26, 2004 at 04:23:01, Gerd Isenberg wrote: >> >>><snip> >>>>>To avoid the white versus side to move score chaos, it makes sense to mirror all >>>>>>current black to move positions vertically with white to move. >>>>> >>>>>Oups, with mirror vertically i mean white pawn on e2 becomes black pawn on e7 >>>>>and so on. >>>> >>>>i prefer also left right mirrorring. I bet diep is one of the few programs still >>>>displaying the same score then :) >>> >>>A good point, with respect to castle rights of course. >>>In my current program there are some asymmetries in (early) middlegame :-( >>> >>>In my new approach i will implement FRC-castles, but mirroring castles (rights) >>>is still not possible since king/rooks target squares are still asymmetric. >>> >>>Gerd >> >> >>I assume he is talking about post-castling positions or mirroring makes no >>sense. But in Crafty, I regularly run a "flip/flop" test on my evaluation. >> >>I set up a position and compute the static evaluation, then I flip (this mirrors >>along the line between the 4th and 5th ranks, so that white becomes black and >>vice-versa. The eval should change its sign but nothing else. I then use a >>"flop" which mirrors along the line between the e and f files, so that a pawns >>move to the h-file, etc. The eval must not change at all. I have a function >>"ev test" that does this for me, so that it flips, flops and flips again, to >>produce the four different positions you can get by mirroring along the two >>lines given above. The scores for flops can't change, but the scores for flips >>much change nothing but signs. I do eliminate the two asymmetry points in the >>eval when I test this because crafty evaluates blocked pawns asymmetrically, and >>I comment that out, and the king safety asymmetry is simply set to zero with the >>"eval asym 0" command... >> >>It catches some bugs now and then, that might be overlooked. > >I have that features too, even rotating a->h or h->a, which is particular >interesting with some endings. > >Anyway, i still have some slight left-right asymmetries in post-castling >positions. Exactly as Vincent pointed out. Queen- and kingside issues even in >late middlegame and probably isolated d- versus e-pawn in middlegame. > >[D]8/k1p1r3/p2b3p/5ppq/1np5/6P1/1R1N1P1P/R4QK1 b - - ; am c3 >[D]8/3r1p1k/p3b2p/qpp5/5pn1/1P6/P1P1N1R1/1KQ4R b - - > >difference: 0.04 similar search behaviour... > >Gerd OK. I don't do any of those, although my blocked pawn asymmetry stuff has to be turned off for the test to work cleanly...
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.