Author: Gerd Isenberg
Date: 12:01:17 10/06/04
Go up one level in this thread
On October 06, 2004 at 04:38:28, Richard Pijl wrote: >On October 05, 2004 at 16:43:38, Gerd Isenberg wrote: > >>What, if best move after null is the reverse previous move, >>like doing three alternated null moves in a row. >>What can we conclude then? >> >>Thanks, >>Gerd > >Assume for arguments sake that white plays a move, black nullmoves and white >reverses his move. > >Then I think there are two possibilities: Ok, nullmove fails high or not. > >Either that the first move by white was really bad (worse than nullmove), and >black is likely to get a nullmove cutoff, e.g. a position with a back-rank mate >threat. Moving a rook off the back-rank is clearly about the worst move on the >board, so after an opponent nullmove it is moved back. Seems logical to trust the nullmove cutoff here. If the reverse move is best after nullmove and if a confirmation search fails, it might be worth to add some fractional plies. Nullmove's "child" has searched all successors and found the reverse move as best. > >Or that white is in zugzwang (and does the reverse move to do a nullmove >himsel), in which case the nullmove (the real one) will fail to produce a cutoff >as playing a move is better, or, when it does produce a cutoff, playing a move >is even better, so the cutoff is ok anyway. > >There may be some horizon effects with the reductions used though (as >effectively white nullmoved with R=normalR+2, but I guess this is normally not a >problem as we expect this 'nullmove' to fail low. > >Adjusting the score (or even not considering the reverse move) can be dangerous >as the scores propagated to the parent nodes will be hashed, increasing search >instability. No score adjustment in mind here. The reverse move refutes nullmove. May be there is a better refutation. It seems interesting, probably with some other conditions, to extend a bit too. > >Hope this helps >Richard. Yes, thanks Richard. It was only a hasty thought, but is probably worth to play a bit around. Gerd
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.