Considering whether it would make sense to permit nullmoves for both sides within a subtree, I expect that then a lot of unwanted position repetitions would occur. Thus equal nodes would be evaluated by very different search levels. Thus I think it would be better to ban such concurring nullmoves by permitting nullmoves for one color side only inside of any subtree.
I have been told that there is a double nullmove heuristic intended for a better handling of zugzwang positions. But that seems to cause just a special case of such concurring nullmoves and thus doubling nodes at different plys.
Have there been already thoughts on that matter?
concurring nullmoves (both colors) in subtrees?
Moderators: hgm, Rebel, chrisw
-
- Posts: 484
- Joined: Mon Mar 13, 2006 11:08 am
- Location: Klein-Gerau, Germany
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: concurring nullmoves (both colors) in subtrees?
First, you can solve the draw problem if you set the 50 move counter to zero when you try a null-move, and you use this to limit how far you go back in the rep list.smrf wrote:Considering whether it would make sense to permit nullmoves for both sides within a subtree, I expect that then a lot of unwanted position repetitions would occur. Thus equal nodes would be evaluated by very different search levels. Thus I think it would be better to ban such concurring nullmoves by permitting nullmoves for one color side only inside of any subtree.
I have been told that there is a double nullmove heuristic intended for a better handling of zugzwang positions. But that seems to cause just a special case of such concurring nullmoves and thus doubling nodes at different plys.
Have there been already thoughts on that matter?
However, I have tried that and it actually costs a couple of Elo. In thinking about it, terminating the search due to a repetition is not a bad event. The tree below a null-move is bogus anyway, and the quicker we "get out" the better.
Finally, I don't think your concern is something to lose sleep over. The sub-tree below a null-move is just another sub-tree. In many branches, the _other_ side could suddenly be winning, because so many moves in the tree are just outright blunders, and a null-move refutes blunders quicker than anything. I don't see why you would want to disable this quick refutation idea for one side, because the other side has previously tried a null earlier in the tree.
-
- Posts: 484
- Joined: Mon Mar 13, 2006 11:08 am
- Location: Klein-Gerau, Germany
Re: concurring nullmoves (both colors) in subtrees?
Does this mean not to reuse or store node results within subtrees below nullmoves?bob wrote:... The tree below a null-move is bogus anyway, and the quicker we "get out" the better. ...
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: concurring nullmoves (both colors) in subtrees?
No. I do exactly the opposite and do everything completely normally. Repetition checks. Hash probes and stores. Etc.smrf wrote:Does this mean not to reuse or store node results within subtrees below nullmoves?bob wrote:... The tree below a null-move is bogus anyway, and the quicker we "get out" the better. ...