Author: Vincent Diepeveen
Date: 03:27:20 10/16/02
Go up one level in this thread
On October 16, 2002 at 03:47:01, Vladimir Medvedev wrote: >In his online interview Mr. CrazyBird told that DB did not use nullmoving >because it is not compatible with singular extensions. Could somebody explain >what are "singular" extensions (which difference with general meaning of >extensions technique? Are singular extensions used only for one move in the >node, while "common" - for all moves from the node? ) > >Why are they incompatible with nullmoving? Singular extensions are forced moves. However humans see forced moves way better than programs. For example after 1.e4,h5 2.qxh5 then for a computer rxh5 is also a forced move. For a human 2.Qxh5 is simply pathetic. But basically forced moves = a singular extension. Of course that's completely different from nullmove. In fact it works great together with nullmove. The problem is not so much nullmove, but singular extensions are very expensive. Suppose that you not only extend the search a move deeper after 1.e4,h5 2.qxh5,rxh5 but also in another zillion lines. That means your lines get longer and longer without being useful. Nullmove is good in limiting the length of not useful lines. It is questionable how good singular extensions are. I can definitely do a statement about this, because i use singular extensions in my own program. In the 80s, singular extensiosn were presented as some kind of holy grail which deep blue used. By then the first version of deep blue was chiptest getting 500k nodes a second. Despite that this is still more than what i get today with diep (about 70-80k nps single cpu k7 or new P4 northwood core), obviously we get far beyond the 8 ply search depth chiptest got by then. However in the 80s and also a the 90s up to about 1998 the word 'tactics' was important in chess. Software didn't search very deeply and it was common for programs to give away queens by a simple knightfork. That's unthinkable nowadays of course, but it was daily business in the 80s. so tactics was important. Singular extensions is of course good to get tactical stronger. So that's why singular extensiosn then were presented as getting a stronger program. However the stronger programs already got tactical stronger and stronger and we can argue about 1995-1997 where the first signs were visible that just tactical strength wasn't important (like deep blue never joining computer events anymore as it used to lose there). But it is very clear that in 1999 search depths were so big that tactical strength was less important. Now your question is of course why i in 2002 still have them in DIEP and others in their programs. The reason is dead simple: people kick on solving tactical jokes fast. So that's why i use them. But when search depths get deeper and deeper, i am sure i must throw them out one day. They are too expensive when search depths get above 11 ply. Deep Blue got 12.2 ply on average. the last few plies deep blue was forward pruning moves based upon just trying a few tactical moves. So it wasn't getting 12.2 ply positional. More like 10 ply positional. But tactical for sure it was strong. He's not giving arguments why he didn't use nullmove other than 'zugzwang' problems. they hardly tested things. he still to today doesn't even know what recursive nullmove is i bet.
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.