Author: Robert Hyatt
Date: 20:56:18 03/30/04
Go up one level in this thread
On March 30, 2004 at 15:01:21, Dan Honeycutt wrote: >On March 30, 2004 at 14:14:24, Robert Hyatt wrote: > >>On March 30, 2004 at 13:55:47, Andreas Guettinger wrote: >> >>>On March 30, 2004 at 13:16:00, Robert Hyatt wrote: >>> >>>>On March 30, 2004 at 12:29:42, Renze Steenhuisen wrote: >>>> >>>>> >>>>>Hi all, >>>>> >>>>>today I wanted to implement SEE, and use it in the Qsearch. My SEE isn't using >>>>>X-ray (discovered attacks) at the moment, but I already expected to see a boost >>>>>in FH-%. Unfortunately I didn't see it yet... >>>>> >>>>>Using it in the main-search did give an advantage of some 5% >>>>> >>>>>1> Is SEE not that important in Qsearch? >>>>> >>>>>2> With MVV/LVA I had about 15-20% Quiscence nodes, with SEE I have 25-30% >>>>> Qnodes!? >>>>> >>>>>Could some one give an overview of which move-ordering technique is responsible >>>>>for how much FH-%?! >>>>> >>>>>Thanks! >>>>> >>>>>Renze >>>> >>>>If you use SEE everywhere for move ordering, it will reduce the total tree size >>>>by about 10% over MVV/LVA. But if you use SEE to eliminate hopeless captures in >>>>the tree search, you will reduce the tree size by more than 50%, which is very >>>>significant. MVV/LVA can't be used to do this since it is a poor estimator for >>>>expected gain or loss... >>> >>> >>>Do you mean by eliminate hopless captures in the tree search: >>> >>>- give them the smallest priority in move ordering? >>> >>>or >>> >>>-prune them in the search tree? >> >>Prune in q-search. I am having a hard time typing until my hand recovers. I >>didn't write that very clearly. I eliminate losing captures _only_ in >>q-search... >> >> >> >>> >>>regards >>>Andy > >Related question which takes little typing - is it worth the cost to include >pins in the SEE? > >Dan H. I don't believe so. But this should be tested of course... q-search is already way inaccurate.
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.