Author: Uri Blass
Date: 02:53:26 06/26/04
Go up one level in this thread
On June 25, 2004 at 23:00:15, Dann Corbit wrote: >On June 25, 2004 at 22:48:03, Uri Blass wrote: > >>On June 25, 2004 at 16:46:31, Dann Corbit wrote: >> >>>On June 25, 2004 at 16:36:19, Stuart Cracraft wrote: >>> >>>>Just a note that I got my hashing working in a >>>>quick hack program I've put together for other >>>>reasons and it was nasty even though I've been >>>>through it over and over through the decades. >>>>I empathize with other hashers who have felt or >>>>are feeling hashed. >>>> >>>>It is well worth it. A typical 8 ply search early >>>>in the game might reduce 50%+ in total nodes searched >>>>and 50%+ in total time. >>>> >>>>These are dwarfed by null move's effect though in >>>>the same positions (90% and 90%). >>>> >>>>So my question is, these are well-known methods to >>>>substantially reduce the number of nodes and amount >>>>of time for most searches -- I wonder if there is >>>>anything else that is as large and as comparable >>>>at these large 50%/90% types of reductions. >>> >>>What is your branching factor? >>> >>>The best programs have a branching factor between 2 and 3. >>> >>>If you already have a branching factor in that range, do not expect any more >>>really dramatic speedups. >> >>2 mistakes: >>1)it is possible to make the program worse by more pruning and a smaller >>branching factor so you cannot use the branching factor to decide if it is >>impossible to make dramtic speedup. > >If you have a big branching factor, then you can have a huge speedup without >damaging the search. All the best programs have a branching factor less than >3.0. If your program has a branching factor larger than this, then you can make >your program faster. Now, you can do stupid pruning and make it smaller and the >program will not be better. So it is necessary but not sufficient for a >condition. > >>2)I do not think that the best programs cannot expect more dramatic speedups in >>the qsearch thanks to new ideas. > >More than null move or hashing? >I doubt it. I do pruning that is not by hashing. I believe that it can be significantly improved. There are some known techniques like history based pruning. Note also that branching factor between 2 and 3 is a big range If you have branching factor of 2.94 and improve it to 2.1 then in case that you did not do stupid pruning you did a very big improvement If you search 14 plies forward then (2.94/2.1)^14=1.4^14 is the speed improvement that you get. practically you may search more than 14 plies so it is going to be equivalent to being 100 times faster. Uri
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.