Author: Michel Langeveld
Date: 12:18:44 01/18/00
Go up one level in this thread
On January 18, 2000 at 14:22:12, Steve Coladonato wrote: >On January 18, 2000 at 14:06:14, José de Jesús García Ruvalcaba wrote: > >>On January 18, 2000 at 12:33:56, Steve Coladonato wrote: >> >>>On January 18, 2000 at 11:40:00, Michel Langeveld wrote: >>> >>>>On January 18, 2000 at 10:40:30, Steve Coladonato wrote: >>>> >>>>>Are tablebases basically a set of finite positions that have pointers to >>>>>subsequent positions (most probably positions leading to a win)? And if so, is >>>>>the basic algorithm to go to the next position that in turn will have a pointer >>>>>to a "won" position? I am also concluding that once a program starts to use a >>>>>tablebase, it no longer does any "real" processing, just pointer evaluation. Is >>>>>this basically it or am I way off the mark here? >>>>> >>>>>Thanks. >>>>> >>>>>Steve >>>> >>>>TB's are files which has no pointer at all. It contains just all positions of a >>>>certain endgame with saved the number of moves it takes to reach mate. Because >>>>we haven't exceed mate in 255 with TB's we can put save 1 byte to an epdrecord. >>>> >>>>Since all the positions are saved sequential it's possible to only save 1 byte >>>>because the physic position of the record tells which position it is. >>>> >>>>The number of positions is reduced because we can convert a Black-to-move >>>>positions, to a white-to-move position. We can flip position between y-axe, >>>>x-axe or even x/y-axe. >>>> >>>>Michel Langeveld >>> >>>Thanks Michel. >>> >>>It isn't what I was thinking at all. Is it correct to say that once the TB >>>starts to be used that the progam doesn't do any real processing anymore, just a >>>scan of the positions? >>> >>>Steve >> >> No, that is not correct. When the program begins to hit the tablebases, it >>works as hard as before using an alpha-beta search (I am thinking on leaf >>probers). It might work less when the root score is solved as a win. >>José. > >Jose, > >I just posted a message in response to Dave's post, but it looks like it's all >going over my head. Perhaps I should pick up one of the "computer chess" books >that were mentioned on the forum a few days ago and read up on the subject >before making these assumptions. > >Thanks to all for responding and perhaps I'll be able to make more sense of this >later on. > >Steve I think you don't understand Steve's story (which is true) so I try to explain it again in another way. If you have a TB (KPK King + Pawn vs King) and your program will search for a position KPK then it will immediatly find the solution (win/lost in x or draw) from the tablebase. IF you have only this table and try to find calculate a position like KPKP (king + pawn vs King + pawn) then your tablebase will only be used if one pawn is captured in the searchproces. So the tablebase is used but much less. This above is exactly what steve is saying but than in a more technical view.
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.