Author: Robert Hyatt
Date: 17:09:27 09/10/02
Go up one level in this thread
On September 10, 2002 at 18:42:18, Uri Blass wrote: >On September 10, 2002 at 18:21:10, Robert Hyatt wrote: > >>On September 10, 2002 at 15:53:23, Uri Blass wrote: >> >>>On September 10, 2002 at 12:49:43, Rolf Tueschen wrote: >>> >>>>On September 10, 2002 at 09:25:57, Uri Blass wrote: >>>> >>>>>On September 10, 2002 at 09:18:50, Rolf Tueschen wrote: >>>>> >>>>>>On September 10, 2002 at 09:06:27, Uri Blass wrote: >>>>>> >>>>>>>The value of the endgame tablebases in chess is also small and programs probably >>>>>>>get no more than 20 elo from them. >>>>>>> >>>>>> >>>>>>Researched against human chessplayers? Or from your experience in clean comp vs. >>>>>>comp practice? >>>>>> >>>>>>Rolf Tueschen >>>>> >>>>>I do not think that there is a big difference. >>>> >>>>I don't know. Nut let's see. >>>> >>>> >>>> >>>>> >>>>>based on comp-comp games tablebases changed the results only in minority of >>>>>the cases. >>>> >>>> >>>>Ok, at least with both sides having tables... or at least input about the >>>>importance of certain endgames. See below. >>> >>>I remember that there was tests when only one programs used tablebases. >>>There are also commercial programs that do not use tablebases(I am not sure if >>>the last version of chessmaster or Rebel but clearly previous versions of them >>>and Hiarcs7 that does not use tablebases is at the same level as hiarcs7.32 that >>>is using tablebases). >>>> >>>> >>>>> >>>>>Humans usually do not get simple endgames against computers. >>>> >>>> >>>>This might be true but what does it mean? I'm referring to the fact that the >>>>leading programs use a feature for the early detection of certain endgame >>>>possibilities. So progs do not only rely on the tables themselves but also on >>>>early help for judging much later and only "possible" endgames. At least I >>>>understood it this way. Now my point is, that if the tables are not allowed, the >>>>mentioned control is impossible what increases the possibilities to make basic >>>>mistakes in the middle game. Seems trivial enough. >>> >>> >>>I know but in most cases the games are decided before tablebases are relevant >>>Tablebases are usually not relevant in the middlegame because there is no >>>logical line in the tree of the programs that leads to position with 5 or 6 >>>pieces(if there are few tablebases hits because of some illogical lines in the >>>search tree they usually change nothing and may only do the program 0.01% >>>faster). >>> >>>Even in games that tablebases are relevant programs in most cases do not blunder >>>in the relevant positions even without tablebases thanks to search. >>> >>>Uri >> >> >>I wouldn't go quite that far. I've seen more than one program, commercial >>or amateur doesn't matter, enter a dead lost KRP KR ending, and doing so at >>_their_ option, not because it was forced... >> >>Ditto for KNN vs KP... > >I said in most cases and I talk only about good programs that at least at the >level of the top amateurs. > >No doubt that there are cases when tablebases help but I do not think that they >increase the rating of programs by more than 20 elo. > >Uri I was talking (in one case) about a commercial program in the top-3 on the current SSDF list. :) It simply traded into a dead lost KRP vs KR ending. Against my program. On ICC. It wasn't using tables while mine was, and mine was saying things were pretty equal (mine was a pawn up but it was somewhat drawish). The opponent just traded the last pawn and piece for a pawn and piece (equal trade) and ended up in a dead lost KRP vs KR ending. I've also seen this (or other) programs trade from a winnable KRBPP vs KRPP into a dead drawn KRB vs KR ending and be happy while doing so. So the top programs are not always that "smart" and the tables can partially fill in for that lack... not completely of course.
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.