Author: Jim Bond
Date: 00:06:25 05/04/03
Go up one level in this thread
On May 04, 2003 at 02:33:35, Uri Blass wrote: >On May 04, 2003 at 02:10:17, Jim Bond wrote: > >>On May 04, 2003 at 01:07:52, Uri Blass wrote: >> >>>On May 03, 2003 at 23:47:13, Jim Bond wrote: >>> >>>>On May 03, 2003 at 23:22:05, Tom Kerrigan wrote: >>>> >>>>>On May 03, 2003 at 21:28:30, Jim Bond wrote: >>>>> >>>>>>On May 03, 2003 at 17:52:40, Uri Blass wrote: >>>>>> >>>>>>>On May 03, 2003 at 17:50:36, Uri Blass wrote: >>>>>>> >>>>>>><snipped> >>>>>>>>I know about programs that tablebases were counter productive for mchess because >>>>>>>>it probed them too much and was slowed down by asignificant factor. >>>>>>> >>>>>>>I meant here that mchess is an example for a program that tablebases was counter >>>>>>>productive for it. >>>>>>> >>>>>>>Uri >>>>>> >>>>>>If you believe that more TB probing is counter productive, you are contradicting >>>>>>with Shredder - the top program. Take this position for example: >>>>>> >>>>>>2k5/8/7p/8/5qP1/1Q5K/8/8 w - - 0 72 >>>>>> >>>>>>If you run infinite analysis on it with Shredder 7.04, Fritz 0.008 and >>>>>>ChessTiger 15, you will find that the Shredder accumlates TB counts about 8 >>>>>>times more than Fritz and about 16 times more than ChessTiger. If TB probing is >>>>>>counter productive, how come Shredder does it so much more and can still be at >>>>>>the top? >>>>>> >>>>>>Jim >>>>> >>>>>Just because Shredder is at the top doesn't mean it plays this particular >>>>>position better than other programs, i.e., the fact that it does so many more >>>>>probes does not mean that more probes are good. >>>>> >>>>>-Tom >>>> >>>>You are saying more probes does not mean that more probes are good, but you >>>>cannot prove that more probe is bad either. The fact is Shredder does probes >>>>and it is the top program where others does less probes and are less strong. >>>>There is a correlation here wouldn't you agreed? >>>> >>>>Jim >>> >>>If I see A and I see B then it does not mean that A is the reason of B. >>> >>> >>>Shredder can be better than the opponents because of different reasons. >>>I believe that shredder7.04 is better than the opponents also when the programs >>>do not get tablebases. >>> >>>Uri >> >>I am saying "A" can be a likely possible reason for "B" given the observations. >>For example, if I pour water and the floor is wet, you would say the floor is >>wet possibly because I poured water right? If the sun came out the east one day >>and sun also came out the east on the second day, you would say the sun most >>likely is going come out the east on the third day right? Can anyone know >>really the truth? Why do you do tree search in movei anyway? Because you don't >>know the truth right? >> >>Jim > >I do not see a proof that suggest that shredder is better in endgames thanks to >better use of tablebases relative to the opponents. > You do not see a proof but is there a proof? > >Fritz and Tiger and other programs also use tablebases and my opinion is that >tablebases do not give more than 20 elo relative to no tablebases. > Is there a proof for that? >More probes does not mean better use of tablebases and it can be also a result >of lack of knowledge. > Not implementing knowledge might simplify the code. This could be a trade off between accuracy and complexity. >The only way to check if shredder does better use of tablebases is by comparing >the level of programs with tablebases and without taablebases. > Is there really an official test for that? Who's is going to do it? >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.