Author: Ryan B.
Date: 03:13:52 11/27/05
Go up one level in this thread
On November 27, 2005 at 06:00:46, Peter Skinner wrote: >On November 27, 2005 at 04:48:37, Ryan B. wrote: > >>Yes a bitbase standard will very soon be public. Bitbases are Superior to EGTB >>because they probe faster and work in qsearch. The .dll file to use Danial's >>bitbases is already done and circulating. It will just take some time for the 5 >>piece bitbases to be generated. I am open to the possibility that the use of >>both bitbases and EGTB together could be usefull but this would require a large >>amount of space for the 6 piece TB's for a small gain. > >Then here lies the a problem. > >If the access code to use the new bitbases is in the form of a dll, how do linux >program authors use it? > >Would it not be better to release the code as a c or cpp module? Or in some form >that is cross platform? > I think it would not be hard to make a .so file for linux and mac users. I am not sure is we can expect this but it would be nice for linux and mac users. >More and more engines and programs are being cross ported, why limit one's >choices to a single operating system? > >Purely just thinking off the top of my head... I don't want to argue which >operating system is better, but rather the productive choices given to program >authors. > >>Scale down towards draw per pawn ram when pawns >= 14 or no open files. This >>will motivate computer to open the game up. Other scale down factors are >>obviously need as well. > >I have thought of this as well. > >Would this not require a vast amount of knowledge outside the opening book of >opening theory and when it is fine to have an open position, or whether it is >better to keep it closed? > >I am assuming as well you would have to make this code identify whether it was >playing another program or a human, and what positions for either side would be >open/closed. > >If such knowledge is needed, it would noticibly slow down the engine and may >slow it to the point of being counter-productive. No? > >Taking the example of the latest Fritz, if you believe the advertsing/author >comments that he tuned the program to play regardless of the type of opponent, >and the strength improvement over the previous version, would you not tune your >program to play the same way? > >Let's try Crafty for an example. > >If you tell Crafty that it is playing a computer (or an interface tells Crafty), >it plays less random and choses it's book moves with less width. If you also use >it's bookc.bin and tell it is playing a computer, things become even more >restrictive. I have always felt that _not_ telling Crafty it is playing a >computer (and modifying winboard not to tell it in ICS mode), it plays a wider >selection of openings, and frankly plays better. (In my opinion) > >If we could do that to all engines, would all show the same improvement like >Fritz? Or would some by severely hurt by it? > >Deciding whether to have an open file or position against x or y opponent is the >exact slippery slope that I am referring about. > >Peter
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.