Author: Eric Oldre
Date: 18:30:29 12/23/05
Go up one level in this thread
On December 23, 2005 at 20:15:09, GeoffW wrote: >Hi Eric > > >>Latista does not allow the hash table size to be configured. That is probably >>something that I should have considered. When the bitbases are loaded in memory >>it uses 50 mb, and without them less. So I thought that is probably small >>enought for most people. >I was trying to increase it :-) > >>If you have a tournament you want to enter it in and have a specific size, I >>could do another point release in the near future, or send you a special >>version. >No its ok thanks, just playing a few games for fun, not a serious tourney > >>Logging is also not configurable. You are right that it really should be though, >>I will make sure it can be disabled in the next release. > >Yes that should be done really >I must have played some games with your old version when I put the new version >into the arena folder it was already over 80 Megs in size !! > >> >>About the bitbase dll. Daniel Shawul created a modified version of TSCP that >>interfaces with the bitbase dll. It really did not take very much extra code to >>get it working. (thanks again daniel). I simply saw what daniel did in TSCP and >>did almost the same thing in Latista. > >Oh thats interesting, should be easy for me to do also then as my prog is just a >generously hacked TSCP > >>two things to be careful of when programming with the bitbases. 1) the dll >>considers zero to be A1, and 63 to be H8. but the original tscp source used A8 >>for zero and H1 for 63. (like latista does). > >thanks for the tip, I wonder where the modified TSCP is, I will go looking for a >link when I have posted this, or do you know what the link is please ? > >Just to finish here is an interesting game for you, nmight help you improve >Latista. It was +4.3 up and a winning position of 3 pawns up and still managed >to lose. I havent studied the game so dont know where it went wrong > >Sorry for the long pgn, but I had some debug output enabled in my prog > > Regards Geoff > Geoff, Thanks for the PGN. that is one of the more "educating" games I've seen. I've done some analysis and found some blunders. Things tend to "go pear shaped" for latista more often in the early end game than in any other part of the game. Like when there are 2 minors and a few pawns left on each side. I'm not sure if it's: a) bad evaluation in that part of the game b) bugs in the search c) too conservative about when null-move is safe to do. and it gets out-searched that part of the game. Anyway, about the endgame bitbases, I would email daniel about getting his version of TSCP. Since he created it I'm not sure I should redistribute. Also there were a few bugs in the version he gave me which he may have fixed by now. Which will save you some BIG headaches. If you see more games like this from Latista feel free to email me the PGN. Games like yours are VERY helpful to me. Eric
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.