Author: Peter McKenzie
Date: 11:06:05 12/16/99
Go up one level in this thread
On December 16, 1999 at 10:37:37, Brian Richardson wrote: >A few weeks ago there were some posts about an ICC Winter tournament with >tentative dates in late Jan and early Feb at 75 10, as I recall. > >What is the status of this event? Good question, it seems to have stalled. > >Also, I am hoping to get some suggestions on the "most important" next features >to add to my program, either in general, or to prepare for this tournament. > >My (as yet unnamed--I know boring--I can really relate to Insomniac), program >has the following features: basic TSCP-type search (iterative deepening and >history with full quiesce) with usable subset of Winboard support, non-rotated >bitmaps, simple hashing (only one table, simple draft replacement), null moves, >killers, internal iterative deepening, and some (hopefully significant) eval >changes. Also draws by repitition are supported and rudimentary time controls. >Enormous thanks to Crafty, TSCP, and Winboard authors. > >Using LCT II and BT2630 shows it has improved to about 2200, although I have yet >to really test it on ICC. > >Of course, there is still a long list of things to add (nevermind to tune), >including at least: pawn hashing, pondering, books, and learning. > >So, my question is what would people recommend that I work on next? I was Most important is that you get it onto one of the servers and see how it performs. This will give you a decent clue as to what to work on next. Of the list you give, I'd priortise them as follows (most important first): book, pawn hashing, pondering, learning. Once you play on the server you'll see humans start repeating wins if you don't have a book with some randomness in it. Learning would solve this, but not as well or as easily. Pawn hash is great because it basically gives you pawn structure evaluation for free, meaning that once you have it you can add alot of pawn structure stuff without slowing the program down at all. You can also hash stuff in there that will greatly add the rest of your evaluation like bitmaps of open files etc. Pondering is nice, but also a bit tricky and a good source of bugs :-) Probably best to leave until the rest of the program is more fully tested in real games - same for learning. >thinking pondering would be essential for a tournament (and ICC use), but there >may be other more significant things. Incidentally, I still find bugs like >2 >instead of >>2, or = instead of ==, or unsigned int vs just int playing havoc, >and "little" changes like doing hash probes BEFORE in check testing (saved 30%, >sigh). Do others with more experience still run into these situations ? I've been workong on my program for at least 5 years and I still run into the occasional nasty bug. They tend to be logic errors instead of language specific problems though. Chess programs are a tremendous source of bugs, no doubt about that!
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.