Author: Mark Boylan
Date: 07:05:40 03/06/06
Go up one level in this thread
On March 06, 2006 at 09:20:41, Jon Dart wrote: >Looks fairly reasonable. > >I have often thought a SQL database would be useful for chess >programs partly because it makes it easy to insert new moves >in the tree. It would be nice if book learning could include >picking up moves opponents have played. > >But this schema seems to have a mix of info, some >of which might be useful for a human browsing the book and >some of which looks more suited for a program. But it doesn't >look complete for either use. For example, a human would probably >want to have a game reference if there is one. For both uses, >win/loss and frequency is useful. > Yep. There should be a game references. In fact, when I was scrawling my diagram, I first came up with the concept of ComputerPlayer and HumanPlayer thinking that they would be useful for both analyses and games played. But I thought that was too ambitious for a midnight modeling session, so I decided to stay within the scope of Dann's original database. >For programs, a "weight" factor is useful - this can be auto >computed from things like win/loss frequency and adjusted as >a result of book learning (my program keeps learning info in >the book itself - although many programs separate it out. Keeping >it together is more feasible with a SQL DB). > I'm sure there are many details missing. I'm also looking at a couple of relations that I would probably do differently. It definately needs some work. Just a start really. >Also I don't understand why pv in the evaluation is a varchar(8). The only excuse I can make for that is that I did it after working all day and two glasses of wine :) - mark
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.