Author: Jay Scott
Date: 14:57:39 02/20/04
Go up one level in this thread
On February 20, 2004 at 14:36:28, Jaime Benito de Valle Ruiz wrote: >If I (or anyone else) provide you with a list of positions, would you be >interesting in providing the static values that your engine get for each >position? Good idea. One source of positions is well-known opening positions with established evaluations. We could choose some by looking at openings played in recent computer competitions. A sufficiently smart evaluator (if it exists) should understand that the King's Gambit Accepted position after 1. e4 e5 2. f4 exf4 is OK for white. I'd like to see that. Interesting subsets: - positions with material imbalance - positions with an established positional advantage to one side (e.g. more space) - dynamically equal positions - positions with known tactical wins (does the evaluator "smell" them coming?) - positions with known endgame wins or draws (does the evaluator have the endgame knowledge?) - fortress draws (search normally can't recognize them, so the evaluator will have to if anything does) We can expect evaluators to be somewhat "noisy", since search multiplies their effectiveness so greatly. Therefore, to get ideas about the quality of the evaluator we'll need a large supply of test positions. Fortunately, running them should be plenty fast. Jay
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.