Author: James Kukula
Date: 22:27:26 01/06/05
Go up one level in this thread
Here is a really simple way to see that some kind of "computer aided explanation" is possible. Suppose one actually had a small collection chess engines. At the beginning of a play, one simply ran every engine in the collection. So then one would have a list that gave next to the name of the engine the move that the engine recommended. Maybe every engine recommends the same move! In which case the choice is easy. But generally I expect the different engines will come up with a few moves. One could assign each engine a voting weight, and then sort the moves by the total weight of all the engines that voted for that move. Still, one would see next to each move the list of engines that voted for it. The name of an engine is a peculiar sort of "explanation" to provide for a move, but I bet the folks here are very familiar with the quirks of the most popular engine, so if the list says that "crafty" votes for some move, that would be a meaningful explanation. Here's the next level of interaction that I envision. Perhaps one actually has quite a large collection of engines, so running the whole set would soak up too much cpu resource for a given move. To start a move, one could pick some subset of the engines to run. I imagine folks get a feel for engines, for knowing in certain types of situations which engines are most likely to give the best moves. Maybe one would run the engines sequentially, picking the next engine to run after seeing the moves picked by the engines already run. If all the engines so far have picked the same move, maybe some radically different engine will pick a different move - if not, one really can conclude the single move picked is the best way to go. Does this direction seem like it ought to work, or what am I missing? Thanks, Jim
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.