Author: Gian-Carlo Pascutto
Date: 09:58:06 09/12/03
Go up one level in this thread
On September 12, 2003 at 12:36:08, Bob Durrett wrote: >Actually, GCP, you and I are talking past each other. You are talking about >apples and I am talking about oranges. We are each discussing a different >topic. I don't think so at all. >The SSDF tests are timed tests. The engines are compared based on whatever >moves they can come up with in a given amount of time. > >What do you suppose the results would be if each engine were allowed to reach >some preset search depth [>16] before each move? In this proposed >[hypothetical] competition, the engines would be forced to move as soon as they >reached the required search depth but not allowed to move prior to reaching >that depth. [As a refinement, the pre-determined depth might be set to be >greater in endgames, perhaps >18.] > >In such a competition, I suspect that Crafty would come out on top of the >rating list simply because Hyatt says Crafty does very little pruning. [This >assumes that the Crafty operator does not shut the engine off due to >boredom. : )] I would presume Crafty to come out at the very bottom because it has an inferior evaluation, inferior extensions, and the little pruning that it does do is not very safe (compared to the top professionals - I don't want to be rude to Bob but let's not deny the facts). I don't see the value of such a comparison or why you bring it up either - it is completely useless and besides the point. >Also, note my comment regarding whether or not any kind of "bias" would affect >the selection of the move. Consider two hypothetical engines identical except >that one engine's position evaluation software gives position evaluations >exactly double those of the other. Would you expect one engine to perform >better than the other insofar as selection of the move to play is concerned? I >would not. The user, however, cares about the accuracy of the evaluation >score. Double the correct value is no good for the user. Junior 7 had a bug in this regard. (Always divided scores by half). >My concern was with what the GUI displays to the user. What happens inside the >guts of the engine is for engine programmers to consider. I am a user and you >are a chess programmer. ["I am a pear and you are a peach."] What you see is exactly what the engine works with in this case. My comment still stands. If Shredder 7.04 gives scores that disagree with other engines, I would not conclude they must necessarily be wrong, because Shredder 7.04 is an extremely strong engine (hence the SSDF remark). I would rather conclude that the others are more likely to be wrong. -- GCP
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.