Author: Shep
Date: 04:17:21 11/22/99
Go up one level in this thread
On November 21, 1999 at 11:12:55, odell hall wrote: >that Fritz searches. Without being to "technical" how is it possible for Chess >tiger to outplay and outsearch Fritz if it does not search as many nodes? I Because it's not important *how many* nodes you search but if you search the correct nodes. If you waste time on the "wrong" ones, you end up seeing less than a slower program (remember how amazingly strong Hiarcs and MChess are at tactics, given their slow speed). have >noticed in many positions using Fritz to analyze the games of Tiger, that Fritz >will completely miss a move that tiger made, even after thinking five minutes. >Yet when I enter the move in fritz it's score will skyrocket indicating it >thought the move very good. I soppose that your program is proving once and That's typical Fritz behaviour, a result of its root node evaluation principle (piece-square tables IIRC!?). You can see this more often that you'd like, e.g. evaluating a position as +0,00 in 5 minutes, then jump to -2 immediately after making the predicted move... for >all that nodes per second is not the most important thing in chess strength. Is >Tiger 12.0 Faster or slower than your previous versions? The actual nps count is slightly slower than e.g. 11.5 or 11.7. However, Christophe made some speed optimizations since then and invested the gained speed again in extending the eval function. Notably, the Rebel Tiger version is about 20-25% faster than the DOS version (at least on many of the systems it has been tested on) due to a better compiler (the actual engine is the same). --- Shep
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.