Author: William Penn
Date: 04:51:04 12/03/05
Fruit 2.2.1, Toga II 1.1, and Shredder 9.1 UCI analyze a position Shredder Classic 9.1 GUI; multi-PV mode, 3 lines; let each engine run for the times indicated; analyses are concatenated; AMD Athlon XP 2400+ 2GHz, 1G RAM [D]2r5/4bprk/pB2b1p1/P2pP2p/3N1R1P/q2PQ3/6P1/5RK1 b - - Engine: Fruit 2.2.1 (704 MB) by Fabien Letouzey - 105 minutes 20 54:35 -0.10 33...Rc3 34.Nxe6 fxe6 35.Qh3 Rc6 (1.928.851.276) 591 20 94:55 -0.12 33...Kg8 34.Nxe6 fxe6 35.Kh2 Kh7 (3.367.131.355) 591 20 78:19 -0.13 33...Re8 34.Qf3 Qc3 35.Qg3 Rc8 (2.758.128.282) 591 Engine: Toga II 1.1 (704 MB) by Thomas Gaksch and Fabien Letouzey - 140 minutes 20 121:00 -0.12 33...Rc3 34.Nxe6 fxe6 35.Qh3 Rc6 (4.468.526.882) 615 19 44:47 -0.23 33...Kg8 34.Nxe6 fxe6 35.Rf6 Bxf6 (1.710.805.862) 615 19 115:42 -0.24 33...Re8 34.Nxe6 fxe6 35.Qf3 Rc8 (4.283.119.413) 615 Engine: Shredder 9.1 UCI (704 MB) - 287 minutes 22 179:10 -0.26 33...Rc3 34.Nxe6 fxe6 35.Qf2 Rc1 (3.833.918.156) 356 22 207:32 -0.53 33...Kg8 34.Nxe6 fxe6 35.Qe2 Qc3 (4.471.816.313) 359 22 225:21 -0.85 33...Re8 34.Nxe6 fxe6 35.Qe2 Rc8 (4.867.391.722) 359 I like this position for a sample analysis. It's difficult. There is no clear "best move", so no exact solution to be discovered as far as I know. It is representative of what happens in chess much of the time. Neither humans nor chess engines know for sure what is the best move! These three analyses were run for somewhat different times, for no particular reason, but because that was most convenient. They show what I want to show, that these engines produce very similar results in lengthy analyses. They all agree that 33...Rc3 is the first choice, but there's very little difference from the other candidate moves, so the confidence level is not high. Any of these three candidate moves might be best after more extensive evaluation, or perhaps even a different move. I have used Shredder 9 for a long time, since it first came out in Feb'05. When I switched to Fruit 2.2.1 for analysis recently, I noticed their similarity right away. Watching the analyses in progress makes me think they are using similar algorithms. And the evaluation numbers are similar, although a little different in magnitude. Shredder usually has larger numbers, but the candidate moves and their relative evaluations are very close. If I didn't know differently, I might think Fruit is a clone of Shredder!? Sometimes they produce different candidate moves but usually they are the same. What is the point? Nothing in particular, except to characterize what happens when you use different engines trying to discover the truth in a difficult chess position. A lot of the time you don't know what is best, or which engine will produce the best move, and they don't know what is best either! It's like the blind leading the blind. I guess the only point is that chess is still largely a game of blind man's bluff... :) WP
This page took 0.01 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.