Author: Dann Corbit
Date: 23:41:52 07/25/03
Go up one level in this thread
On July 26, 2003 at 00:55:19, Dana Turnmire wrote: >"Some people say that they can watch a single move by a program and judge how >strong it is. I'm not one of those people." > >Maybe not a single move but how about a much weaker program such as an old >dedicated unit playing against one of the top programs on a current PC. Would >one really have to play 100 games or even 20 games to know which program really >is the strongest? Where can the line be drawn? Mathematics can draw the line for you. After "k" number of games, you can have a certainty of some number. If the number is good enough for you then stop there. With the SSDF list (for instance) Rating + - Games Won Av.opp 1 Shredder 7.04 UCI 256MB Athlon 1200 MHz 2810 37 -34 465 76% 2607 2 Shredder 7.0 256MB Athlon 1200 MHz 2770 27 -25 801 70% 2622 3 Fritz 8.0 256MB Athlon 1200 MHz 2762 26 -25 821 68% 2627 4 Deep Fritz 7.0 256MB Athlon 1200 MHz 2761 28 -27 694 69% 2622 5 Fritz 7.0 256MB Athlon 1200 MHz 2742 30 -29 574 64% 2637 6 Shredder 6.0 Pad UCI 256MB Athlon 1200 2724 23 -22 991 63% 2634 7 Shredder 6.0 256MB Athlon 1200 MHz 2721 31 -30 547 62% 2632 8 Chess Tiger 15.0 256MB Athlon 1200 MHz 2720 26 -25 784 61% 2641 9 Shredder 7.0 UCI 128MB K6-2 450 MHz 2717 46 -43 258 63% 2624 9 Chess Tiger 14.0 CB 256MB Athlon 1200 2717 30 -30 557 61% 2638 11 Deep Fritz 256MB Athlon 1200 MHz 2715 30 -29 571 61% 2639 12 Gambit Tiger 2.0 256MB Athlon 1200 2712 29 -29 583 58% 2653 13 Junior 7.0 256MB Athlon 1200 MHz 2697 25 -25 761 55% 2663 14 Hiarcs 8.0 256MB Athlon 1200 MHz 2682 23 -23 952 53% 2659 15 Rebel Century 4.0 256MB Athlon 1200 MHz 2675 29 -29 590 60% 2604 We know that Shredder 7.04 is stronger than Rebel Century 4, since 2675+29=2704 and 2810-34=2776. Hence, we can say with 97% accuracy that Shredder 7.04 is probably stronger than Rebel Centrury 4 (on that particular hardware and under those particular conditions). However, we cannot know if Shredder 7.04 is really stronger than Chess Tiger 15. The error bars overlap, and so we are left without a decision. You are (of course) free to guess which one is stronger.
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.