Author: Steve Glanzfeld
Date: 16:13:58 09/12/05
Go up one level in this thread
On September 09, 2005 at 23:48:47, Tanya Deborah wrote: >Speaking about the strengh of the programs, The SSDF results are much more real >than any other tournament in the world (equal hardware + Much,much more games, >etc). We have to wait and see if Zappa can play 400-500 games, and has a better >performance than Shredder 9, also if Zappa can beat the strongest program >Shredder 9 of the first place, to say that Zappa is the clear #1 in the world. What is the definition of "real"? (I.) If you want to know which engine is strongest against other engines at 40/2h, 1200 MHz, own books and with two comps/ponder on, wait for SSDF results. (II.) If you want to know which engine is strongest at blitz, maybe with own books or maybe with equalized openings, various other CPUs and ponder off at a single CPU computer - like probably most users have only - search other rankings and results. (III.) If you want to know which engine is best for analysis, take a look how it performs in analysis. That will be mainly at testsuits where we can have other engine's results to compare with, but also single "critical" positions like posted in CCC often. (IV.) If you want to know which engine is strongest against human masters, forget it, you'll never know. Man versus Machine competition does never produce statistcally relevant numbers of games for a particular engine version. The list is incomplete.. I.e. for online computerchess competition (or for 90% of it) performance at very short time controls, typical is 3+0, and ponder = on is what matters. OTOH a correspondence player wants strong analysis performance at long thinking time probably, and the program's opening book is irrelevant to him. What is the chess software user profile of a typical club player, who wants to analyse his games (somewhat below IM level :) ) and practise a bit? What is the chess software user profile of a typical engine fan, who wants to match the strongest engines on his machine against the latest releases and particpates in online competition? These are all different setups (time/cpu(s)/book/ponder..). If someone wants to base a decision what to buy on rating lists, it makes sense to look at those ratings which come from a test setup most similar to his own practise of how he uses an engine, most often. That includes test position sets, because running a testsuite of selected positions is much more comparable to the analysis situation, than playing a practical game from the opening to the endgame when 90+ % of the positions are so boring that you would never do engine analysis at it (in a normal interactive analysis of your latest club game or the like). Steve
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.