Author: Shep
Date: 04:21:12 07/29/99
Go up one level in this thread
On July 29, 1999 at 06:23:42, Peter McKenzie wrote: >Please tell us more about this Aemis test. 100 tactical positions in Aemis 1, 100 endgame positions in Aemis 2. Compiled by Covax about 1.5 years ago. Taken from the Informant series, I believe. >Is it well tested? Some people claim that a test suite where the top 5 programs are only a few points apart (117-126 of 200) is not a good one. However, the Aemis suite is very diverse. By this I mean that while the total number of solved positions does not differ much, the positions found by each program differ quite a lot. (Imagine a 100 positions suite where program A solves 1-50 and program B solves 51-100, this would be interesting, right?) The diversity index (number of positions solved by best program divided by number of positions solved by all programs together) is 126/182, thus very low compared to the 28/30 (or so) in Louguet II. (Lower=more interesting) There are some positions where I have doubts the given move is the only solution, though. >Is there any further information about it on the web? You can get the whole suite in PGN and CBV (EPD follows soon) plus results on SCCS: http://sccs.cjb.net -> Test suites -> Aemis test suite --- 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.