Author: Jim Monaghan
Date: 20:11:28 03/26/03
Go up one level in this thread
On March 26, 2003 at 22:27:01, Mike Siler wrote:
>does anyone know (or know where I can find) the formula for estimating a
>program's playing strength based on the BS-2830 test suite?
From Rebel Documentation:
The new BS2830 (Bednorz-Schumacher) test suite can be downloaded from the Rebel
10 subscription aera. This suite is designed for programs, with an expected
SSDF-ELO rating of 2300 or higher. The test suite contain 27 positions. Select
each position and let Rebel think for 15 minutes. If a position is solved, write
down its solution time in minutes. It doesn't count as a solution if Rebel
finds, the move and then changes its mind. If after finding a move, then
changing its mind, then finding it again, you should use the last time found.
Any solution that is not found, score as 15 minutes. Add up all the times (in
minutes) to the total time TT (in minutes). You get the (BS-) ELO rating with
the following formula (updated in CSS 2/1998):
BS2830-ELO = 2830 - TT/1.5 - (TT * TT)/(22 * 22)
Be aware, that the authors changed the formular in April 1998. The older
formular (CSS 4/97) was
BS2830-ELO = 2830 - (TT * TT)/(17 * 17)
So all older results must be recalculated.
Be cautious in interpreting the BT/BS results. The BT/BS values are only
indicators for the tactical strenght of a chess engine. You can't measure the
positional strength of a chess engine with this test suites.
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.