Author: Mike S.
Date: 07:46:10 06/23/02
Go up one level in this thread
On June 23, 2002 at 10:08:29, Robert Henry Durrett wrote: >I would be interested in knowing the names of people who make these. > >The main interest is in publicly available suites. One might reasonably surmise >that engine designers/developers have their own "secret" suites as well. > >Also, which currently available suites are the very best? [Maybe they should >get ratings too?] I'm afraid it's only a small minority of users who is really interested in using test suites. I mean people who test every new engine, calclulate or even publish results, etc. Even fewer people actually design test suites. A good overview of popular computer test suites: http://www.andreas-schwartmann.de/chess.html#Test Some test suites which were/are popular in german speaking regions, were designed by H.Bednorz and Tönissen (BT-Tests), Bednorz & H.J.Schumacher (BS-Tests), and recently by Schumacher & M.Gurevich, supported by Manfred Meiler who also posts here and is very helpful always. http://www.computerschach.de/test/index.htm The Quicktest (which I trust most, for obvious reasons :o)): http://meineseite.i-one.at/PermanentBrain/quick/quick1.htm To judge the quality of test suites, the first idea would be to do intensive analysis to see, if (and how many) positions are bogus, or at least doubtful. But it may be very difficult to judge. Furtheremore, as mentioned in a previous thread, a strong and correct move/idea/variation does not yet make a good test position IMO, if the testing character of the position is missing. But this is also difficult to see sometimes. For example, in this position #7 from the WM-Test, I thought the testing character would be missing. I was wrong: [D]r4k1r/pp2pp1p/8/2PPb3/Q7/4p3/B2q1PPP/2R2RK1 w - - 0 1 The game (Petrosian-Fischer 9171) went 1.Rcd1 Qe2 2.d6 Qh5 3.f4 e2 4.fxe5 exd1Q 5.Rxd1 Qxe5 6.Rf1 f6 7.Qb3 1-0 I asked, isn't 1.Rcd1 a simple attacking move (only), which could be played without seeing anything of the following variation? But the move offers a sacrifice of the exchange: 1.Rcd1 e2 2.Rxd2 Bxh2+! 3.Kxh2 exf1Q - Surprising, but easily visible for programs. So they won't play 1.Rcd1 without additional reason. Which means it takes much time & chess strength to try to "debug" large, difficult test suites. Also I'm afraid you can't rely on automatic analysis for that purpose, because i.e. no engine will ever point you to the above variation except you enter 1...e2 *manually*. Regards, M.Scheidl
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.