Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is Your program falling in these traps (12 positions)?

Author: Robert Hyatt

Date: 10:05:18 09/23/98

Go up one level in this thread


On September 23, 1998 at 01:18:29, Jouni Uski wrote:

>Here are 12 positions from old CSS magazine (my own permission):
>
>2Q5/5Rpk/8/1p2p1qp/1P2Pn1P/5PP1/4r3/7K b - -
>r1b1r1k1/1p1npp1p/2ppnqp1/4b3/p1P1P3/5P2/PPNQNBPP/1R2RBK1 b - -
>8/8/8/N1K5/P7/1Pb5/2k5/8 b - -
>7Q/kp3ppp/8/8/q3p3/P5PP/5b2/R1R3K1 w - -
>6k1/6p1/4q2p/5b2/1p1R1Q2/pP6/P3rPPP/K2R4 w - -
>6k1/pp1b2r1/1q1p1pRQ/4p3/4P3/2r1PP2/PP6/1K6 b - -
>4r1k1/3b1p2/5qp1/1pPpn2p/7n/r2BP1N1/2QBRPPP/1R3NK1 w - -
>3r2k1/p4ppp/1q6/1Nrn4/8/2p1P3/P1R1QPPP/2R3K1 w - -
>1r6/5pk1/2p1pbpq/3p3p/2P4P/1PNP1QP1/1B2PP2/6K1 b - -
>r4r2/ppp2p1k/3p1Rpp/3P4/1Q1Nq3/3N4/PPP3PP/6K1 b - -
>r4rk1/5p1p/p2RbnpQ/q1p1p2P/1pP1P3/1P3P2/P1N3P1/2K2B1R b - -
>8/7p/8/p4p2/5K2/Bpk3P1/4P2P/8 w - -
>
>The task is to AVOID these moves:
>-Qxg3?
>-Bxb2?
>-Bxa5?
> Kxf2?
> Rxb4?
>-Rxg6?
> Bxb5?
> Nxc3?
>-Rxb3?
>-Qxd5?
>-Nxh5?
> Kxf5?
>
>For comparison here some results in P90 and 3 minute time limit:
>
>                 1 2 3 4 5 6 7 8 9 10 11 12  total solved
>
>Nimzo98 (Pad.)   - - - - X X X X X -  -  -     5
>Mchess 7         X X X - X X X X X X  -  X    10
>
>Jouni

Here's my results, but presented just a little bit differently.
I give 3 columns...  the depth at which it switches away from the
bad move, the time when it fails low (realizes the bad move is bad)
and the time when it actually announces a new best move.  The fail low
time is interesting, because if it fails low at exactly 3 minutes, this
means that *if* it has at least 3 minutes to choose a move, it won't play
the bad move, because the fail low will trigger a deeper think to find a
better move.

position     depth     Tfail     Tnew
   1.          10        4         7    Qxg3 = 0.00, Qg6=+4.4
   2.          10       18       121    Bxb2 = +.70, Ndc5=+.94
   3.          11        3         8    Bxa5=-6.41
   4.          --        -         -    finished 12, still +8.00
   5.           8        4         9    Rxb4 drops from +3.5 to +2.0, R4d2
                                        is +3.09
   6.          11       27        37    older versions find this at
                                        depth=3 where Rg6 drops to about
                                        +5.5 while Rc1 is +5.6.  Rg6 drops
                                        to 0.00 here at depth=11, 27 seconds.
   7.           8        1         7    Bxb5 -Mat05
   8.           6        -         0    likes a4 positionally, instantly,
                                        never considers Nxc3 after 5 plies,
                                        0 seconds.  Never considers Nc3 after
                                        that point.
   9.           7        0         2    Rxb3 -3.04
  10.          --        -         -    Score =-.7 thru depth 11.
  11.           2        -         0    Likes Qxa2 from ply=2 on.
  12.           4        0         0    likes g4

all times are on my P6/200 X 4 machine...



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.