Author: Albert Silver
Date: 06:49:24 08/06/04
I ran a test on Thorsten Czub's Mach10 style for Pro Deo against Junior 8 using
the Nunn set. In the SSDF list, Junior 8 was one of the programs that gave Rebel
12 its worst beating with a score of 24.5-16.5 and as everyone seems to be
testing against either Fritz or Shredder, it seemed like an interesting idea. I
will now run them on the Nunn2 set and see how it goes. The match conditions
were:
Athlon 2400 XP+ (2GHz)
64 Mb hash each, Ponder off, no EGTB, Fritz interface.
Some of the games between the two led to some rather entertaining chess, even in
their draws.
For example, in game 2, they reached this position:
[D]1r6/7p/3pk3/2r1bp2/8/5BP1/P3R1KP/3R4 w - - 0 40
Pro Deo noticed the pin and played 40.Rd4. Junior promptly replied with a
hanging piece of its own with 40...Rb2!
[D]8/7p/3pk3/2r1bp2/3R4/5BP1/Pr2R1KP/8 w - - 0 41
Quite the picture. :-) And 4 moves later, in this particular pas-de-deux, they
reached this position:
[D]8/7R/3p1k2/4bp2/8/5BP1/r1r1R2P/5K2 w - - 0 45
And sure enough, Pro Deo played 45.Rf7+! This game had me really chuckling.
Another interesting aspect was how much the two disagreed on position
evaluations. As an example, in the same aforementioned game, the following
position came up at move 32:
[D]5r1k/7p/1q1p4/2rBbp2/Q3p3/6P1/P3RP1P/3R2K1 w - - 0 33
Here, Pro Deo Mach 10 is happy with its position and says:
33.Qb3 Qa5 34.Red2 Rc3 35.Qb1 Qa7 36.Qb7 Qxb7 37.Bxb7
² (0.26) Depth: 12 00:00:55 52199kN
33.Qb3 Qa5 34.Qb7 Qc7 35.Qxc7 Rxc7 36.Kg2
= (0.21) Depth: 13 00:01:50 107340kN
Note that Pro Deo default is no different:
33.Qb3 Qa5 34.Be6 Rc3 35.Qd5
= (0.23) Depth: 11 00:00:14 10365kN
33.Qb3 Qc7 34.Red2 Bc3 35.Rc2
= (0.25) Depth: 12 00:00:51 45824kN
33.Qb3 Qc7 34.Qb7 Qxb7
= (0.25) Depth: 13 00:02:16 128082kN
However, if one gives the position to Junior 8, it thinks Black is doing better:
33.Rc2 Rxc2 34.Qxc2 e3 35.Qb3 exf2+ 36.Kg2 Qc5 37.Rd2 Bd4 38.Qf3 Be3 39.Re2 Re8
40.Rxf2 Bxf2
³ (-0.42) Depth: 15 00:00:05 8636kN
33.Red2 Rb5 34.Kg2 Rd8 35.Rc2 Rb4 36.Qc6 Qxc6 37.Rxc6 Ra8
³ (-0.27) Depth: 15 00:00:23 41003kN
33.Red2 Rc3 34.Kg2 Qc5 35.Qd7 Rc1 36.a4 Qc7 37.Qxc7 Rxc7
³ (-0.28) Depth: 17 00:01:20 139730kN
And of course, heh-heh, Shredder 7.04 concludes it's about equal:
33.Qb3 Qa7 34.Qe3 f4 35.Qxe4 fxg3 36.hxg3 Bxg3 37.Qe3
= (-0.01) Depth: 12/31 00:01:03 24305kN
33.Qd7 Rxd5 34.Rxd5 Qb1+ 35.Kg2 f4 36.Rxe5 f3+ 37.Kh3 Qf1+ 38.Kh4 dxe5 39.Qe7
Ra8 40.Rxe4 Kg8 41.Rxe5 Qh3+
= (0.00) Depth: 12/37 00:01:13 28384kN
33.Qd7 Rxd5 34.Rxd5 Qb1+ 35.Kg2 f4 36.Rxe5 f3+ 37.Kh3 fxe2 38.Rh5 e3 39.Qe7 Rxf2
40.Qe8+ Kg7 41.Rg5+ Kf6 42.Qd8+ Kf7 43.Qd7+ Kf6 44.Qd8+ Kf7 45.Qd7+ Kf6 46.Qd8+
Kf7
= (0.11) Depth: 12/38 00:01:22 32453kN
33.Qd7 Rxd5 34.Rxd5 Qb1+ 35.Kg2 f4 36.Rxe5 f3+ 37.Kh3 fxe2 38.Rh5 e3 39.Qe7 Rxf2
40.Qe8+ Kg7 41.Rg5+ Kf6 42.Qd8+ Kf7 43.Qd7+ Kf6 44.Qd8+ Kf7 45.Qd7+ Kf6 46.Qd8+
Kf7
= (0.11) Depth: 13/40 00:01:47 42746kN
So much for depending on one engine for an evaluation. The above examples are
limited to 1-2 minutes as this is the time they thought for at most in the
games. Note that I saw a number of instances where the added difference went to
0.80. Meaning one thought it was up 0.40 and the other thought the same about
itself, and this went on for several moves.
Well, now for Nunn2 to see how it goes there.
Albert
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.