Author: Albert Silver
Date: 06:44:46 08/18/04
I've been having fun playing around with the settings, but have noticed some
really strange things happening in the automated matches played in the Fritz
GUI.
I looked over the games, enjoyed Pro Deo's victories, and checked the losses of
course. In one game, I saw Pro Deo had a huge edge, according to both it and
Junior 8, yet lost. I was intrigued. Had they both seriously misevaluated the
position or did PD blunder horribly?
Example:
[D]4r1k1/1p4pp/p7/2QNpb2/4q3/1P6/1PP5/1K1R4 b - - 0 31
The game is hardly over, but there isn't an engine that doesn't think Black has
a huge edge. Adn why not? Black *has* a huge edge. Black didn't blunder here,
but already some odd things appear. Pro Deo played 31...b5 here (with -1.91 for
black and 12 plies depth, and Junior expected 31...Qe2. So I run Pro Deo, with
my settings, and see its analysis.
Analysis by Pro Deo 1.0:
31...Qe2
-+ (-1.67) Depth: 1 00:00:00
31...Qg2
-+ (-1.76) Depth: 1 00:00:00
31...Qg2 32.Qc7
-+ (-1.64) Depth: 2 00:00:00
31...Qg2 32.Ne3 Qf3
-+ (-1.66) Depth: 3 00:00:00
31...Qg2 32.Ne3 Qf3 33.Rg1
-+ (-1.60) Depth: 4 00:00:00
31...Bg4
-+ (-1.60) Depth: 4 00:00:00
31...Bg4 32.Qg1 Rf8 33.Rd3
-+ (-1.80) Depth: 4 00:00:00
31...Bg4 32.Rg1 Kh8 33.Ne3 Bf3
-+ (-1.55) Depth: 5 00:00:00
31...Kh8 32.Rf1 Rd8 33.Ne3 Be6 34.Rf8+
-+ (-1.73) Depth: 5 00:00:00
31...Kh8 32.Ne3 b5 33.Rg1 Qf4 34.Qc7
-+ (-1.68) Depth: 6 00:00:00 78kN
31...Kh8 32.Ne3 Be6 33.Qb6 Qc6 34.Qb4
-+ (-1.68) Depth: 7 00:00:01 220kN
31...Kh8 32.Ne3 Be6 33.Qb6 Qc6 34.Qb4 Qf3 35.Qb6
-+ (-1.70) Depth: 8 00:00:01 657kN
31...Kh8 32.Ne3 Be6 33.Qb6 Qc6 34.Qb4 b5 35.Qh4 Rf8
-+ (-1.73) Depth: 9 00:00:02 1509kN
31...Kh8 32.Ne3 b5 33.Qe7 Bg6 34.Rf1 Kg8 35.Qg5 Qd4 36.Ka1
-+ (-1.69) Depth: 10 00:00:04 4017kN
31...Qe2 32.Rg1 Kh8 33.Qc7 Rg8 34.Nb6 Qd2 35.Rh1
-+ (-1.73) Depth: 10 00:00:06 5415kN
31...Qe2 32.Rg1 Kh8 33.Qc7 Bg6 34.Rc1
-+ (-1.78) Depth: 11 00:00:12 11377kN
31...Qe2 32.Rc1 Kh8 33.Ne7
-+ (-1.68) Depth: 12 00:00:23 22071kN
31...Kh8 32.Ne7 h6 33.Rc1 Bh7 34.Qd6
-+ (-1.71) Depth: 12 00:00:35 34791kN
31...Kh8 32.Ne3 Be6 33.Qb6 g6
-+ (-1.73) Depth: 13 00:01:41 97962kN
No 31...b5 in sight, and mind you this is single CPU with no permanent brain.
Hmmmm!
I pinpoint the losing blunder, squandering the draw to clear loss, and no, I
don't have any odd contempt values on:
[D]4r2k/4N1pp/p7/1p2R3/8/PP6/1Kb5/8 b - - 0 38
In this position, White is threatening Ng6+ picking up the rook. Black must keep
the bishop on the diagonal so as to meet Ng6 with Bxg6 protecting the rook.
White can therefore draw by attacking the bishop with 38.Kb2 as here, 38...Bd3
only move, 39.Kc3 Bb1 40.Kb2 and a repetition is in order. Instead I see
38...g6?? with -0.23 at 14 plies, 39.Kxc2 Kg7 40.Nf5+ Kf8 41.Rxe8 and things
went downhill with white winning the endgame. I honestly don't care how the
endgame went since when I ask Pro Deo (same settings) what it thinks of the
position after 38.Kb2 it says:
38...Bxb3 39.Kxb3
= (0.04) Depth: 1 00:00:00
38...Rxe7 39.Rxe7
³ (-0.31) Depth: 1 00:00:01
38...Bf5 39.Rxf5 Rxe7
-+ (-1.94) Depth: 1 00:00:01
38...Bg6
-+ (-2.04) Depth: 1 00:00:01
38...Bg6 39.Nxg6+ hxg6 40.Rxe8+ Kh7
+- (2.99) Depth: 2 00:00:01
38...Bd1 39.Ng6+ hxg6 40.Rxe8+ Kh7
= (0.08) Depth: 2 00:00:01
38...Bf5 39.Rxf5 Rxe7
-+ (-1.94) Depth: 2 00:00:01
38...Bf5 39.Rxf5 g5
² (0.38) Depth: 3 00:00:01
38...Bd1 39.Ng6+ hxg6 40.Rxe8+
= (0.08) Depth: 3 00:00:01
38...Bd3 39.Kc3 Rd8
-+ (-1.94) Depth: 3 00:00:01
38...Bd3 39.Kc3 Rd8 40.Nd5
-+ (-1.89) Depth: 4 00:00:01
38...Bd3 39.Kc3 Rd8 40.Nc6 Rc8 41.Kxd3
-+ (-1.86) Depth: 5 00:00:01
38...Bd3 39.Kc3 Rd8 40.Nc6 Rc8 41.Kxd3 Rxc6
-+ (-1.86) Depth: 6 00:00:01 13kN
38...Bd3 39.Kc3 Rd8 40.Nc6 b4+ 41.Nxb4 Bb5
µ (-0.87) Depth: 7 00:00:01 25kN
38...Bd3 39.Kc3 Rd8 40.Nc6 b4+ 41.Nxb4 Bb5 42.Nd5
µ (-0.89) Depth: 8 00:00:01 45kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 9 00:00:02 119kN
38...Bxb3 39.Kxb3 Rf8 40.Nd5 Rf3+ 41.Kb4 g6 42.Re6 Rf5
= (0.00) Depth: 9 00:00:02 369kN
38...Bxb3 39.Kxb3 Rf8 40.Re6 Rf3+ 41.Kb4 Rf4+ 42.Kc5 Ra4 43.Nf5 h5 44.Re8+ Kh7
= (0.00) Depth: 10 00:00:02 730kN
38...Bxb3 39.Kxb3 Rf8 40.Re6 Rf6 41.Rxf6 gxf6 42.Kb4 Kg7 43.Kc5 h5 44.Nf5+ Kg6
= (0.00) Depth: 11 00:00:03 2473kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 11 00:00:04 2811kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 12 00:00:04 3593kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 13 00:00:07 6793kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 14 00:00:13 12772kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 15 00:00:29 31455kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 16 00:01:09 76135kN
38...Bd3 39.Kc3 Bb1 40.Kb2 Bd3
= (0.00) Depth: 17 00:03:18 224497kN
So how the heck did 38...g6 get played?? Ed? Anyone?
My settings have been on my Athlon 2400XP+ with 256 MB ram:
Nunn I set, 30 mins + 2 sec increment, 64 MB hash for Junior 8 and w6 setting
(60 MB hash) for Pro Deo, permanent brain off.
Albert
This page took 0.01 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.