Author: Eelco de Groot
Date: 19:12:17 03/27/05
Go up one level in this thread
>
>New Pro Deo Betazoid T5m personality, also known as 'Jaws 0.1', is like a fish
>in the water in this position and reaches a search efficiency of 97.5%. It is
>not so remarkable that it can find Nf5 but it does not use too aggressive
>settings I hope and also "regular" null-move. Heavily borrowed from Albert's
>settings, and Dieter Eberle's e3. No idea how good this plays games as it hasn't
>been tested yet.
>
Pro Deo Betazoid T5m (Jaws 0.1) lost two tournament games against Chess Tiger
14.0. That was not very heartening but it also does not say very much of course.
Some tactical positions analysis went better. It may turn out however that
either
[Mobility = 120]
[Attractiveness = 110]
[Selective Search = 80]
[Pawn Formation = 115]
are too risky, especially against Angstgegner Chess Tiger (very hard to beat).
Attractiveness can be lowered to 100 or 103 in the few tactical positions I
tried, without much tactical loss.
As another testposition, I would like to find a version of Jaws 0.1 that plays
12..Nxe5 in Ed's immortal game that he played against his own brainchild.
[D]rnbq1rk1/pp1n1p1p/4p1p1/3pP1NP/1b3P2/1PpB4/P1P3P1/R1BQK2R b KQ - 0 1
I am assuming that 12..Nxe5 is the only move here that can draw. But maybe there
are a few others, I am not totally sure sure about 12..Qb6 that also got a
draw-score in some searches.
Several versions came close but would fail low, others would play Qc7 but with
very small changes never choose that move. It seems a good testposition, but the
scores I saw I could not always understand.
My original notes about this position were as follows:
------------------------------------------------------------------------------
S&D 6_7e3, modified from Albert Silver's settings, does not consider 12...Qc7.
It wants to return material with 12...Nxe5, but this fails low.
00:40:28 12.00 0.17 12..Nxe5 13.fxe5 Qc7 14.Nxh7 Qxe5 15.Kf1 Nc6
16.Nxf8 Qf6 17.Qf3
White sacs a Knight on h7 to sustain the attack! The scores at twelve ply are
close together now so maybe Nxe5 can come back as first choice, on deeper plies?
In tournament conditions this would certainly be played. (I see this agrees more
or less with the analysis from Albert with Fritz and Shredder 7.04, they also
play 12...Nxe5. Shredder thinks White can press on after 12.Nxe5, Fritz thinks
Black is a little better. They play 14.Qe2 Nc6 15.Nxh7 where S&D 6_7e3 goes
straight to 14.Nxh7 and thinks it's a draw, see below)
After a long computation to resolve the fail-low then 12...Nc6 is chosen, but
the advantage for Black is not enough anymore to think that Black can play for a
win. A pity that resolving the fail-low of 12Nxe5 took rather long, I have not
tried any other settings here.
Celeron 500 MHZ, 28 Mb HT, Pro Deo 1.0 S&D 6_7e3
00:00:01.7 1,88 1 26 gxh5 Bxh7+ Kh8 Qxh5
00:00:01.8 3,72 1 28 Nc6
00:00:01.8 3,80 1 48 Qb6
00:00:01.8 2,10 2 485 Qb6 hxg6 hxg6 Rh7
00:00:01.8 2,90 2 567 Bc5 hxg6 hxg6
00:00:01.8 3,23 2 1142 h6 a3
00:00:01.8 3,44 3 2817 h6 Nf3 gxh5 Rxh5
00:00:01.8 3,23 4 9863 h6 Nf3 Nc5 hxg6 Nxd3+ cxd3
00:00:01.9 3,23 5 18078 h6 Nf3 Nc5 hxg6 Nxd3+ cxd3
00:00:02.0 3,26 5 30846 Be7 hxg6 hxg6 Nf3 Nc5 Bb5
00:00:02.3 3,10 6 74365 Be7 Nxh7 Bh4+ Kf1 Kxh7 Qg4 Nxe5 hxg6+
00:00:06.7 1,73 7 611512 Be7 Nxh7 Bh4+ Kf1 Kxh7 Qg4 Kg8 Rxh4 g5 fxg5
00:00:18.9 1,73 8 2127857 Be7 Nxh7 Bh4+ Kf1 Kxh7 Qg4 Kg8 Rxh4 g5 fxg5
00:00:36.8 1,27 9 4529618 Be7 hxg6 Bxg5 Qh5 h6 fxg5 Nxe5 Qxh6
00:01:02.3 1,45 9 8231431 Nxe5 fxe5 Qc7 Qe2 h6 Nf3 g5 Be3 f5
00:01:39.3 1,16 10 14072186 Nxe5 fxe5 Qc7 Qe2 f6 exf6 Rxf6 hxg6 hxg6 Kd1 Qg3 Rh3
Qc7
00:05:25.1 1,11 11 47895984 Nxe5 Nxh7 Qc7 Nxf8 Nxd3+ Qxd3 Bxf8
00:40:29.7 0,16 12 361714317 Nxe5 fxe5 Qc7 Nxh7 Qxe5+ Kf1 Nc6 Nxf8 Qf6+ Qf3
00:48:43.8 0,24 12 438308310 Bc5 hxg6 fxg6 Nxh7 Qe7 Bxg6
00:52:05.2 0,28 12 469869732 Nc6 Nxh7 Bc5 hxg6 Ndxe5
{
How does the position after 12..Nxe5 13.fxe5 Qc7 14.Nxh7 Qxe5 turn out on deeper
analysis? It seems an interesting alternative, but I'm not very good at tactics.
Is 12...Nxe5 better than 12...Qc7 or 12...Nc6?
Pro Deo 1.0 S&D 6_7e3 sees a draw as White's best hope after a possible 12..Nxe5
13.fxe5 Qc7 14.Nxh7 Qxe5
[D]rnb2rk1/pp3p1N/4p1p1/3pq2P/1b6/1PpB4/P1P3P1/R1BQK2R w KQ - 0 1
00:09:10 12.00 0.00 15.Kf1 Kxh7 16.hxg6 Kg8 17.g7 Qf6 18.Ke1 Qxg7
19.Bh7
}
------------------------------------------------------------------------------
Betazoid T5m gives the following analysis in the first position, this is about
the only setting I have found so far that really sticks to Nxe5, but only at
great depth. I let the position run about 13 hours, much longer here than with
the other settings, maybe they all settle eventually to Nxe5
00:00:00.4 1,89 1 26 gxh5 Bxh7+ Kh8 Qxh5
00:00:00.4 3,73 1 28 Nc6
00:00:00.4 3,80 1 48 Qb6
00:00:00.4 2,82 2 195 Qb6 hxg6 hxg6
00:00:00.5 2,91 2 313 Bc5 hxg6 hxg6
00:00:00.5 2,91 2 425 Nc6 hxg6 hxg6
00:00:00.5 3,17 2 1052 h6 a3
00:00:00.5 3,39 3 2091 h6 Nf3 gxh5 Rxh5
00:00:00.6 3,24 4 7491 h6 Nxf7 Kxf7 Qg4 g5 fxg5
00:00:00.7 1,07 5 23475 h6 hxg6 Nxe5 fxe5 hxg5 g7 Kxg7
00:00:00.8 2,64 5 33046 Nc6 Nxh7 Kxh7 hxg6+ Kg8 gxf7+ Rxf7 Qg4+ Rg7
00:00:01.0 3,03 5 54758 Be7 hxg6 hxg6 Nf3 Qb6 a4
00:00:01.4 3,29 6 74853 Be7 hxg6 hxg6 Nf3 Qb6 a3 Nc6
00:00:02.7 2,36 7 216434 Be7 Nxh7 Bh4+ Kf1 Kxh7 hxg6+ Kg7 gxf7 Rh8 Qg4+ Kxf7
00:00:07.4 2,35 8 729340 Be7 Nxh7 Bh4+ Kf1 Kxh7 hxg6+ Kg7 Qg4 Nxe5 fxe5 fxg6+
00:00:17.0 1,60 9 1896036 Be7 hxg6 fxg6 Nxe6 Bh4+ Kf1 Qb6
00:01:13.5 1,19 10 8598790 Be7 Nxh7 Bh4+ Kf1 Kxh7 Qg4 Kg8 Rxh4 g5 h6 f5
00:05:20.6 -0,35 11 39223313 Be7 hxg6 fxg6 Bxg6 Bxg5 Bxh7+
00:14:16.4 0,48 11 107787856 Nxe5 fxe5 Qc7 Qg4 Qxe5+
00:22:31.8 0,65 11 164268315 Nf6 exf6 Qxf6 hxg6 hxg6 Be3
00:24:25.6 -1,19 12 179657482 Nf6 exf6 h6 hxg6 fxg6 f7+ Kg7
00:29:13.9 0,17 12 216798395 Nxe5 fxe5 Qc7 Qg4 Qxe5+
00:45:13.0 0,36 12 343604723 Nc6 Nxh7 Qe7 Nxf8 Nxf8
01:33:15.4 0,17 13 735741739 Nc6 Nxh7 g5 h6 Kh8 Nxf8 Qxf8
03:32:40.6 -0,11 14 1710597700 Nc6 hxg6 fxg6 Nxh7
04:41:56.9 0,00 14 -1995759887 Nxe5 fxe5 Qc7 Nxh7 Qxe5+ Kf1 Kxh7 hxg6+
13:06:34.8 0,00 15 -1900258246 Nxe5 fxe5 Qc7 Nxh7
Now I made a setting with exactly the same numbers as T5m, plus a few switched
off settings inside , that still should produce exactly the same but does not;
slightly different nodecounts and Nxe5 only gets 0.00 instead of 0.17 at twelve
ply. I'll see if I can reproduce it next time I try this. This is what I got
with the second version:
00:00:00.5 1,89 1 26 gxh5 Bxh7+ Kh8 Qxh5
00:00:00.5 3,73 1 28 Nc6
00:00:00.6 3,80 1 48 Qb6
00:00:00.6 2,82 2 195 Qb6 hxg6 hxg6
00:00:00.6 2,91 2 313 Bc5 hxg6 hxg6
00:00:00.6 2,91 2 425 Nc6 hxg6 hxg6
00:00:00.6 3,17 2 1052 h6 a3
00:00:00.6 3,39 3 2091 h6 Nf3 gxh5 Rxh5
00:00:00.7 3,24 4 7491 h6 Nxf7 Kxf7 Qg4 g5 fxg5
00:00:00.8 1,07 5 23475 h6 hxg6 Nxe5 fxe5 hxg5 g7 Kxg7
00:00:00.9 2,64 5 33046 Nc6 Nxh7 Kxh7 hxg6+ Kg8 gxf7+ Rxf7 Qg4+ Rg7
00:00:01.1 3,03 5 54758 Be7 hxg6 hxg6 Nf3 Qb6 a4
00:00:01.6 3,29 6 74853 Be7 hxg6 hxg6 Nf3 Qb6 a3 Nc6
00:00:02.9 2,36 7 216434 Be7 Nxh7 Bh4+ Kf1 Kxh7 hxg6+ Kg7 gxf7 Rh8 Qg4+ Kxf7
00:00:07.6 2,35 8 729413 Be7 Nxh7 Bh4+ Kf1 Kxh7 hxg6+ Kg7 Qg4 Nxe5 fxe5 fxg6+
Kg1
00:00:17.4 1,60 9 1907942 Be7 hxg6 fxg6 Nxe6 Bh4+ Kf1 Qb6 Rxh4 Qxe6 Be3
00:01:13.5 1,19 10 8551998 Be7 Nxh7 Bh4+ Kf1 Kxh7 Qg4 Kg8 Rxh4 g5 h6
00:05:05.1 -0,28 11 37008205 Be7 hxg6 fxg6 Bxg6 Bxg5 Bxh7+
00:12:51.1 0,44 11 96624198 Qc7 hxg6 fxg6 Rxh7 Qc5
00:14:03.9 0,48 11 105961726 Nxe5 fxe5 Qc7 Qg4 Qxe5+ Kd1 Bc5
00:20:57.7 0,65 11 155547966 Nf6 exf6 Qxf6 hxg6 hxg6 Be3
00:23:07.5 -1,19 12 172727365 Nf6 exf6 h6 hxg6 fxg6 f7+ Kg7
00:28:14.8 0,00 12 214198450 Nxe5 fxe5 Qc7 Nxh7 Qxe5+
Second version of T5m looked like this:
[Personality = Betazoid T6c] * Pro Deo 1.1 (February 2005)
[Pawn Value = 100]
[Knight Value = 101]
[Bishop Value = 100]
[Rook Value = 100]
[Queen Value = 100]
[My Queen = 101]
[King Safety = 160]
[Mobility = 120]
[Pawn Structure = 110]
[Passed Pawns = 100]
[Pins = 101]
[Bishop Pair = 101]
[Chess Knowledge = 200]
[Attractiveness = 110]
[Attacking = 100]
[Strength of Play = 100]
[Draw Contempt Factor = 0.00]
[Selective Search = 80]
[Search Technique = NULLMOVE]
[Engine Master = off] off
[Engine Learner = off] off|on
[Book Learner = off] off|passive|moderate|strong|aggressive
[Position Learner = off] off|on
[Extended Book Learner = off] off|read|write|read&write
*
[Pruning = MISC_25] *
[Pruning = MISC_37] *
[Pruning = MISC_40] *
[Pruning = MISC_52] *
[Pruning ? MISC_53] *
[Pruning = MISC_66] *
[Pruning = MISC_69] *
[Pruning = MISC_78] *
*
[Center Control = 30] *
[Bishop Mobility = 25] *
[Right to Move = 75] *
[Strong Squares = 128] *
*
[Pruning = MISC_02] *
[Pruning = MISC_03] *
[Pruning = MISC_39] *
[Pruning = MISC_08] *
[Pruning = MISC_12] *
[Pruning = MISC_16] *
[Pruning = MISC_22] * Slightly safer Selectivity
[Pruning ? MISC_30] * extend singular checks
[Pruning ? MISC_28] * (2,2,1,1) >> in test <<
[Blitz Time Control = 120] * use more time at blitz levels
*
[Pawn Pressure midg = 165] * weight pawn pressure middlegame
[Pawn Pressure end0 = 140] * weight pawn pressure early endgame
[Pawn Pressure end1 = 125] * weight pawn pressure endgame
[Pawn Pressure end2 = 110] * weight pawn pressure simple endgame
*
[Knight Mobility = 88] * 88
[Double Pawns = 112] * 112
[Pawn Formation = 115] * 108
[Weak Pawns = 112] * 112
[Minimum Knight Mobility = 100] * 100
[Minimum Bishop Mobility = 100] * 100
[Passed Pawn King Tropism = 100] * king tropism to own passed pawns
[Progressive Isolated Pawns = 100] * weight isolated pawns (new style)
*
[ANTI-GM = OFF]
[EVALUATION = SMART]
[Extensions (remaining)= 3]
[Extensions (checks)= normal] * low|moderate|normal|strong|all
[Extensions (captures)= 0]
Last try was with a lower value for the queen, as in the Q3 and Q5T settings.
This version also chooses Nc6 at thirteen ply but this time only with a score of
0.01. Nxe5 only came on top with 0.00 after 50 minutes on the Celeron 500, then
Nc6 topped that with just 0.01 point. Grrr...
If you happen to have a version of Pro Deo that plays 12..Nxe5, please report
immediately here or to the nearest policestation!
Anyway I'll continue my searches but for today it's been enough!
Eelco
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.