Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Relationship of Ratings verses Time for Thought

Author: Jim Monaghan

Date: 11:16:10 08/07/02

Go up one level in this thread


On August 07, 2002 at 13:18:35, William H Rogers wrote:

>Your test closely resembles the test done years ago where they tested program
>strength in relation to ply depth. Up to about 8 or 9 plys the programs strength
>increased quite rapidly then it got smaller and smaller with each increase of
>ply depth.
>Is there a way that you could show the approximate depth obtained for various
>times? It would be nice.
>Then all we would have to do is to adjust for computer speed to come close to
>the ratings on our machines.
>Good work
>Bill

Hi Bill,

Thanks for your's and James' comments. I'm not sure what you are asking Bill
about the ply depth. I ran the test in two phases. First, all 360 positions were
pumped through at 10 seconds per move. I then took the missed.epd positions and
ran those through at 2 minutes per move. So I have two log files. Actually, four
files, the other two are copy and pastes from the GUI of the last 50 or so
positions, which I've pasted as well. Here's an entry from all four:

Solved within 10 seconds:

rq1n1bk1/3b1pp1/3p3p/pppP3N/8/1P4PP/PBPQ1PB1/4R1K1 w - - bm Qxh6; id "IQ.1268";
ce 462; acd 12; acn 3154671

I understand this to mean that ce=the eval of the PV, acd= the depth achieved
that was being worked on, not completed, and acn= the nodes. So I'm a little
confused on how ply depth achieved can be useful to you. Maybe the difficulty is
that the program didn't stop at one second in the first example after 9 ply and
go on to the next one. So I think my answer is that with the data I have with
this engine, I can't develop a ply table verses time for the entire test. I
would have to use the test on another engine like Crafty, I think, which saves
time and skips to the next position on finding the right answer. Shredder is set
up to use a constant time. Do I understand correctly? Would a time/ply/rating be
a more useful table for comparisions with other programs or hardware?

Position:  bm Qxh6; id "IQ.1268";
Solution : Qd2xh6!
  9.01	 0:01 	+3.21 	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ Kg8 4.Nxb8 Rxb8 5.g4 a4
6.bxa4 bxa4 (270.577) 252.6
 10.01	 0:01 	+3.46++	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ Kg8 4.Re8 f6 5.Nxb8 Kg7
6.Rxd8 Kg8 (381.389) 268.2
 10.01	 0:01 	+3.67 	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ Kg8 4.Re8 Nc6 5.Nxb8 Nxb8
6.c4 bxc4 7.bxc4 (476.263) 284.8
 11.01	 0:03 	+3.92++	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ Kg8 4.Re8 Nc6 5.dxc6 Qa7
6.Nf6+ Kg7 7.c7 Qxc7 8.Nd5+ Kg6 9.Nxc7 Rxe8 10.Nxe8 (1.164.756) 295.9
 11.01	 0:04 	+4.42++	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ Kg8 4.Re8 Nc6 5.dxc6 Qa7
6.Nf6+ Kg7 7.c7 Qxc7 8.Nd5+ Kg6 9.Nxc7 Rxe8 10.Nxe8 (1.330.923) 298.0
 11.01	 0:05 	+4.62 	1.Qxh6 gxh6 2.Nf6+ Kh8 3.Nxd7+ f6 4.Bxf6+ Kg8 5.Nxb8 Rxb8
6.Re8 Nc6 7.Rxb8 Nxb8 8.c4 bxc4 9.bxc4 (1.727.874) 311.4
best move: Qd2xh6 time: 0:10.284 min  n/s: 307.352  CPU 99.4%  nodes: 3.154.671

Solved within 2 minutes:

5rk1/2rnqp1p/b3p1pP/2ppP3/1p3QN1/pP1P1BP1/P1P2P2/R3R1K1 w - - bm Bxd5; id
"IQ.1126"; ce 142; acd 14; acn 32752047; pv Bxd5 Bb5 c4 bxc3 Be4 Bc6 Bxc6 Rxc6
Rac1 g5 Qe3 f5 exf6 ;

Position:  bm Bxd5; id "IQ.1126";
Solution : Bf3xd5!
 11.01	 0:00 	+0.17 	1.Nf6+ Nxf6 2.Qxf6 Qxf6 3.exf6 (205.580) 244.4
 12.01	 0:11 	+0.13 	1.Nf6+ Nxf6 2.exf6 Qd8 3.g4 Rc8 4.Rac1 Qb6 5.g5 Rfd8 6.Rcd1
Bb7 7.Qe5 d4 (3.302.414) 285.7
 13.01	 0:18 	+0.17 	1.Nf6+ Nxf6 2.exf6 Qd8 3.g4 Rc8 4.Rac1 Qb6 5.g5 Rcd8 6.Bg2
Bc8 (5.159.009) 285.0
 13.02	 0:28 	+0.18++	1.Bxd5 exd5 2.e6 Nb6 3.Nf6+ Kh8 4.Qxc7 Qxc7 5.e7 Re8
6.Nxe8 Qd8 7.exd8Q (8.195.705) 286.0
 13.02	 0:32 	+0.53++	1.Bxd5 exd5 2.e6 Nb6 3.Nf6+ Kh8 4.Qxc7 Qxc7 5.e7 Re8
6.Nxe8 Qd8 7.exd8Q (9.379.447) 285.4
 13.02	 1:50 	+1.42 	1.Bxd5 Bb5 2.c4 bxc3 3.Be4 Bc6 4.Bxc6 Rxc6 5.Rac1 g5 6.Qe3
f5 7.exf6 (30.321.661) 273.4
best move: Bf3xd5 time: 2:00.363 min  n/s: 272.133  CPU 99.8%  nodes: 32.752.047

Thanks,
Jim



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.