Author: Javier Gamero
Date: 21:14:32 12/25/04
Go up one level in this thread
I have done some calculations of Nodes-per-Sec and "intensity" parameter. The
results are somewhat unexpected (in my mind, at least).
I have used the position after 1. a3 , a6 (LOL, the double Anderssen system), I
have configured Crafty Classic 2004 with this crafty.rc:
log on
krafty kramnik
hash=0K
hashp=0K
intensity 1000
exit
(The choice of Kramnik is for no particular reason!). I have used Arena as GUI
(again for no particular reason, but I had some problems in my winboard with the
"intensity" command). I put 5 seconds as fixed time per movement. I got the
following results:
intensity NpS
===============
10000 2067291
9999 2067291
5000 2067291
3000 1095714
2000 501312
1000 501312
500 137632
300 137632
200 29007
100 8148
50 2190
30 2190
20 2190
10 57
5 57
3 57
2 57
1 57
1) NpS don't increase smoothly with intensity. It seems to be discrete, not
continous. Some of the "jumps" are very big (for example, from 57 to 2190). This
implies there are perhaps a dozen significatives choices for intensities.
2) I had hoped NpS were proportional to intensity, but it is only approx
proportional in the range 200 - 10000. For lower intensities NpS are less than
proportional respect intensity.
It is possible that these results are product of some particular effect of
position/configuracion/GUI, I don't know Crafty source at all and I cannot know
why it happens.
Interestingly these results fit nicely with Michael's (Byrne) estimation of
strengths (calculating ELOs in function of NpS instead of intensities):
int NpS ELO(FIDE?)
1 57 1250
50 2190 1580
10000 2067291 2200
I hope this small "research" (rolleyes) is useful for something.
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.