Author: Dan Homan
Date: 06:46:31 06/18/98
Go up one level in this thread
On June 18, 1998 at 07:39:28, Peter McKenzie wrote:
>Bob Hyatt wrote:
>
>>How about this:
>>
>>Pick a position from the first 12 test cases, ignoring number one which
>>should never have been in the test, and then run the test position for
>>depth=1, then 2, then 3, then 4, and so forth, and publish the node counts
>>here. I'm fixing to do the same... in fact, here are my numbers for
>>position #5, from 1 to 10 plies deep:
>>
>> depth total nodes ratio <added to original quote>
>> 1 61
>> 2 299 4.9
>> 3 1,939 6.5
>> 4 9.052 4.7
>> 5 41,606 4.6
>> 6 121,430 2.9
>> 7 459,585 3.8
>> 8 1,244,527 2.7
>> 9 2,935,151 2.4
>> 10 6,494,133 2.2
>
>Node counts from lambChop:
>
> depth total nodes
> 1 63
> 2 328 5.2
> 3 3,870 11.8
> 4 15,552 4.0
> 5 68,331 4.4
> 6 264,674 3.9
> 7 1,520,524 5.7
> 8 3,691,857 2.4
> 9 16,369,682 4.4
>
My program, EXchess gets:
depth total nodes
1 119
2 805
3 3653
4 17205
5 46397
6 94002
7 349740
8 1320973
9 3695135
10 9998085
So my program starts our worse than crafty, gets better by the
middle depths, and gets worse again. The only part I am concerned
about it the getting worse again at greater depths. I think my
extensions may get a bit out of hand... I'll have to play around
with them.
I can also change these numbers significantly by changing my
futility cutoff in the quiescent search. The cutoff is now
set pretty agressively.
- Dan
>looks like I have some testing to do...
>
>>those are using a 12mb hash table (crafty) and a 3m pawn hash
>>table. the 10 ply search took 1:16 (all were run with one cpu).
>>these node counts are the *total* node counts, so the 10 ply
>>count includes all the ones before it, plus the nodes added by
>>the 10 ply iteration...
>>
>>
>>Bob
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.