Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Branching factor, etc

Author: Omid David

Date: 13:54:53 08/30/02

Go up one level in this thread


On August 30, 2002 at 16:12:38, Dann Corbit wrote:

>On August 30, 2002 at 13:56:09, Omid David wrote:
>
>>On August 30, 2002 at 13:42:29, Dann Corbit wrote:
>>
>>>On August 30, 2002 at 09:56:21, Omid David wrote:
>>>[snip]
>>>>How do you calculate branching factor here?
>>>
>>>How long does it take to get from ply n to ply n+1
>>>
>>>Or
>>>
>>>How many nodes in ply n verses ply n+1
>>>
>>>Both work out about the same.
>>
>>But there is the odd/even issue. For example how do you count the b-factor in
>>this example (Genesis, depth 10):
>>
>>---------------
>>created 64 MB hash table, 4194304 entries.
>>using NegaScout search.
>>quiescence search on [check on]
>>null move search on [R = 3] [v]
>>aspiration window search off.
>>opening book off.
>>resign on [score < -1200]
>>searching 9 plies per move.
>>
>>White(1): load test
>>can't load file: test
>>File loaded: test.txt
>>
>>       +---+---+---+---+---+---+---+---+
>>    8  |   |   | *B|   |   |   | *K|   |
>>       +---+---+---+---+---+---+---+---+
>>    7  |   | *P| *P| *P|   | *P| *P| *P|
>>       +---+---+---+---+---+---+---+---+
>>    6  | *R|   |   |   |   |   |   |   |
>>       +---+---+---+---+---+---+---+---+
>>    5  | *P|   | P |   | P | P | *R|   |
>>       +---+---+---+---+---+---+---+---+
>>    4  |   |   |   | B |   |   |   |   |
>>       +---+---+---+---+---+---+---+---+
>>    3  | R |   |   |   |   |   |   |   |
>>       +---+---+---+---+---+---+---+---+
>>    2  | P |   | P |   |   |   | P | P |
>>       +---+---+---+---+---+---+---+---+
>>    1  |   |   |   | R |   |   | K |   |
>>       +---+---+---+---+---+---+---+---+
>>         a   b   c   d   e   f   g   h
>>
>>White(1): sd 10
>>searching 10 plies per move.
>>White(1): p
>>depth   time  score   variation
>>  1->   0.00  -0.13   1. a3f3
>>                      nodes 107  nps 0  q 68%  hhits 0%
>>  2->   0.02  -0.14   1. f5f6 g5g6
>>                      nodes 618  nps 30900  q 71%  hhits 0%
>infinite
>
>>  3->   0.04  -0.07   1. f5f6 a5a4 2. e5e6
>>                      nodes 1460  nps 73000  q 56%  hhits 4%
>2x
>
>>  4->   0.12   0.00   1. f5f6 g5f5 2. a3d3 a6e6
>>                      nodes 6908  nps 86350  q 58%  hhits 3%
>3x
>
>>  5->   0.31  -0.09   1. f5f6 g7f6 2. e5f6 d7d6 3. d4e3
>>                      nodes 17859  nps 93502  q 52%  hhits 6%
>2.6x
>
>>  6->   0.69   0.12   1. f5f6 g7f6 2. e5f6 d7d5 3. c5d6 a6d6
>>                      nodes 38199  nps 100523  q 36%  hhits 6%
>2.2x
>
>>  7->   1.73   0.02   1. f5f6 g7f6 2. e5f6 a6e6 3. a3e3 d7d5 4. c5d6
>>                      nodes 112116  nps 107596  q 45%  hhits 9%
>2.5x
>
>>  8->   4.38   0.08   1. f5f6 g7f6 2. e5f6 a6e6 3. a3a5 e6e2 4. d4f2 e2c2
>>                      nodes 299957  nps 113448  q 35%  hhits 8%
>2.5x
>
>>  9->  17.56   0.16   1. f5f6 g7g6 2. a3a4 g5g4 3. c2c4 g4e4 4. d4c3 c7c6
>>                      nodes 1484609  nps 112649  q 41%  hhits 10%
>4x
>
>> 10->  42.40   0.20   1. f5f6 g7g6 2. a3a4 d7d5 3. e5d6 c7d6 4. d4e3 g5e5
>>                      5. e3f4 e5c5
>>                      nodes 2838138  nps 114233  q 28%  hhits 7%
>2.4x
>>----------------------------------------------------------
>>nodes 4799971  time 42.41  nps 113177
>>leaf 3357952  q 1603786  eval 3824785
>>null 2576438  null-cut 307402
>>hhits 395391  hexact 253858  hleaves 253524
>>hput 435376  hoverwrites 1328
>>leaves 69%  quiesc 33%  null 53%
>>null-cut 77%  aspwin-suc 100%
>>hhits 8%  hoverwrites 0%  husage 8%
>>White(1): f5f6
>
>Looks like 2.5x answers pretty well.

Thanks Dann.

Calculating the b-factor based on time (from ply n-1 to n) seems more
practically correct because of different node count methods people use. But in
published papers I've only seen b-factor based on node count (which if node
count is done properly they shouldn't be much different).



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.