Reducing branching factor

Discussion of chess software programming and technical issues.

Moderator: Ras

bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Reducing branching factor

Post by bob »

Kempelen wrote:Idem as having hash hits or null move success hits, I am calculating % of nodes which are reduced or extended.

Of the total nodes (excluding quiesce nodes, as there I dont use extension), I am getting around 2% of total nodes are extended and arund 1% reduced.
These datas sound very few for me.

What are good % for extensions and reduccions, in average?

thx.
some samples from a recent game on ICC at about 3 minutes a move:

total nodes 390 million check extensions 24.7M, qchecks 23.9M, 1 legal reply extensions, 3.3M, mate threat extension 2.5M, I attempted to reduce 174M moves and actually reduced 33.1M.

total nodes 446M, check extensions 32M, qchecks 39.8M, 1 reply 3.6M, mate threat 2.2M, attempted to reduce 170M, reduced 39M

So yes, your numbers sound _way_ low...
User avatar
Kempelen
Posts: 620
Joined: Fri Feb 08, 2008 10:44 am
Location: Madrid - Spain

Re: Reducing branching factor

Post by Kempelen »

bob wrote:total nodes 390 million check extensions 24.7M, qchecks 23.9M, 1 legal reply extensions, 3.3M, mate threat extension 2.5M, I attempted to reduce 174M moves and actually reduced 33.1M.
Hi Bob,
When you say 390 million total nodes, are you saying....

a) 390M = nodes + qnodes? or
b) 390M = nodes?

and.... are those number attemps or only succesfulls hits?

thx
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Reducing branching factor

Post by bob »

Kempelen wrote:
bob wrote:total nodes 390 million check extensions 24.7M, qchecks 23.9M, 1 legal reply extensions, 3.3M, mate threat extension 2.5M, I attempted to reduce 174M moves and actually reduced 33.1M.
Hi Bob,
When you say 390 million total nodes, are you saying....

a) 390M = nodes + qnodes? or
b) 390M = nodes?

and.... are those number attemps or only succesfulls hits?

thx
total nodes including q-search which do not get reduced. Not sure what you mean about "hits" however. In the above, 24.7M moves triggered the "give check" extension... I'm not sure what the "hit" means in this context...
User avatar
Kempelen
Posts: 620
Joined: Fri Feb 08, 2008 10:44 am
Location: Madrid - Spain

Re: Reducing branching factor

Post by Kempelen »

bob wrote: total nodes including q-search which do not get reduced. Not sure what you mean about "hits" however. In the above, 24.7M moves triggered the "give check" extension... I'm not sure what the "hit" means in this context...
OK Bob, very clear now. I was confused with another concept. Thanks a lot.... as always.