Don Dailey wrote:
BOTH versions have NullMove Pruning and History Pruning turned off
because I feel that it would bias the test due to interactions
between selectivity and evaluation quality (I believe it would make
the strong version look even more scalable than it is.)
There is nothing in nullmove pruning or history pruning that depends on
the features you want to remove, so this sounds a bit weird and
arbitrary to me.
Don't forget to measure the effort needed to reach the fixed depths and
compare that too. A change there has real-life implications.
Thanks for wanting to do that study and Try To Settle The Matter Once
And For All.
I answered that above - in summary it's pretty difficult to assess an
evaluation function due to so many factors that interact. I'm not a
big knowledge advocate, I believe that the more knowledge you have, the
more difficult it is to get it right - and indeed it can conflict with
other knowledge.
This I agree with 100%.
But note that a limited amount of well-tuned knowledge doesn't prevent
being fast, on the contrary. So the trend I see is towards smart and
fast, but not too knowledgable. (That is my answer to your story)
--
GCP
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
