Gian-Carlo Pascutto wrote: > 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. Ok, then I will run it with the selectivity left on.
> > Don't forget to measure the effort needed to reach the fixed depths > and compare that too. A change there has real-life implications. Let me first just focus on this aspect which is more easily testable, the assertion that the quality of the evaluation function becomes more important with depth, independent of time. I think we would all agree that they are highly correlated even if not perfectly correlated. I can provide rough measures of average game time per player but it won't be perfect since my machine does not always have the same load - but it should be roughly correct. > > Thanks for wanting to do that study and Try To Settle The Matter Once > And For All. I don't really think it will settle it because it will always be possible to criticize the way the test was done no matter how I might have constructed it. But it will be a starting point and the burden of proof, whatever happens, will fall more on the critic of the study. > >> 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) That's always been the right formula, you just misunderstood me. I think you are focused more on recent trends but I am speaking to a long term trend - there is no question about the fact that evaluation functions in chess are far more sophisticated (and at least somewhat slower) than they used to be. I remember the days when a certain program amazed everyone for capturing a pawn with a knight in a knight vs pawn ending because it realized that it couldn't win with just a knight. That program was considered knowledge intensive for actually having this little bit of specialized knowledge. We have come a very long way since then. - Don _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
