How do I know if I have a new feature? For instance, I never did nullmove pruning. Since start, I did what I would say is nullmove reductions. I never thought this was new, but apparently it might have been. Later, nullmove verified research was published, which is very, very close. I always thought that my handling of checkmates was different. I found recently that HGMuller is doing something either similar or identical. I can go on and on. How do you define new? Anyway, I may have new things and I do not even know it [1]. I hope I have something new in the evaluation, otherwise it will be kind of sad, but who knows?sje wrote:It seems to me that a program written from scratch without access to other sources should contain at least one new item: a move ordering trick, an evaluation term, or maybe a novel user interface feature. Now the new item may not be superior to what wight be available elsewhere, but it word be new.michiguel wrote:How do you define feature?
But maybe I'm asking for too much here. I'll guess that back in the pre Internet days of punch cards and paper tape, it was easier to write a de novo chess program just because there were no other sources available.
Well, Greenblatt's MacHack VI was generally available to the AI and DEC communities. But that was about it for nearly a decade.
If someone were to compile the Truly Exhaustive List of Chess Programs, it would be nice if each program had a list of new techniques that it first introduced.
It is not that straightforward IMHO to know "novelty". It is simpler in Academia because we keep meticulous records, not always in CC, let alone when it is treated as hobby. I will define myself a computer chess enthusiast even if it turns out that there is no novel technique in Gaviota.
Miguel
[1] It is easy to be certain that something is NOT new.
