A:
===
well, imo the first question should be what "innovation" means today or
maybe, what it means in general. Introducing new techniques like nullmove,
lmr,razoring,futility and so on is only one step of innovation. We do
only accept these techniques as innovation when one of us is getting them
to work. Then, suddenly , everyone knows that it must work somehow and
finds a way to adapt it in its own framework.
But there is not only innovation by introducing new techniques, also
refinements can be ( must be ) considered as innovative work imo.
As all you chess programmers out there know: the devil is in the details.
Of course there will never be a discussion about the alphabeta vs minimax
algorithm. From implementation view it is a tiny step, the idea is enormous.
But 300 elo is the sum of dozens little improvements which can also be
the result of innovative work.
Where is the line between improvement and innovation exactly ?
Isnt the one thing not the other at _some_ points?
eg: having a staged move generation is one thing, but to find some groups
that do not only differ between captures and quiet moves _can_ be
innovative work. Putting the puzzle together, from introducing gain
move statistic into an existing framework (staged mg scheme) is innovation too,
especially if such a statistic wasnt meant to use it for this purpose.
So, combination of already existing ideas is innovation too of course.
B:
===
Nonetheless, beside dozens of details there is one eyecatcher imo.
(i cannot say it is from ippo or rybka, never saw the source, but
i know it from robbo).
"The search classification"
Yes, we all know about root,cut,all,pv,quies and horizon nodes,
but to design own search functions for all these node types is a great
idea. I speak of idea because it opens my mind to create a complex
search that handles special types of nodes in its special way.
(there is no limit in building any kind of groups, like capture nodes,
known wins, maybe in some years nodes that are handeld with a montecarlo
technique or whatever) and mix them all together to get a better
search result.)
I only know that it is very hard to figure out what is typical for
a node and what is a general node feature, even in the most common node
classification like mentioned above. I am pretty sure that i can
have a compact search in my engine that is as good as a sub-divisioned
search, but there is not the same room for new ideas.
So innovation is what you make out of an idea and you finally need
a successful implementation.
There are small,intermediate and big innovations. The big ones are
already explored by the pioneers of chess programming. So our task
is to sum up the small ones with hard work. Fortunatelly, the small
ones are making the big picture, LOL (
)
Michael
