Evaluation vs. Search

[Graham Banks]

I don't see the Zappa/Rybka match as something intended to prove what "engine" is better anyway. It has many other factors that could not be reproduced. It is about how good the overall team is in the given conditions.

Exactly.

[bob]

My thoughts have always been pretty consistent:

The difference between selective searching and full-width searching with extensions and reductions is nil. these are two different ways to reach the same result. One program spends more time trying to figure out what to search, the other program spends more time searching more stuff. It tends toward equality, with the simpler search getting the nod IMHO (and that would _not_ be the selective version which gets too complex).

Ditto for eval vs search. You can either go with a simpler eval, and a more complex search, or vice-versa. Or a hybrid. I've sort of taken the middle of the road here. Some things are convenient to find with search (tactics for one). Some things are easier to find via evaluation (weak squares). Some things respond better to a hybrid approach (king safety for example, where static analysis is good, but deep tactics need to be examined to make sure the static eval is not way off-base.)

So it is not necessarily a matter of "either-or". It becomes more a matter of "how much of each"...

This is a complex game. Playing against another computer is a different thing entirely from playing against a human GM...

[bob]

I don't see the Zappa/Rybka match as something intended to prove what "engine" is better anyway. It has many other factors that could not be reproduced. It is about how good the overall team is in the given conditions.

Exactly.

Actually there is a far larger luck component at play that most realize... that has more to do with the outcome than anything else when two engines are reasonably close in skill and the number of games is this tiny.