My two cents on the whole ippolit thing

[adieguez]

Beware of copying without verifying. I have verified my last-4-plies pruning code _extensively_. I tried 5 and up and it hurt in every test I contrived. I've not given up, but just because an idea shows up in program X does not guarantee it will work well in program Y. In fact, it does not guarantee it works well in X. I'd bet most every program, IPPOLIT included, has things in it that are detrimental rather than helpful.

agreed agreed, but I don't test that extensively. I confess I play less games than you . And sometimes test more that one thing at the same time.

[Ryan Benitez]

First some history and my opinions pasted to it:
The chess engine Rybka came onto the scene with version 1.0 and it was a stunning world beater.
Sometime later, some enterprising souls did some reverse engineering of Rybka and came to the conclusion that it had striking similarities to fruit.
Here are some links to the analysis:
{analysis links removed}

The inescapable conclusion is that Rybka 1.0 is heavily influenced by Fruit.

wow that's interesting stuff. I don't know if I remember well, but I think before "Rybka 1.0", Vas had already a weaker version right? so he later changed the evaluation then maybe? why would one semicopy an evaluation? including things that may look caprichous, I don't think that's even something that's going to give you too much elo and feels kind of cheap. Was changing the eval an elo-boost? is that Fruit eval specially strong at all?? and would an international master prefer to base heavily its eval in another eval instead of try to write it by himself?

Actually I don't want to "acuse" no one of nothing. I wrote just with my first impression. Taking that aside, I repeat it is a bit strange, coming from an international master one would expect an eval much more complex than an average one. But it's ok. I remember also a webpage where the author did a theory about rybka strength and talk about the better calculation of the potential of the pieces or something like that, pure ilusion?.

Rybka was a lot stronger than Fruit, right off the bat.
Rybka is a bitboard engine so the architecture is different than fruit which does not use bitboards.
Rybka uses the idea of material imbalance in the evaluation as pioneered by Larry Kaufman.

We must not have the same understanding of architecture because many of the architecture differences between Rybka and Fruit seem irrelevant to the use of bitboards to me. Also I can't help but be be annoyed by the myth floating around that board rep changes require a large architecture change or even a rewite of the engine. Converting Fruit to being a bitboard engine is a minor change that is just a few hours of work.

I totally disagree that it is only a few hours of work to convert fruit to bitboard engine because only generating move generator that is based on bitboards with no bugs(part of the work) is clearly more than a few hours of work.

Uri

Most of the time is in debugging and having the old board rep in place until the conversion is done speeds up debug time by a lot. Also the second time you do something is much faster than the first. It was not my first bitboard move gen. Yes I agree that the first board rep in a new engine takes far more than a few hours to crate and debug.