talk about IPP's evaluation

[liuzy]

Can anyone explain the evaluation of IPP ?

[bob]

Can anyone explain the evaluation of IPP ?

It looks at the static position of the pieces on the board, and converts this to a numeric number that approximates the goodness or badness of the overall position.

[rvida]

Can anyone explain the evaluation of IPP ?

What exactly do you want to know? It's just the usual stuff... Pawn structure, piece mobility, king safety, etc...
It's not the most readable code on the planet. For studying evaluation function I would recommend more sane (human readable) sources like Stockfish, Crafty.

[Dann Corbit]

Can anyone explain the evaluation of IPP ?

The one thing it has that is new is material imbalance.
Read Kaufman's paper to understand that.

Consider (for instance) that an A or an H pawn is not as valuable as the others.

[jarkkop]

It has some terms that are strangely multiplied by 65536(<<16) and added together and then at the final stage divided by same 65536(>>16).
What does this achieve?

[Don]

It has some terms that are strangely multiplied by 65536(<<16) and added together and then at the final stage divided by same 65536(>>16).
What does this achieve?

That could be some mechanism to reduce rounding error while dealing with fractions as integers. That's my guess.

[Karlo Bala]

It has some terms that are strangely multiplied by 65536(<<16) and added together and then at the final stage divided by same 65536(>>16).
What does this achieve?

That could be some mechanism to reduce rounding error while dealing with fractions as integers. That's my guess.

Maybe it deals with opening and endgame value(like Fruit), but both at same time(pack two values into 32bit).

[Don]

It has some terms that are strangely multiplied by 65536(<<16) and added together and then at the final stage divided by same 65536(>>16).
What does this achieve?

That could be some mechanism to reduce rounding error while dealing with fractions as integers. That's my guess.

Maybe it deals with opening and endgame value(like Fruit), but both at same time(pack two values into 32bit).

Yes, that's probably it.

[bob]

It has some terms that are strangely multiplied by 65536(<<16) and added together and then at the final stage divided by same 65536(>>16).
What does this achieve?

That could be some mechanism to reduce rounding error while dealing with fractions as integers. That's my guess.

I think I heard someone mention that it might keep the EG and MG scores combined, 16 bits per value, so that one is left shifted 16 bits. You could then add a constant to both EG and MG scores with one add, even if the values you add are different, since the constant could itself be the results of a shift/add which would still be a constant.

[mcostalba]

I think I heard someone mention that it might keep the EG and MG scores combined, 16 bits per value, so that one is left shifted 16 bits.

Yes you guessed right. It saves both mid game and end game score in a 32 bit integer with a 16 bit word for each score.

At the end the combined score is splitted and the two values merged in the usual way to get the final score.