Does adapting piece values to the variant work?

[HGMuller]

I am not yet familiar enough with the code to say with certainty this is a problem. But in the init function of psqt.cpp the piece values get adapted to the variant, e.g. for the board size. I would have expected that these recalculated values are then assigned back to PieceValue. But I cannot find that anywhere. They are assigned to CapturePieceValue and EvalPieceValue (for both colors!). The former is only used in SEE, for the latter I could not find any usage at all! This is a bit suspect. Shouldn't the modified values have been assigned to PieceValue instead of EvalPieceValue? That would seem a more natural solution to the bug you just fixed. Why recalculate all these values, and then do not anything with them?

[ianfab]

The PieceValue entries are very sensitive for the standard chess pieces, therefore I was very hesitant to dynamically change them. There is a huge number of places of the code that rely on the constants PawnValueMg, etc. all equal their respective entry from the PieceValue array. If they don't then endgame, eval, and SEE logic breaks.

In order to be able to start using the adjusted piece values in more and more places without being forced to do a huge single switchover I at some point introduced these additional piece values to be able to test them in some places. However, because it affects basically all games, testing is very difficult. There is also still some old branch lying around from when I was testing that, but I never arrived at a conclusion. ac31744