On reaching maximum ply

[MartinBryant]

I know this occurs rarely, if ever, but I was intrigued when I noticed that Stockfish and Crafty handle it differently.

In Colossus I've always just returned the static evaluation.

Stockfish has this...
return (ss->ply >= MAX_PLY && !ss->inCheck) ? evaluate(pos) : VALUE_DRAW;

Crafty has this...
return beta;

Thoughts?
Why VALUE_DRAW (rather than the static evaluation) if you're in check?
Why beta?

[Sven]

When you are in check, the static eval is often unreliable so VALUE_DRAW is just "less random". And ignoring the current node by returning beta (which is negated by the parent so it can't raise his alpha) is "even less random".

Just my thoughts, don't know if the original authors had that in mind.

[chrisw]

I know this occurs rarely, if ever, but I was intrigued when I noticed that Stockfish and Crafty handle it differently.

In Colossus I've always just returned the static evaluation.

Stockfish has this...
return (ss->ply >= MAX_PLY && !ss->inCheck) ? evaluate(pos) : VALUE_DRAW;

Crafty has this...
return beta;

Thoughts?
Why VALUE_DRAW (rather than the static evaluation) if you're in check?

SF evaluation function doesn’t handle in-check positions, so I guess they just default to DRAW in the absence of any other v.

Why beta?