After doing some debugging, I started to wonder if a node might put into the table an *exact* score when it is done with its searching, yet in the children nodes that very score was not exact. Could someone look through my pseudo-code and perhaps see if there's a problem with the windows, or such?

My bad, I did indeed not write the pseudo-code out completely correctly. Here it is with that problem corrected. Thanks for looking at it!

int maxd;

tt_replace(u64 zobrist, int score, int scoretype, int depth);

void start(Position pos, int d) {
for (maxd=1; maxd<d; d++)
negascout_best(pos, 0 ...

Hi all.

We're having some problems with a chess program we're developing. The program right now uses negascout (fail-soft) with PV move ordering, iterative deepening and an always-replace transposition table. It also uses SEE-ordered quiescence search (fail-hard).

The problem occurs when both ...