diff options
| author | tslil clingman <tslil@posteo.de> | 2023-01-19 22:40:24 +0100 |
|---|---|---|
| committer | tslil <tslil@posteo.de> | 2026-08-28 19:37:41 +0100 |
| commit | cb2b78ced27fc7996ed11c3450f69138d7d1b61f (patch) | |
| tree | 8d8e00e9125a216184127ee56224e1b68660f791 | |
| parent | 0223a9bec5535fced1a7698b55fd42155d9b0446 (diff) | |
Change transposition table lookup policy
Given the approximate nature of the evaluations and truncated tree
searches, the result of negamax will always be sensitive to the
particulars of the depth bounding and the conditions for referring to
precomputed values.
The tradeoff here is a slight performance penalty (~200ms at depth 6
on my old Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz), but a greatly
increased opponent strength as compared to `>=`, and the old
convolutional network (flawed as the implementation was).
| -rw-r--r-- | include/negamax.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/include/negamax.c b/include/negamax.c index 0f2c450..0ec40e5 100644 --- a/include/negamax.c +++ b/include/negamax.c @@ -83,7 +83,7 @@ static float negamax(tak_state_p state, const uint8_t cur_depth, tt_entry_t *entry = tt_seek(hash); // CAUTION: ≥ breaks search stability (vs =) on shallow depths - if (entry != NULL && entry->depth >= cur_depth) { + if (entry != NULL && entry->depth == cur_depth) { if (entry->flag == TT_EXACT) { return entry->value; } else if (entry->flag == TT_LOWERBOUND && entry->value > alpha) { |