Fix picker issues

src/search/movepick.rs

@@ -8,7 +8,6 @@ use rand::{thread_rng, Rng};
 use std::sync::Arc;
 
 type ScoredVec<Move> = Vec<(Move, ValueScore)>;
-type PickResult = (Move, ValueScore);
 
 const RANDOM_FACTOR: ValueScore = 1000;
 
@@ -40,7 +39,7 @@ impl MovePicker<true> {
 }
 
 impl std::iter::Iterator for MovePicker<true> {
-    type Item = PickResult;
+    type Item = Move;
 
     fn next(&mut self) -> Option<Self::Item> {
         find_next_max_and_swap(&mut self.moves, &mut self.index)
@@ -75,11 +74,11 @@ impl MovePicker<false> {
 }
 
 impl std::iter::Iterator for MovePicker<false> {
-    type Item = PickResult;
+    type Item = Move;
 
     fn next(&mut self) -> Option<Self::Item> {
-        if let Some((mov, score)) = find_next_max_and_swap(&mut self.moves, &mut self.index) {
-            return Some((mov, score));
+        if let Some(mov) = find_next_max_and_swap(&mut self.moves, &mut self.index) {
+            return Some(mov);
         }
 
         match self.stage {
@@ -121,7 +120,7 @@ where
     moves.iter().map(|mov| (*mov, f(*mov))).collect()
 }
 
-fn find_next_max_and_swap(moves: &mut ScoredVec<Move>, index: &mut usize) -> Option<PickResult> {
+fn find_next_max_and_swap(moves: &mut ScoredVec<Move>, index: &mut usize) -> Option<Move> {
     if *index >= moves.len() {
         return None;
     }
@@ -136,5 +135,5 @@ fn find_next_max_and_swap(moves: &mut ScoredVec<Move>, index: &mut usize) -> Opt
     }
 
     *index += 1;
-    Some((moves[*index - 1].0, moves[*index - 1].1))
+    Some(moves[*index - 1].0)
 }

src/search/pvs.rs

@@ -61,19 +61,19 @@ fn quiesce(
 
     let mut count = 1;
 
-    for (mov, _) in picker {
-        // Delta prune move if it cannot improve the score
+    for mov in picker {
         if !is_check && mov.flag().is_capture() {
+            // Delta pruning: if there is no way this capture can improve the score, prune immediately.
             let captured_piece =
                 position.board.piece_color_at(mov.to()).map_or_else(|| Piece::Pawn, |p| p.0);
             if static_evaluation + captured_piece.value() + MAX_POSITIONAL_GAIN < alpha {
                 continue;
             }
-        }
 
-        // Static exchange evaluation: if we lose material, there is no point in searching further.
-        if see::see(mov, &position.board) < 0 {
-            continue;
+            // Static exchange evaluation: if we lose material, there is no point in searching further.
+            if see::see(mov, &position.board) < 0 {
+                continue;
+            }
         }
 
         let (score, nodes) = quiesce(&position.make_move(mov), -beta, -alpha, constraint);
@@ -258,14 +258,14 @@ fn pvs<const ROOT: bool, const MAIN_THREAD: bool, const ALLOW_NMR: bool>(
     // We need to keep track of the original alpha and best moves, to store
     // the correct node type and move in the hash table later.
     let original_alpha = alpha;
-    let mut best_move = picker.peek().map(|(mov, _)| *mov).unwrap();
+    let mut best_move = picker.peek().copied().unwrap();
 
     // Check extension: we are interested in exploring the outcome of this properly.
     if is_check {
         depth = depth.saturating_add(1).min(MAX_DEPTH);
     }
 
-    for (i, (mov, _)) in picker.enumerate() {
+    for (i, mov) in picker.enumerate() {
         // Extended futility pruning: discard moves without potential
         if depth <= 2 && i > 0 && !may_be_zug {
             let move_potential = MAX_POSITIONAL_GAIN * depth as ValueScore

src/search/see.rs

@@ -15,6 +15,8 @@ fn least_valuable(bb: Bitboard, board: &Board) -> Option<(Piece, Square)> {
 }
 
 pub fn see(mov: Move, board: &Board) -> ValueScore {
+    debug_assert!(mov.flag().is_capture());
+
     let mut board = *board;
 
     // Store the "best score" when it is our turn to move.