Fix doubled pawns asymmetry

When evaluating double pawns we use always
lsb() to extract the frontmost square.

This breaks evaluation color symmetry as is
possible to verify with an instrumented evaluate()

  Value evaluate(const Position& pos) {

    Value v = do_evaluate<false>(pos);
    Position p = pos;
    p.flip();
    assert(v == do_evaluate<false>(p));
    return v;
  }

Passed no regression test:

STC
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 21035 W: 4244 L: 4122 D: 12669

LTC
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 39839 W: 6662 L: 6572 D: 26605

bench: 8255966

diff --git a/src/pawns.cpp b/src/pawns.cpp
index 1099c5a48a67a9e5f15f1526ce720c299800c06a..fe989e9b8f24a502f7febb083f4b6733beae5d8b 100644 (file)
--- a/src/pawns.cpp
+++ b/src/pawns.cpp
@@ -165,7 +165,7 @@ namespace {
             value -= UnsupportedPawnPenalty;
 
         if (doubled)
-            value -= Doubled[f] / distance<Rank>(s, Square(lsb(doubled)));
+            value -= Doubled[f] / distance<Rank>(s, frontmost_sq(Us, doubled));
 
         if (backward)
             value -= Backward[opposed][f];