Remove log depth reduction terms.

Motivated by the elo loss of last VVLTC regression test i remove the new log
depth reductions (tuned at STC) to regain strength at VVLTC. The constant terms
are adjusted based on the old values and the changes from the last added VVLTC
tuning.

Passed VVLTC with STC bound:
LLR: 2.95 (-2.94,2.94) <0.00,2.00>
Total: 53802 W: 14030 L: 13740 D: 26032
Ptnml(0-2): 5, 4924, 16754, 5212, 6
https://tests.stockfishchess.org/tests/view/68a9a9f575da51a345a5a675

Passed VVLTC with LTC bound:
LLR: 2.94 (-2.94,2.94) <0.50,2.50>
Total: 53658 W: 14022 L: 13699 D: 25937
Ptnml(0-2): 3, 4894, 16712, 5217, 3
https://tests.stockfishchess.org/tests/view/68a8d2b2b6fb3300203bca77

closes https://github.com/official-stockfish/Stockfish/pull/6257

Bench: 2566780
This commit is contained in:
Stefan Geschwentner
2025-08-24 19:30:31 +02:00
committed by Joost VandeVondele
parent 7fe46b5a70
commit d11f49b790
+8 -8
View File
@@ -1177,27 +1177,27 @@ moves_loop: // When in check, search starts here
// These reduction adjustments have no proven non-linear scaling
r += 700 - 6 * msb(depth); // Base reduction offset to compensate for other tweaks
r -= moveCount * (64 - 2 * msb(depth));
r += 671; // Base reduction offset to compensate for other tweaks
r -= moveCount * 66;
r -= std::abs(correctionValue) / 30450;
// Increase reduction for cut nodes
if (cutNode)
r += 3092 + 2 * msb(depth) + (980 + 15 * msb(depth)) * !ttData.move;
r += 3094 + 1056 * !ttData.move;
// Increase reduction if ttMove is a capture
if (ttCapture)
r += 1467 - 40 * msb(depth);
r += 1415;
// Increase reduction if next ply has a lot of fail high
if ((ss + 1)->cutoffCnt > 2)
r += 1041 + 34 * msb(depth) + allNode * (752 + 226 * msb(depth));
r += 1051 + allNode * 814;
r += (ss + 1)->quietMoveStreak * 50;
// For first picked move (ttMove) reduce reduction
if (move == ttData.move)
r -= 2096 + 27 * msb(depth);
r -= 2018;
if (capture)
ss->statScore = 803 * int(PieceValue[pos.captured_piece()]) / 128
@@ -1208,7 +1208,7 @@ moves_loop: // When in check, search starts here
+ (*contHist[1])[movedPiece][move.to_sq()];
// Decrease/increase reduction for moves with a good/bad history
r -= ss->statScore * (734 - 12 * msb(depth)) / 8192;
r -= ss->statScore * 794 / 8192;
// Step 17. Late moves reduction / extension (LMR)
if (depth >= 2 && moveCount > 1)
@@ -1250,7 +1250,7 @@ moves_loop: // When in check, search starts here
{
// Increase reduction if ttMove is not present
if (!ttData.move)
r += 1178 + 35 * msb(depth);
r += 1118;
if (depth < 5)
r += 1080;