From 4e4768d96b8944c7c2f55e020f789cb53de4c9e9 Mon Sep 17 00:00:00 2001 From: Stefan Geschwentner Date: Mon, 20 Jan 2025 15:40:41 +0100 Subject: [PATCH] Consider more nodes as ttPv nodes. Remove depth condition in propagation rule for ttPv state from a node to it childs. Because this change marks more nodes as ttPv, we have a time sensitive ttPv reduction rule and the STC snd LTC seems to show bad scaling. So i have also submitted a VLTC non-regression to check the scaling at higher time control. The results gives a little indication that we have perhaps good scaling with more ttPv nodes so that could be further explored. Passed non-regression STC: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 82528 W: 21627 L: 21453 D: 39448 Ptnml(0-2): 317, 9809, 20891, 9877, 370 https://tests.stockfishchess.org/tests/view/678e608cd63764e34db49ad7 Passed non-regression LTC: LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 310440 W: 78879 L: 78956 D: 152605 Ptnml(0-2): 255, 34915, 84938, 34876, 236 https://tests.stockfishchess.org/tests/view/678fab89ac8f8f5496155f3c Passed non-regression VLTC for scaling verification: LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 59496 W: 15158 L: 14983 D: 29355 Ptnml(0-2): 15, 6039, 17470, 6204, 20 https://tests.stockfishchess.org/tests/view/6794bd1f4f7de645171fb33b Bench: 1498882 --- src/search.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/search.cpp b/src/search.cpp index e99d5d3fc9f..d6dc29cf06d 100644 --- a/src/search.cpp +++ b/src/search.cpp @@ -1421,7 +1421,7 @@ Value Search::Worker::search( // If no good move is found and the previous position was ttPv, then the previous // opponent move is probably good and the new position is added to the search tree. (~7 Elo) if (bestValue <= alpha) - ss->ttPv = ss->ttPv || ((ss - 1)->ttPv && depth > 3); + ss->ttPv = ss->ttPv || (ss - 1)->ttPv; // Write gathered information in transposition table. Note that the // static evaluation is saved as it was before correction history.