This project is mirrored from https://:*****@github.com/mhouppin/stash-bot.
Pull mirroring updated .
-
-
-
-
-
-
-
-
-
-
-
-
-
v13 Release: v13
Version 13 is out! Here is a list of new things: - LMR depth correction (now applied when depth > 3); - More aggressive reductions on NMP when eval is way above beta; - NMP reverified when beta is above 5000 centipawns (zugzwang-mate checking); - Mate Distance Pruning (if the current line is already longer than a previous "mate in x" found, stop analysis); - Correction of "go depth" and "go nodes" command (also "go infinite" is used by default when no parameters are given).
-
v12 Release: v12
Version 12 released with new components: Null Move Pruning, and Late Move Reductions ! For now, I kept the code very simple: - a Null Move Search is done if depth > 1 and the last null move is at least (ply * 2) distant from current node, to avoid consecutive null moves to abort search too quickly. Depth is reduced by 3, and a recheck with no null move is done with same reduction if depth was greater than 9 (zugzwang checking). - a Late Move Reduction is applied when at least 3 legal moves have been analysed with no fail-high encountered. All subsequent nodes are analysed 1 ply less deeper.
-
v11 Release: v11
Version 11 comes out, with a new component: a Transposition Table! For now, previous search entries are only used as a help for move ordering, as I got problems with 3-fold detection on the opponent moves when the entries were not refreshed.
-
v10 Release: v10
Version 10 is out, with a new component to analyse: Quiescence Search ! The evaluation scores are now completely stabilized, which will allow for more cool things to be correctly implemented (such as a Transposition Table and a Null Move Search). An initiative bonus (15 centipawns) has also been given to the side to move on evaluation.
-
-
-
-