Author: Ryan B.
Date: 14:18:31 11/15/05
Has anyone tested Gambit Fruit with Rebel Reductions on? It is an idea from Ed's web site that looks safe in theory and causes few researches. I have not tested it so it is off by default and I am not sure if I implemented it exactly as intended. The code is very simple. // rebel reduction rebel_reduction = false; if (use_rebel_reduction && remaining_depth > 2 && node_type != NodePV && !in_check && board->square[MOVE_TO(move)] == Empty && !MOVE_IS_PROMOTE(move)) { if (remaining_depth <= 4) margin = 500; else if (remaining_depth <= 6) margin = 700; else if (remaining_depth <= 8) margin = 975; else margin = 1500; opt_value = board->piece_material[White] - board->piece_material[Black]; if (COLOUR_IS_BLACK(board->turn)){ opt_value = -opt_value; } if (alpha > opt_value + margin) { new_depth -= Fractional; rebel_reductions++; rebel_reduction = true; } } I also added researchs to make sure the reduction would be safe. if (rebel_reduction && value >= beta) { ASSERT(node_type!=NodePV); new_depth += Fractional; rebel_researchs++; value = -full_search(board,-beta,-alpha,new_depth,height+1,new_pv,NODE_OPP(node_type)); } A small idea involving the value of passed pawns that I am not using yet in Gambit Fruit. The idea is the measure the pawn pushing power of each side and scale the value of the passed pawn up or down based on this data. Here is an example of how to count for white. // white pawn push power ppp[White] += WhiteKnight * 2; ppp[White] += WhiteBishop * 2; if (WhiteBishops >= 2) ppp[White]++; ppp[White] += WhiteRook * 3; ppp[White] += WhiteQueen * 4; ppp[White] += DefendingPassedPawns * 3; // very strong to have and hard to stop ppp[White] -= WhiteTrapedPieces; With this data for each side you can scale the score up or down based on ppp[me]-ppp[opp]. This should help with issues like 2 bishops and a passed pawn vs queen or sacrificing a minor for connected passed pawns. Ryan
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.