opelly/Stockfish
1
0
Fork 0
mirror of https://github.com/opelly27/Stockfish.git synced 2026-05-20 09:47:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

Futility pruning simplification

Browse Source 
1/ eval margin and gains removed:
16bit are now free on TT entries, due to the removal of eval margin. may be useful
in the future :) gains removed: use instead by Value(128). search() and qsearch()
are now consistent in this regard.

2/ futility_margin()
linear formula instead of complex (log(depth), movecount) formula.

3/ unify pre & post futility pruning
pre futility pruning used depth < 7 plies, while post futility pruning used
depth < 4 plies. Now it's always depth < 7.

Tested with fixed number of games both at short TC:
ELO: 0.82 +-2.1 (95%) LOS: 77.3%
Total: 40000 W: 7939 L: 7845 D: 24216

And long TC
ELO: 0.59 +-2.0 (95%) LOS: 71.9%
Total: 40000 W: 6876 L: 6808 D: 26316

bench 7243575
This commit is contained in:
Lucas Braesch
2013-11-08 18:42:22 +08:00
committed by Marco Costalba
parent 343544f3f7
commit eed508b444
7 changed files with 59 additions and 116 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/evaluate.cpp
+11 -27
View File
@@ -229,7 +229,7 @@ namespace {
// Function prototypes
template<bool Trace>
Value do_evaluate(const Position& pos, Value& margin);
Value do_evaluate(const Position& pos);
template<Color Us>
void init_eval_info(const Position& pos, EvalInfo& ei);
@@ -238,7 +238,7 @@ namespace {
Score evaluate_pieces_of_color(const Position& pos, EvalInfo& ei, Score* mobility);
template<Color Us, bool Trace>
Score evaluate_king(const Position& pos, const EvalInfo& ei, Value margins[]);
Score evaluate_king(const Position& pos, const EvalInfo& ei);
template<Color Us, bool Trace>
Score evaluate_threats(const Position& pos, const EvalInfo& ei);
@@ -264,8 +264,8 @@ namespace Eval {
/// values, an endgame score and a middle game score, and interpolates
/// between them based on the remaining material.
Value evaluate(const Position& pos, Value& margin) {
return do_evaluate<false>(pos, margin);
Value evaluate(const Position& pos) {
return do_evaluate<false>(pos);
}
@@ -307,19 +307,14 @@ namespace Eval {
namespace {
template<bool Trace>
Value do_evaluate(const Position& pos, Value& margin) {
Value do_evaluate(const Position& pos) {
assert(!pos.checkers());
EvalInfo ei;
Value margins[COLOR_NB];
Score score, mobility[2] = { SCORE_ZERO, SCORE_ZERO };
Thread* th = pos.this_thread();
// margins[] store the uncertainty estimation of position's evaluation
// that typically is used by the search for pruning decisions.
margins[WHITE] = margins[BLACK] = VALUE_ZERO;
// Initialize score by reading the incrementally updated scores included
// in the position object (material + piece square tables) and adding
// Tempo bonus. Score is computed from the point of view of white.
@@ -332,10 +327,7 @@ Value do_evaluate(const Position& pos, Value& margin) {
// If we have a specialized evaluation function for the current material
// configuration, call it and return.
if (ei.mi->specialized_eval_exists())
{
margin = VALUE_ZERO;
return ei.mi->evaluate(pos);
}
// Probe the pawn hash table
ei.pi = Pawns::probe(pos, th->pawnsTable);
@@ -353,8 +345,8 @@ Value do_evaluate(const Position& pos, Value& margin) {
// Evaluate kings after all other pieces because we need complete attack
// information when computing the king safety evaluation.
score += evaluate_king<WHITE, Trace>(pos, ei, margins)
- evaluate_king<BLACK, Trace>(pos, ei, margins);
score += evaluate_king<WHITE, Trace>(pos, ei)
- evaluate_king<BLACK, Trace>(pos, ei);
// Evaluate tactical threats, we need full attack information including king
score += evaluate_threats<WHITE, Trace>(pos, ei)
@@ -401,7 +393,6 @@ Value do_evaluate(const Position& pos, Value& margin) {
sf = ScaleFactor(50);
}
margin = margins[pos.side_to_move()];
Value v = interpolate(score, ei.mi->game_phase(), sf);
// In case of tracing add all single evaluation contributions for both white and black
@@ -414,9 +405,7 @@ Value do_evaluate(const Position& pos, Value& margin) {
Score b = ei.mi->space_weight() * evaluate_space<BLACK>(pos, ei);
Tracing::add(SPACE, apply_weight(w, Weights[Space]), apply_weight(b, Weights[Space]));
Tracing::add(TOTAL, score);
Tracing::stream << "\nUncertainty margin: White: " << to_cp(margins[WHITE])
<< ", Black: " << to_cp(margins[BLACK])
<< "\nScaling: " << std::noshowpos
Tracing::stream << "\nScaling: " << std::noshowpos
<< std::setw(6) << 100.0 * ei.mi->game_phase() / 128.0 << "% MG, "
<< std::setw(6) << 100.0 * (1.0 - ei.mi->game_phase() / 128.0) << "% * "
<< std::setw(6) << (100.0 * sf) / SCALE_FACTOR_NORMAL << "% EG.\n"
@@ -640,7 +629,7 @@ Value do_evaluate(const Position& pos, Value& margin) {
// evaluate_king() assigns bonuses and penalties to a king of a given color
template<Color Us, bool Trace>
Score evaluate_king(const Position& pos, const EvalInfo& ei, Value margins[]) {
Score evaluate_king(const Position& pos, const EvalInfo& ei) {
const Color Them = (Us == WHITE ? BLACK : WHITE);
@@ -735,12 +724,8 @@ Value do_evaluate(const Position& pos, Value& margin) {
attackUnits = std::min(99, std::max(0, attackUnits));
// Finally, extract the king danger score from the KingDanger[]
// array and subtract the score from evaluation. Set also margins[]
// value that will be used for pruning because this value can sometimes
// be very big, and so capturing a single attacking piece can therefore
// result in a score change far bigger than the value of the captured piece.
// array and subtract the score from evaluation.
score -= KingDanger[Us == Search::RootColor][attackUnits];
margins[Us] += mg_value(KingDanger[Us == Search::RootColor][attackUnits]);
}
if (Trace)
@@ -1024,8 +1009,7 @@ Value do_evaluate(const Position& pos, Value& margin) {
stream << std::showpoint << std::showpos << std::fixed << std::setprecision(2);
std::memset(scores, 0, 2 * (TOTAL + 1) * sizeof(Score));
Value margin;
do_evaluate<true>(pos, margin);
do_evaluate<true>(pos);
std::string totals = stream.str();
stream.str("");