Re: [gnugo-devel] connection tuning

[nando]

Hi,

I've noticed that my patch was not good enough (see new testcase
connection:112). So I revised it a bit. Breakage unchanged besides the PASS
in the new testcase and performance impact even more negligible.

nando

- connection tuning

Index: engine/readconnect.c
===================================================================
RCS file: /cvsroot/gnugo/gnugo/engine/readconnect.c,v
retrieving revision 1.77
diff -u -r1.77 readconnect.c
--- engine/readconnect.c        29 Apr 2004 23:59:21 -0000      1.77
+++ engine/readconnect.c        16 May 2004 15:07:11 -0000
@@ -2525,6 +2525,7 @@
   for (r = 0; r < num_moves; r++) {
     int move = moves[r];
     int adjacent_to_attacker = 0;
+    int bonus_given = 0;
 
     for (k = 0; k < 4; k++) {
       int pos = move + delta[k];
@@ -2537,12 +2538,14 @@
          distances[r] -= 0.2;
          if (verbose > 0)
            gprintf("%o%1M -0.2, adjacent to attacker string with at most two 
liberties\n", move);
-         if ((conn1->distances[move] - conn1->deltas[move] <= 0.5
-              || conn1->distances[pos] - conn1->deltas[pos] <= 0.5)
+         if ((connect_move || !bonus_given)
+             && (conn1->distances[move] - conn1->deltas[move] <= 0.5
+                 || conn1->distances[pos] - conn1->deltas[pos] <= 0.5)
              && (conn2->distances[move] - conn2->deltas[move] <= 0.5
                  || conn2->distances[pos] - conn2->deltas[pos] <= 0.5)
              && conn1->distances[pos] < total_distance
              && conn2->distances[pos] < total_distance) {
+           bonus_given = 1;
            distances[r] -= 0.7;
            if (verbose > 0)
              gprintf("%o%1M -0.7, capture or atari of immediately connecting 
string\n", move);
@@ -2555,10 +2558,29 @@
          if (verbose > 0)
            gprintf("%o%1M -0.2, adjacent to defender string with at most two 
liberties\n", move);
        }
+       /* The code above (in the 'board[pos] == other' branch) makes
+        * perfect sense for the defender, but has a tendency to
+        * overestimate solid connection defenses when the attacker's
+        * stones happen to be in atari, specially when capturing some
+        * defender stones instead would help just as well, if not better.
+        * The following code compensates in such kind of situations.
+        * See connection:111 and gunnar:53 for example.
+        */
+       if (!connect_move && countlib(pos) == 1
+           /* let's avoid ko and snapbacks */
+           && accuratelib(move, other, 2, NULL) > 1) {
+         int adjs[MAXCHAIN];
+         float bonus;
+         bonus = 0.1 * chainlinks2(pos, adjs, 2);
+         bonus += 0.5 * chainlinks2(pos, adjs, 1);
+         distances[r] -= bonus;
+         if (verbose > 0)
+           gprintf("%o%1M -%f, capture of defender string\n", move, bonus);
+       }
       }
     }
     if (adjacent_to_attacker
-       && color != color_to_move
+       && !connect_move
        && is_edge_vertex(move)) {
       distances[r] -= 0.1;
       if (verbose > 0)
Index: regression/connection.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/connection.tst,v
retrieving revision 1.73
diff -u -r1.73 connection.tst
--- regression/connection.tst   23 Apr 2004 16:11:57 -0000      1.73
+++ regression/connection.tst   16 May 2004 15:07:23 -0000
@@ -391,6 +391,13 @@
 popgo
 popgo
 
+# See also nngs1:11
+loadsgf games/nngs/camel-gnugo-3.1.26-200203022335.sgf 104
+trymove B S6
+112 connect S3 S6
+#? [0]*
+popgo
+
 # Report number of nodes visited by the tactical reading
 10000 get_reading_node_counter
 #? [0]&