Kerry Handscomb  vs  Debby (Mona + YL)           April 6-30, 2000

   1. d1-b3 h7-f7   2. g8-g6 h4-f2   3. b8-b5 h2-e2   4. c8:a6 f7-c4
   5. b1-b4 a7-d4   6. f8-c5 a5-c7   7. c1:a3 h5-f3   8. e1-c3 h3:c3
   9. g6-d6 f3:b3  10. a6-b7 a4-c2  11. a3-e7 h6-f6  12. e8-c6 a2-d5
  13. d8-e8 d4-d1  14. Resigns 0-1

  Move by Move:

 Note that Kerry's comments are taken from the e-mail messages
 during the game, without the benefit of hindsight.  This allows
 some insight into how a top-level Lines of Action player thinks.

      +-----------------+
    8 | . B B B B B B . |
    7 | W . . . . . . W |
    6 | W . . . . . . W |
    5 | W . . . . . . W |
    4 | W . . . . . . W |
    3 | W . . . . . . W |
    2 | W . . . . . . W |
    1 | . B B B B B B . |
      +-----------------+
        a b c d e f g h	 

    1. d1-b3 h7-f7

 Darse:  The move h7-f7 was made shortly after the introduction of a
 new evaluation procedure, and is the product of a wildly paranoid bias
 against getting blockaded.  Having moved past the testing stage, I've
 now removed this big bias, and h7-f7 is no longer the move selected.
 It now prefers to make its own blockade with h4-f2, just as Yngvi's
 program did, before recently changing to a5-c7.  So h7-f7 might be a
 little inconsistent with the new approach, but it probably didn't do
 much harm, either.

      +-----------------+
    8 | . B B B B B B . |
    7 | W . . . . W . . |
    6 | W . . . . . . W |
    5 | W . . . . . . W |
    4 | W . . . . . . W |
    3 | W B . . . . . W |
    2 | W . . . . . . W |
    1 | . B B . B B B . |
      +-----------------+
        a b c d e f g h  

    2. g8-g6 h4-f2

 Darse:  We expected this move because of Kerry's article in the first
 issue of "Abstract Games" magazine.  He also mentions 2. b3:f7 as a
 more dynamic approach.  The programs don't mind b3:f7 at all, thinking
 it is close to equality, but their judgement is rather suspect at this
 stage of the game.

 2... h4-f2 might be playing into the strength of a good human player
 (a strategic position rather than tactical), but that will give us the
 opportunity to learn more about this aspect of the game.

      +-----------------+
    8 | . B B B B B . . |
    7 | W . . . . W . . |
    6 | W . . . . . B W |
    5 | W . . . . . . W |
    4 | W . . . . . . . |
    3 | W B . . . . . W |
    2 | W . . . . W . W |
    1 | . B B . B B B . |
      +-----------------+
        a b c d e f g h  

    3. b8-b5 h2-e2

    depth value best  principle variation (last 4-ply deleted)

 YL:   14  -33  h2e2 b1b4 a7d4 f8c5 a2d2 e8e5 a5c7 g1g3 f7d7 c8e8
 Mona: 11  -28  h2-e2 c8-c6 a2-d2 b1-b4 a3-a8 e8-e5 a8:c6

      +-----------------+
    8 | . . B B B B . . |
    7 | W . . . . W . . |
    6 | W . . . . . B W |
    5 | W B . . . . . W |
    4 | W . . . . . . . |
    3 | W B . . . . . W |
    2 | W . . . W W . . |
    1 | . B B . B B B . |
      +-----------------+
        a b c d e f g h	 

    4. c8:a6 f7-c4

 Kerry:  Your move is what I expected.  I vacillated between this
 move and b1b4, but chose this because it's more agressive.  Perhaps
 I should have continued to close the game down with b1b4, but then
 I was concerned about h3:f1.

 YL:   13   +3  f7c4 b1b4 a7d4 e8e5 h3e3 e1a1 h6h8 c1e3 a5c7
 Mona: 11 -106  f7-c4 b1-b4 h6:f8 a6-c6 h5-e5 b5:e5 a5-c5
 Mona: 12  -52  f7-c4 b1-b4 a2-d2 c1-c3 h5-f3 f1:h3 a7-d4 f1-c1

 alternate choice (different versions of Mona):
 11  -82  h3-e3 g1-g3 e2-b2 a6-d3 a7-d4 b1-e4 h5-e5
 11   -5  h3-e3 e1-c3 a3:c1 g1:c1 a7-d4 b1-b4 a2:a6

 Darse:  The two candidates were f7-c4 and h3-e3, and they had virtually
 equal averages.  Since the latter seemed to get into messier tactical
 positions, and we were hoping to learn something about strategy, we
 elected to go with the more positional looking move, 3... f7-c4.

      +-----------------+
    8 | . . . B B B . . |
    7 | W . . . . . . . |
    6 | B . . . . . B W |
    5 | W B . . . . . W |
    4 | W . W . . . . . |
    3 | W B . . . . . W |
    2 | W . . . W W . . |
    1 | . B B . B B B . |
      +-----------------+
        a b c d e f g h  

    5. b1-b4 a7-d4

 Kerry:  This seems like the obvious move to me.  Now I'm looking at
 f8-c5 (threatening c1:c4), followed by d8-b6.  Also now if a3:c1,
 g1:c1 is a good move.

 YL:   14  -22  a7d4 f8c5 a5c7 a6d6 h3e3 e1c3 c7d7 c5f5 d7c7 f1g2
 Mona: 13 -106  a7-d4 f8-c5 h3-e3 g1-g3 a3:c1 f1:c1 e3:b3 c1:c4 h6-g5

 alternate choices (different versions of Mona):
 13  -77  a3:c1 g1:c1 a2-d2 d8-d6 h6:f8 g6-g5 d2:g5 a6-c6 a7-c5
 11  -73  h6:f8 g1-g3 a2-d2 c1:a3 a7-c5 d8-d6 h3-g4
 11 -112  a7-d4 f8-c5 h3-e3 c1:c4 e3:b3 e8-e5 h6-g5

 Darse:  For the first time in these e-mail games, we had more than
 two candidate moves, because the best move for one program wasn't
 the second best for the other, and vice-versa.  In this case it was
 more like first and third, with the programs agreeing on the second
 best.  The other moves were 5... h6:f8 and 5... a3:c1 (with the reply
 g1:c1, as Kerry suggested).  All of the averages were quite close,
 making me wonder if they even have a clue about what's going on. :)

      +-----------------+
    8 | . . . B B B . . |
    7 | . . . . . . . . |
    6 | B . . . . . B W |
    5 | W B . . . . . W |
    4 | W B W W . . . . |
    3 | W B . . . . . W |
    2 | W . . . W W . . |
    1 | . . B . B B B . |
      +-----------------+
        a b c d e f g h  

    6. f8-c5 a5-c7

 Kerry:  Seems to be the natural move.  I would estimate I have the
 advantage at this stage.

 YL:   15  -18  a5c7 a6d6 a3c1 g1c1 f2c5 c1b2 d4b2 g6g5 c5g5 e8e5 a2d2
 Mona: 13  -91  a5-c7 e1-c3 a3:c1 e8:h5 a4-a1 f1:h3 h6-f4 b4-d2 c7-e5

 alternate choice:
 13  -93  h3-e3 g1-g3 a3:c1 f1:c1 e3:b3 c1:c4 f2:c5 g3-g5 a5-c7

 Darse:  The programs agree that Black has an advantage, but are slow
 to appreciate the full extent of the problems.  Without any explicit 
 knowledge, it is hard for them to see what's coming.

 Mona prefered h3-e3 up until the last iteration, aiming for e3:b3.
 The move played appears to get into some boot ugly positions (the
 kind of positions that only a computer could love:), but at least
 they look a little complicated.

      +-----------------+
    8 | . . . B B . . . |
    7 | . . W . . . . . |
    6 | B . . . . . B W |
    5 | . B B . . . . W |
    4 | W B W W . . . . |
    3 | W B . . . . . W |
    2 | W . . . W W . . |
    1 | . . B . B B B . |
      +-----------------+
        a b c d e f g h	 

    7. c1:a3 h5-f3

 Kerry:  It's now impossible for the piece on a4 to escape -- Debby needs
 to attack my formation from the outside.  I think I can defend, unless
 I've missed something.  It will be interesting.  I also considered e1c3
 and c1e3, but I think the move I chose will give a more interesting game.

 Darse:  The programs considered 7. e1-c3 to be the main line, but
 trapping the piece on a4 is precisely the kind of thing they do not
 understand very well.

 YL:   15  -25  h5f3 a3b2 d4b2 a6d6 f2c5 g1e3 f3f5 e3c5 f5c5 e8e5 a2d2
 Mona: 13  -42  h5-f3 a3-d6 d4-c3 d6-a3 h3-g4 a3-b2 h6-g5 c5:f2 a4-a1
 Mona: 13  -49  h5-f3 a3-b2 d4-d2 a6-d6 h3-e3 d6-d3 a4-a6 g6-e4 c7-d6

 alternate choice:
 YL:   15   -2  a2d2 e8h5 h6f4 a3d6 c7d7 h5e5 a4a6 g6f7 c4g4 f7d7 h3e6
 Mona: 13  -72  a2-d2 e8:h5 h6-f4 d8-d5 h3:h5 g6-f5 c7-e5 g1-d1 h5-h4

      +-----------------+
    8 | . . . B B . . . |
    7 | . . W . . . . . |
    6 | B . . . . . B W |
    5 | . B B . . . . . |
    4 | W B W W . . . . |
    3 | B B . . . W . W |
    2 | W . . . W W . . |
    1 | . . . . B B B . |
      +-----------------+
        a b c d e f g h  

    8. e1-c3 h3:c3

 Kerry:  I would like to move g1g3, but then h3g4 is a killer.

 Darse:  The programs agree that h3-g4 is strong for White.  However,
 in their opinion, not much changes after giving up this piece, so it
 is considered to be a concession, reducing Black's advantage.

 YL chooses h3:c3 at all depths through to 14-ply (output unavailable)
 Mona: 13  100  h3:c3 a3-b2 c7-d6 d8-e7 a2-b1 e7-f7 b1-d3 b2-a3 h6-g5

      +-----------------+
    8 | . . . B B . . . |
    7 | . . W . . . . . |
    6 | B . . . . . B W |
    5 | . B B . . . . . |
    4 | W B W W . . . . |
    3 | B B W . . W . . |
    2 | W . . . W W . . |
    1 | . . . . . B B . |
      +-----------------+
        a b c d e f g h  

    9. g6-d6 f3:b3

 Kerry:  g1g3 doesn't look good because of e2g4. I considered a6b7 but
 went with this move as it starts to block in c7, brings another piece
 into the group, and makes it easy to connect the pieces on d8, e8.

 YL:   15  +97  f3b3 a6b7 a4c2 a3a5 h6g5 a5e5 g5f4 f1d1 b3d3 e8c6 f4e3
 Mona: 11  170  f3:b3 a6-b7 a2-b1 b5-d5 b3-d1 e8-f7 c7:f7
 Mona: 13  255  f3:b3 a6-b7 a2-d5 b7:d5 h6-h5 d5-c6 h5-e5 d8-d5 b3-e3

 Darse:  The programs expected either a3-d6 or a3-b2, preventing the
 capture on b3.  The e2-g4 plan did come up in a few PVs, but usually
 after something like h6-g5 and g5-d5.  After 9... f3:b3 both programs
 believe they have an advantage, and it is growing at greater depths,
 but this could simply be a reflection of their unsophisticated method
 of evaluation.

 However, from a practical point of view, allowing the piece on a4 to
 easily re-enter the game allows the programs to avoid their weaknesses.
 If the position should ever become a wide-open tactical affair, the
 deep searchers can play at a level that greatly exceeds anything
 humans are capable of...

      +-----------------+
    8 | . . . B B . . . |
    7 | . . W . . . . . |
    6 | B . . B . . . W |
    5 | . B B . . . . . |
    4 | W B W W . . . . |
    3 | B W W . . . . . |
    2 | W . . . W W . . |
    1 | . . . . . B B . |
      +-----------------+
        a b c d e f g h	 

   10. a6-b7 a4-c2

 Kerry:  I still think it's wide open, but then I'm not looking 10 moves
 into the future.

 Darse:  Well, seven and a half moves into the future (full width, all
 possible lines), but I sure wouldn't want to play this beast! :)

 YL:   14  +90  a4c2 a3a5 h6f6 e8c6 a2b1 a5d5 b1d3 b7b3 c2e4
 Mona: 13  304  a4-c2 b7:b3 c7-d7 d8-c7 h6-g5 c7-e5 c2-d3 b5-b2 g5-g3

 alternate choices:
 13  256  a2-d5 b7:d5 h6-h5 d5-c6 h5-e5 f1-d1 f2-d2 g1:d4 e2:b5
 13  275  b3-e3 g1-h2 h6-f4 d8-e7 f2-f5 d6:d4 f5:c5 e7:e3 a2-d2
 13  279  a2-b1 b5-d5 h6-g5 d8:d4 g5:d5 b7:b3 c7-b6 f1-c1 e2-c2

      +-----------------+
    8 | . . . B B . . . |
    7 | . B W . . . . . |
    6 | . . . B . . . W |
    5 | . B B . . . . . |
    4 | . B W W . . . . |
    3 | B W W . . . . . |
    2 | W . W . W W . . |
    1 | . . . . . B B . |
      +-----------------+
        a b c d e f g h  

   11. a3-e7 h6-f6

 Kerry:  I think I have to work on trapping the c7 piece.  My big problem
 now is that it is difficult to move the f1 and g1 pieces.

 YL chooses h6-d6 at all depths (output unavailable)
 Mona: 11  454  h6-f6 e8-c6 a2-d5 f1-h1 c7-a5 h1-f1 c2-d3

 Darse:  It looks like trapping the c7 piece will be critical, since
 White is ahead in the race to connect otherwise.  The programs now
 understand physical blockades better, thanks to the lessons from the
 first game, but this problem is trickier.  If a piece like the one on
 c7 isn't physically blocked, and has good mobility within a restricted
 region, then the evaluation function may not sense the danger.  It may
 continue to improve the rest of its position, and turn to the real
 problem only after losing valuable time.

 Both programs were very quick at electing 11... h6-f6, so they must
 consider it to be clearly best here.

      +-----------------+
    8 | . . . B B . . . |
    7 | . B W . B . . . |
    6 | . . . B . W . . |
    5 | . B B . . . . . |
    4 | . B W W . . . . |
    3 | . W W . . . . . |
    2 | W . W . W W . . |
    1 | . . . . . B B . |
      +-----------------+
        a b c d e f g h  

   12. e8-c6 a2-d5

 Kerry:  I don't think I have much option here. I was expecting c7a5
 last time.  Maybe now....

 Darse:  That move has appeared in the PVs, but apparently they want
 to consolidate first.

 YL:   14 +159  a2d5 c6f6 c2d3 d8e8 c3f3 e8c6 f3c3 f6h4 e2e4 h4f2
 YL:   15 +215  a2d5 f1h1 e2e4 h1f1 c2e2 b5d7 b3e6 f1g2 f2g3 g1h2 f6f5
 Mona: 11  534  a2-d5 d8-e8 d4-d1 c5:f2 c3-e5 f2-d4 c2:c6

      +-----------------+
    8 | . . . B . . . . |
    7 | . B W . B . . . |
    6 | . . B B . W . . |
    5 | . B B W . . . . |
    4 | . B W W . . . . |
    3 | . W W . . . . . |
    2 | . . W . W W . . |
    1 | . . . . . B B . |
      +-----------------+
        a b c d e f g h

   13. d8-e8 d4-d1

 Kerry:  Kind of a zugzwang situation for me.  I want Debby to let me
 off the hook!

 YL:   13 +29987  d4d1 c5f2 c3e5 e8d8 c2c6 b7b3 c7e7 g1d1 f6d4 f2d4
                   e7e4 d8e7 e2e6
 Mona: 13  32000  d4-d1 g1-g2 f6-d4 g2-g1 b3-d3 e8-d8 c7-a5 b7-d7
                   a5-a4 d8-f6 c3-e3 g1-g2 d3-b3

 Darse:  No such luck, I'm afraid.  Both programs were able to search
 to the end of the game, announcing a mate in 13-ply after 13... d4-d1.
 Examples of the principle variations include:

 c5:f2 c3-e5 e8-d8 c2:c6 b7:b3 c7:e7 g1:d1 f6-d4 f2:d4 e7-e4 d8-e7 e2-e6

 c5:f2 c3-e5 b4-c3 c7-b6 e7-c5 b6-a5 e8:e5 f6-f3 c3-a1 c2-d3 g1-d4 a5-b4

 g1:d1 f6-d4 b7:b3 f2:c5 b3-b6 c7:e7 f1-f2 e2-e5 b6-e6 e7-f6

 g1-g2 f6-d4 g2-g1 b3-d3 e8-d8 c7-a5 g1:d1 a5-a4 f1-h1 a4:d1 h1:d5 a4:d1
                                     b7-d7 a5-a4 b5-b3 a4-a3 b4-b2 f2:b2
                                     c6-a4 d1:a4 g1-g2 f2:c5 d8-c8 d5-b3
                                                 b7-a6 c3-e3 g1-e1 d3-b3
                                                 f1-d1 c3-e3 b4-d2 d3-b3
                                                 b7:d5 c3-e3 d8-b6 d3-b3

      +-----------------+
    8 | . . . . B . . . |
    7 | . B W . B . . . |
    6 | . . B B . W . . |
    5 | . B B W . . . . |
    4 | . B W . . . . . |
    3 | . W W . . . . . |
    2 | . . W . W W . . |
    1 | . . . W . B B . |
      +-----------------+
        a b c d e f g h

   14. Resigns 0-1

 Kerry:  I'm tempted to play again with something other than 7. c1:a3.
 I probably overreached at that point, and if I had done enough analysis
 I would have seen the position was too difficult to defend.  My guess
 is that a human player can still outplay Debby strategically, but he
 has to play very accurately as Debby is unforgiving of moves which are
 slightly inferior.  With a game like LOA it is difficult for a human
 to see very far down the game tree.

      +-----------------+       +-----------------+       +-----------------+
    8 | . . . . B . . . |     8 | . . . B . . . . |     8 | . . . B . . . . |
    7 | . B W . B . . . |     7 | . B W . B . . . |     7 | . . W . B . . . |
    6 | . . B B . W . . |     6 | . . B B . W . . |     6 | . . W B . W . . |
    5 | . B . W . . . . |     5 | . B . W W . . . |     5 | . B . W W . . . |
    4 | . B W . . . . . |     4 | . B W . . . . . |     4 | . B W . . . . . |
    3 | . W W . . . . . |     3 | . W . . . . . . |     3 | . B . . . . . . |
    2 | . . W . W B . . |     2 | . . W . W B . . |     2 | . . . . W B . . |
    1 | . . . W . B B . |     1 | . . . W . B B . |     1 | . . . W . B B . |
      +-----------------+       +-----------------+       +-----------------+
        a b c d e f g h	          a b c d e f g h           a b c d e f g h  

  YL PV:        c5:f2
          c3-e5 e8-d8
          c2:c6 b7:b3

      +-----------------+       +-----------------+       +-----------------+
    8 | . . . B . . . . |     8 | . . . B . . . . |     8 | . . . . . . . . |
    7 | . . . . W . . . |     7 | . . . . W . . . |     7 | . . . . B . . . |
    6 | . . W B . W . . |     6 | . . W B . . . . |     6 | . . W B W . . . |
    5 | . B . W W . . . |     5 | . B . W W . . . |     5 | . B . W W . . . |
    4 | . B W . . . . . |     4 | . B W B . . . . |     4 | . B W B W . . . |
    3 | . B . . . . . . |     3 | . B . . . . . . |     3 | . B . . . . . . |
    2 | . . . . W B . . |     2 | . . . . W . . . |     2 | . . . . . . . . |
    1 | . . . B . B . . |     1 | . . . B . B . . |     1 | . . . B . B . . |
      +-----------------+       +-----------------+       +-----------------+
        a b c d e f g h	          a b c d e f g h           a b c d e f g h  

  YL PV:  c7:e7 g1:d1
          f6-d4 f2:d4
          e7-e4 d8-e7 e2-e6++