🚀 go-pugleaf

After having thought of doing it for a long time, I finally set aside an
evening and played MaxII against CM4000.  MaxII is the version of gnuchess
4.0 that I have been tinkering with and which I run on ICS.  It is
basically patchlevel 68 code, but with lots of modifications.  Some of
the modifications which I've found beneficial have been posted to gnu.chess.
Others deal with tuning of evaluation parameters and minor bug fixes.  
MaxII also uses its own book, which is based on BCO 2 and contains about
32K positions.

Anyway MaxII is running on an IBM RS6000/370 and calculating about 20K
nodes/sec.  On a 486/33, it would be calculating at about 5K nodes/sec.
CM4000 Turbo runs on a 486/33 with 16 MB ram and under WIN-OS/2.  I set
MaxII to 20 secs per move and CM4000 to 30 secs per move.  Both programs are
playing without thinking on opponent's time.  During the initial stages of
the game, both programs typically calculates up to 6 plies deep, but as more
pieces are exchanged, MaxII tends to get more plies.

With the "tournament" background made clear, here is the game itself, with
some of my own comments.


White: CM4000
Black: MaxII (gnuchess4.0pl68 with mods)

1.	d2d4	g8f6
2.	c2c4	c7c5
3.	d4d5	e7e6
4.	b1c3	e6d5
5.	c4d5	d7d6
6.	g1f3	g7g6
7.	g2g3	f8g7
8.	f1g2	o-o
9.	o-o	f8e8	

Both programs have been using their books until now.  CM4000 thought for
almost 2 mins before playing

10.	c1g5	d8b6 (-11)

The (-11) means that MaxII thinks it is -0.11 pawns down.  Suddenly 
CM4000 decides to give up its Bishop.

11.	g5f6	g7f6 (33)
12.	d1c2	c8f5 (46)
13.	e2e4	f6c3 (83)

CM4000 decides to sac a pawn for reasons unknown to me.  It thinks it is
ahead by about 0.8+ pawns.

14.	c2c3	f5e4 (72)
15.	f1e1	b8d7 (111)
16.	f3h4	e4g2 (148)
17.	h4g2	b6a6 (144)
18.	b2b3	e8e1 (157)
19.	a1e1

CM4000 sacs its a-pawn.  It's rook is heading for the 7 rank and it 
probably feels it has enough compensation.

19.	...	a6a2 (221)
20.	e1e7	d7e5 (238)
21.	e7b7	a8e8 (243)

Although MaxII last move shows a score of +2.43 pawns, in actual fact,
it has failed high, but did not complete the re-search due to timing out.
In reality, white rook is lost!

22.	g2f4	a2a6 (552)	* Note the increase in score
23.	b7c7	a6b6 (540)
24.	c7c6	e5c6 (540)
25.	d5c6	b6c6 (541)
26.	f2f3	e8e5 (554)

Now MaxII just shuffled pieces instead of trying to advance that d-pawn.
The passed pawn evaluation code needs some work here.

27.	h2h4	a7a6 (553)
28.	g1f2	c6b5 (554)
29.	c6b5	g8f8 (556)
30.	c2d2	b5b6 (526)
31.	f2g2	b6c6 (519)
32.	f4d3	e5d5 (526)
33.	d2c3	c6b5 (525)
34.	d3f2	b5e2 (576)

CM4000 last move is a mistake and now MaxII capitalized on it.

35.	f3f4	d5d2 (748)
36.	c3f3	e2f3 (765)
37.	g2f3	d2b2 (748)


At this point, I stopped the game.  I was surprised to discover that
MaxII could win and pretty convincingly too.  Maybe its because of its
much faster hardware.  But considering that in the initial stages, both
programs are looking at 6 plies, this might not be the reason.  Anyway
for the next game, I'll give CM4000 45 secs and see if that will even
things out a bit.

Kong Sian