tcollins@mentor.com (Truman Collins) (12/17/90)
Archive-name: gnu/chess/gnuchess-analysis/1990-12-13
Original-posting-by: tcollins@mentor.com (Truman Collins)
Original-subject: Chess analysis program.
Reposted-by: emv@ox.com (Edward Vielmetti)
There has been some discussion recently about a chess computer or
program that would do some sort of analysis on a game already played.
A while back I made some modifications to gnuchess 3.1 to do something
similar.
Basically, what I did was to make a new front end to gnuchess that
reads a file containing a game in long algebraic notation. It then
plays through the game and for each move gives the move it would have
made along with the score it would give the position after its move.
At the end it gives for each side the percentage of the actual moves
that corresponded to the computer's recommended move, which can be
interesting but doesn't seem to be very meaningful with small
percentage differences.
Here is an example of the output on a short game from a book:
Move Actual Suggest Score Depth Actual Suggest Score Depth
------------------------------------------------------------------
1 e2e4 g1f3 12 4 e7e5 g8f6 -10 4
2 g1f3 b1c3 12 4 d7d6 b8c6 -16 4
3 f1c4 d2d4 29 4 h7h6 b8c6 4 4
4 b1c3 d2d4 54 4 c8g4 b8c6 -24 4
5 f3e5 f3e5 152 4 g4d1 d6e5 -122 4
6 c4f7 c4f7 9996 2 e8e7 e8e7 -9997 1
7 c3d5 c3d5 9998 1
White's percentage was 42.86%. Black's percentage was 16.67%.
You can specify a maximum depth for the analysis and a maximum
time. I find that it is most useful for showing where you missed
winning combinations or where your opponent did. It has helped me some
in showing general situations where I tend to miss attacking
possibilities. It's mainly good for tactics rather than strategy.
I did this on a PC, and I uploaded it along with all of gnuchess
3.1 to Compuserve. It is in the Chess Club library of the Gamers
forum. The name of the file is GNUCH3.ARC, and it has been
compressed using pkpak which is compatible with arc. The file is 274k
long. It comes with all of the source code, and it should be easily
portable to another machine. I have a copy that runs on an Apollo
workstation, but there is no place I can put it for FTP access.
It has a problem in some situations involving possible pawn
promotions; it gets lost and can't go past a specific move. I disabled
the opening book so it would do more tactical analysis.
If anyone makes any additions or improvements to it, I would like
to hear about them. I would also like to hear about any suggestions,
although I won't have time to make any major changes in the near
future.
-Truman
tcollins@pdx.mentor.com
71121.3343@compuserve.com