Just a thought not a bug. I haven't played your program in years.

[CCamp81318]

  After the latest two human v. computer matches I have been thinking about 
computer chess a lot more. I am nearly home bound and play against my 
computer opponents quite a bit. I am not in either Kasparov's or Kramnik's 
class but I get a respectable score against all of the commercial programs 
(about 40%).
    What has come to my attention and I am finally able to put into words are 
the two reasons that I do well against the bots.
    The first reason is that the bots have no concept of never. I frequently 
get them into positions where they are tied to the defense of something and 
can never give that up. I just go about improving my position for the next 20 
moves while they shuffle a piece back and forth until they run out of moves. 
Kind of a middlegame version of Zugswang. The bot has been lost for 20 or 
more moves before it starts to realize that it is in danger.
    The second is the way the evaluation function works. Humans look at (or 
at least I do) with how close the game is getting to winning or drawing while 
the computers sort of emulate this they fall down very badly at converting 
the advantage. By this I mean that they will let you into a bad but drawable 
ending. This is frequently the same thing as the first problem, the defender 
can set up a fortress and no improvement can be made.
    My suggestion is an added evaluation routine to check for positions where 
one side is not doing anything, where no progress is being made. I have no 
idea how this could be written at this point but if you are interested I 
would be willing to put forth the effort in an attempt to quantify the 
variable.

CTCampbell