Computer Chess Club Archives


Search

Terms

Messages

Subject: 379 ply search

Author: Dan Honeycutt

Date: 18:39:14 04/02/04


This post is for other beginners like myself.  You veterans are welcome to read
and laugh but no comments are necessary.

I recently checked to see how my engine was doing in a test match and found it
in a dead-drawn, blocked-pawn, opposite-color bishop ending.  The other engine
showed a draw score but poor Bruja, clueless, thought it had a slight advantage
and wanted to fight on.  I thought about adjudicating but browsing the score
showed nothing had happened for some time so I figured the game had to soon end
by 3 rep or 50 move rule.

Then I noticed the output.  Bruja was reeling off plys like pennies at the gas
pump with today's high prices.  My first thought was "wow, hash tables really
help in these positions with so few pieces".  I became uneasy when the depth
passed 20 and downright worried when it passed 40.  Terror gripped my heart when
the depth flew past my array dimensions.  I watched helpless, waiting for the
crash and wondering what was wrong with my bounds checks.

After iteration 379 Bruja decided it had searched deep enough and played.  The
other engine replied and the game ended in a 50 move draw.  The postmortem was,
fortunately, fairly simple.  I had bounds checks in the search but nothing in
the iteration function.  Iterate called search which promptly returned as all
moves led to a draw two plys.  Iterate incremented the depth, recalled search
with the same result until the time allocation finally ran out.  (I should have
noticed that the whole time the pv was only 2 moves but i was fixated on the
depth.)

Others with green engines may want to check to see that they don't have a
similar situation.

Dan H.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.