Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: time management & fail low/high

Author: martin fierz

Date: 13:27:27 05/30/02

Go up one level in this thread


On May 30, 2002 at 12:57:53, Scott Gasch wrote:

>Before I search I set two time limits, a soft limit and a hard limit.  If
>everything goes well I move after the soft limit.  But if:
>
>  1. The root position is "critical"
>  2. We've seen many root fail lows
>  3. We're just out of book
>  4. We're resolving a root fail low / fail high
>  5. We're already done searching the first move at this depth (with the
>     assumption that the rest should come quickly or cause a FL/FH if
>     interesting, might as well search out the ply).
>
>then the engine keeps searching until either the condition disappears (for
>temporary conditions) or the hard timer is exceeded.  The code also has some
>provisions for moving before the soft timer has expired in situations like:
>
>  1. It thinks the root position / move is a simple recapture.  Everyone
>     thinks this is a dumb idea but it has never bitten me.  Simple recapture =
>     there is only one way to recapture a piece > knight and there are no
>     checks against the enemy king.
>  2. I have some code that makes it move faster when time is low in a blocked
>     position when the move it is considering doesn't unblock the position.  I
>     added this as part of an attempt to "fix" anti-computer play.  It also has
>     been programmed to hate totally blocked positions so hopefully avoids this
>
>Hope this is helpful for you.  Someday I may go back and add a second "harder"
>timer... Usually the hard timer is about 2-3x the short timer and I think in

sure - this already is much more sophisticated than what i do...

>Another thing, why are you not playing the tip of the PV at the time you run out
>of time?  If you think about when your PV changes (or, better yet, print out a
>message) at the root you'll see that playing the first move on the PV when you
>are halfway through ply X is the same as playing the move returned by ply X-1 as
>long as the PV hasn't changed.  If it has, I play the new move.

i don't have a reliable PV - i'm using MTD, and there you just have no real PV
except whats in the hashtable - and that can always be overwritten :-(

aloha
  martin



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.