Author: Tony Werten
Date: 00:51:54 04/20/01
Go up one level in this thread
On April 19, 2001 at 14:26:17, Miguel A. Ballicora wrote: >On April 19, 2001 at 10:59:01, Dan Andersson wrote: > >>How many chess programmers do attack extensions other than checking moves. For >>example, when pressing an attack it is often as forcing to threaten the >>opponents queen an thus you gain tempi. Anyone doing that and what are your >>results. >> >>Regards Dan Andersson > >I just tried a little bit of that but I had problems. The way I did was >to look at the value that the nullmovesearch returned. If it is <= -mate_in_100 >I extended. I other words, every time that a side is threatened with checkmate >I extended. I just made my search huge in one position I tried. >I think that it could be useful in tactical situations but using fractional >extensions (that I did not implement it yet) to limit it. I haven't implemented fractional plies either (yet, it's high on the list, just not looking forward to all the testing) but I limit the threat extensions to the lowest couple of plies. The idea is that if I have 4 ply of search left, I might overlook a checkmate in 5 ply but if I have 10 ply left, the search should be able to take care of it without any extra help. > >I think that "Attack" extensions will increase the tree too much unless is >very much controlled doing it with a very low fraction. That were my experiences too. One of the problems: Opponent has done a stupid move; it attacks my undefended queen, but did it with his undefended queen. If I nullmove I get an extensions back, just taking his queen would have stopped the search quite fast. >I will be very interested too in hearing comments and learn about this. Somebody ( D Beal, J schaeffer ? ) has written something about this. He limited it to the qsearch (experiments ...something). The idea was to see if the opponents had any threats. Can't remember if it worked back then, but it's always usefull to relook at those old ideas. I have gotten quite a few good ideas from those oldies. ( wich were not-usefull ideas back then ) Faster hardware and new standards can make the difference. cheers, Tony > >Regards, >Miguel
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.