Author: Steven Edwards
Date: 22:08:19 04/25/05
Go up one level in this thread
On April 23, 2005 at 15:57:59, Peter Skinner wrote: >On April 23, 2005 at 13:19:05, Steven Edwards wrote: >>Some things I've noticed concerning questionable behavior at ICC: >> >>1. Elo 1200 humans mysteriously playing Elo 2400 chess, but only against my >>program and other programs. > >This is the main reason most operators use some sort of rating formula. eg. >rating>myrating-500 I have this now and have set the delta to 600 points. >>2. Human opponents playing five man endgames just as would a tablebase equipped >>program. > >You will also notice the move times are the same as if you were to manually move >the pieces yourself. Perhaps an ICS could randomly sample played games by humans to check for this. (Maybe this is already done silently.) It would be a nearly sure method of detection. The automated checker could insert an indelible finger note to the effect of: "This player plays endgames like a computer" and so be a warning to others. >>3. Opponents who repeatedly adjourn blitz games via disconnection. > >Easily avoided. set noescape 1 I've do this, but I still seem to get adjournments via opponent disconnection. Maybe there's another var setting needed. >>4. Established computer players without program/CPU finger notes, or any finger >>notes at all. > >I just +noplay those individuals. This seems to be the best thing to do. >>5. Computer opponents that waste time by extensively analyzing a position when >>only one move is available, or when a quick forced loss is unavoidable. > >This is unavoidable. Then a program is thinking, there is nothing you can do but >wait for it's move. Yes, but why should a program spend time analyzing when only one legal move is available? >>6. Computer opponents that lack a reasonable resignation capability. > >I used to set my resign factor at -6.5, but I have seen even computer blunder a >won game to a draw or even worse a loss. Symbolic's xboard interface code will give up the ghost when three consecutive searches fall below one rook value. It doesn't care about the clock, the opponent's rating, or anything else. Proper resignation is a manner of professionalism. It shows respect for the opponent. It also earns respect from a decent opponent as it shows that the program is smart enough to know when to give up and considerate enough not to waste the opponent's time.
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.