Author: Alois Ganter
Date: 07:36:36 05/01/01
Go up one level in this thread
On May 01, 2001 at 09:38:05, Robert Hyatt wrote: >On May 01, 2001 at 02:10:40, Alois Ganter wrote: > > >> >>A strong chess player can point out obvious dangers to a chess program during >>the game. E.g. by discreetly adjusting the rest position of his mouse arrow >>during moves. Say the program grabbed a pawn and now desperately needs to >>consolidate instead of going for more material. Any 1800 player can judge that >>better than nowadays top programs. > > >I wouldn't take an 1800's opinion over a program's opinion, _ever_. The >1800 player will make so many more mistakes than the program, he won't be >able to understand the tactics the program sees, etc. Its not about recommending moves. Its about tweaking the evaluation towards consolation and king safety when you smell that the program underestimates imminent dangers. I am 1750 on good days and I have seen so many stupid things by strong chess programs. A typical problem is: You have a d4/d5 center, black pawn on c5, white bishop on d3. The program grabs space with c5-c4, attacking the bishop. Then it loses on the king side with a blocked centre. So after entering White's Bd3 move, the 1750 player positions his mouse on the screen in a way which says "Lower the evaluation for a blocked pawn chain!". Maybe he is off by 0.2 pawns with his judgement but he will not lose. If the program sees that it will win a piece with c5-c4, it will play it anyway. > >And yes, there is always the potential for cheating, for those that are so >inclined. I'm not sure how they get away with it however, since others can >eventually use the same (or similar) engine and find out that it won't reproduce >the move played.. > Nope, those versions would never get published, only a couple of months after the tournament, when everybody has forgotten about it. > > >> >>Manual operation opens a pandora box of disturbing unfairplay which the other >>program cannot defend against. It should e.g. be strictly forbidden to operate >>chess software with a mouse during important matches. >> > > >Why? Do you also propose playing the match in a room surrounded by steel mesh >to block radio transmission? I could influence an autoplayer game just as >easily if I wanted so you have to block RF stuff, IR stuff, even high-frequency >audio stuff. No, you can't. You do not have access to the machines on which the match is played. So you cannot help the program from the outside because you cannot build in a receiver. Unless you bring your own special hardware to a tournament where you operate yourself... And if you try to simply disturb the program or the protocol by irradiating RF stuff it will show it in its logs or crash or whatsoever. > >Oh yes, humans have a tendency to cheat as well, so the opponent has to also >be cut off from the outside world... > > > > >>On the other hand every program can trivially defend against abusing input from >>an autoplayer interface. No human interference, no cheating. > > >It isn't as "trivial" as you think... Why not? It is certainly easier than writing a decent Internet firewall. If the opponent sends something you do not like, e.g. a takeback command during a game you simply log a warning message on the screen: "Opponent sent takeback" and ignore it. My conclusion is that autoplaying is safe and fair because the program has full control about whats happening to it, while manual is an invitation to abuse. Unless it is limited to typing in moves in an e2-e4 style. But even the rhythm of typing could transmit one or two crucial bits of information to the program.
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.