Author: Don Dailey
Date: 15:08:41 11/17/98
Go up one level in this thread
On November 17, 1998 at 05:53:57, Jürgen Hartmann wrote: >On November 17, 1998 at 04:39:08, Ed Schröder wrote: > >>However and here starts the confusion, the communication of >>moves between two Pc's is just ONE aspect of an autoplayer. >> > >Shouldn't the autoplayer not do one and only one thing: Transmit moves until the >game is ended by mate, theoretical draw or mutual agreement, then begin a new >game? Hi Jurgen, I think you are adding validity to Ed's point. There is a LOT more to it than this simplistic naieve view. It appears that you are not qualified to talk about it. I wrote my own DOS autoplayer a few years ago and there are a bunch of issues and things that can go wrong. My autoplayer did not need support from the chess programs, it read the screen, pressed the keys (by stuffing the keyboard buffer) and kept time. Your view of what an autoplayer is not very accurate. Here are some issues: . Does it keep time? What happens when one program exceeds? . Does it stop and restart games accurately? If one of the programs crashes does this get reported as a loss or does it get ignored? . Does it reset the parameters of the program to the state you intended? . Does it do all of these things EVERY time? Does it report every game that's played? If it misses 1 out of 30 how do you know? . Does it do stuff to interefere with the performance of one or both of the programs? How do you know it doesn't? All of these things are really hard to determine without an enormous amount of effort. If it failed to report a game every 50 games how would you know unless you are watching it like a hawk? What kind of system would you use for proving that everything went exactly as planned? When you start an autotest match like this you need to PROVE all this stuff is working properly. How would you find out that program x crashed on the 53rd game and it didn't get reported? Are you watching each and every game to find this out? It's even more complicated that I am telling you and on top of this, if there is a problem there is an excellent chance it was a bug, not by evil intent. You are making a typical error that people make that don't understand. They distill it to some simplistic level that betrays their total lack of understanding. It's typical even among programmers to underestimate the task at hand because even they look at the details and fine points last. >This is what players do over the board. > >Why is it not possible to check this 100% with a communication log? Whatever a >communication link does shows up by definition in the log. If you are a chess programmer you can check this, otherwise you have a very difficult task ahead of you. But I thought Ed checked this, isn't that what he said? >About saving: Why not save the game yourself in Rebel when the game ends? Same issue. Are you going to stand around and save each and every game? You might not be able to save them anyway if the autoplayer has control of the games. How do you propose to do this? A programmer could try to check these things however. It's a lot of effort though. >If Rebel's learning/saving after games depends on another computer being hooked >up to it by a strange cable and sending "Save" commands then I think you should >talk to your programmers to implement it in a more general way. > >This discussion has 'sore loser' written all over it (you invited the flame). > >Jürgen Hartmann > >PS: Does the autoplayer protocol allow draw offers or resigning?
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.