Author: Robert Hyatt
Date: 17:56:07 12/02/03
Go up one level in this thread
On December 02, 2003 at 11:37:08, Bob Durrett wrote: >On December 01, 2003 at 22:14:19, Robert Hyatt wrote: > >>On December 01, 2003 at 19:20:35, Bob Durrett wrote: >> >>>On December 01, 2003 at 12:39:13, Robert Hyatt wrote: >>> >>>>On December 01, 2003 at 11:41:56, Bob Durrett wrote: >>>> >>>>>On December 01, 2003 at 11:30:54, Robert Hyatt wrote: >>>>> >>>>>>On December 01, 2003 at 11:23:09, Bob Durrett wrote: >>>>>> >>>>>>>On December 01, 2003 at 11:00:16, Sven Reichard wrote: >>>>>>> >>>>>>>>On December 01, 2003 at 10:17:16, Robert Hyatt wrote: >>>>>>>>> >>>>>>>>>The operator made the decision to play on. But the operator is _not_ allowed >>>>>>>>>to make _any_ decisions while a game is in progress, as per the rules. >>>>>>>>>Therefore this reasoning simply is unsound because it is based on rules that >>>>>>>>>were not in effect. The operator is passive. He _always_ has been passive, >>>>>>>>>at least when we go by the rules in force for these events. >>>>>>>>> >>>>>>>> >>>>>>>>If he is always passive, I don't see the point of having an operator. Most >>>>>>>>programs run on all-purpose hardware (maybe enhanced by some additional >>>>>>>>circuits) with networking capabilities. Why not have the opponents communicate >>>>>>>>directly, using a standard interface like xboard or UCI, maybe relayed via an >>>>>>>>arbiter program? Then they can decide for themselves whether to offer or claim a >>>>>>>>draw. >>>>>>>> >>>>>>>>If the GUI makes decisions for the engine, the combination GUI/engine should be >>>>>>>>considered the competitor. >>>>>>>> >>>>>>>>Sven. >>>>>>> >>>>>>>My understand is that the operator is there for a good reason similar to why a >>>>>>>wise tournament director must be there. Chess computer tournaments are still >>>>>>>evolving and humans need to be there to correct for errors or oversights of the >>>>>>>programmers. >>>>>> >>>>>>Crafty has played over 1 million games on chess servers with no problems. We >>>>>>want the operators there in case there is a network issue that causes a >>>>>>disconnect, or a hardware problem that might require a reboot (very rare), and >>>>>>to discuss things with other programmers. We don't need operators to handle >>>>>>normal "problems" as they simply don't happen. >>>>> >>>>>Please forgive me for saying this, but: "In all due respect, playing your >>>>>engine on a chess server is simply not the same thing as playing in a chess >>>>>computer tournament. You might use similar technology but there are significant >>>>>differences in the two situations." >>>>> >>>>>Bob D. >>>> >>>>And what would those differences be? FIDE has had events on the net. CCT >>>>is a computer chess tournament that works just fine on the internet. The >>>>main difference is that there are no "operator errors" to contend with because >>>>there are no "operators". >>> >>>The two examples you cite are tournaments. When you said "Crafty has played >>>over 1 million games on chess servers with no problems." I did not think you >>>were talking about tournaments. The non-tournament automated use of chess >>>computers at ICC is a different application from a tournament application. >>> >>>Bob D. >> >> >> >>Perhaps I am just dense, but I don't see how. The first two years crafty >>was on ICC, it played in 1-2 human tournaments _every day_. All it needed >>was for someone to match it, or for me to tell it to match someone, and >>away it went, with absolutely no problems of any kind... >> >> >> >>> >>>> >>>>The server is the final arbiter with respect to time, draws, wins and losses. >>>> >>>> >>>>> >>>>>> >>>>>>> When the available rule set fails to properly cover a new >>>>>>>situation, humans must get involved. Hopefully, their actions will be >>>>>>>reasonable. [Throwing a draw away would not be reasonable.] >>>>>>> >>>>>>>Bob D. > >I believe your main point is quite valid. Computer tournaments of the future >should be automated. If it is possible to anticipate every possible >eventuality, and to automate the appropriate corrective actions, then it should >be done. After much experience, the software should be extremely mature and >problems minimized. However, don't forget: "If something can go wrong, it >will." : ) > >Bob D. The main problem we had in previous CCT events was disconnects. But ICC takes care of that automatically, so it isn't a problem at all. Of course an engine can have a bug, play an illegal move, or whatever, but it will lose on time if it doesn't get fixed quickly, making the event flow smoothly.
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.