Author: Robert Hyatt
Date: 05:22:24 04/30/98
Go up one level in this thread
On April 30, 1998 at 05:20:27, Ulrich Tuerke wrote: >> ... >>Hi Uli, >>I don't understand: >>.."Who for instance is going to check that the programs do not use more >>time than available ..". >> >>Did not every programm test the time of the opponet and say 'Flag >>fallen' ? >> >>.. > >Hi Peter, >you shouldn't forget that there are two clocks, cause each program has >it's own time management. You can never be sure that the programs' >bookkeepings of used times do 100% agree. In fact, deviations wan't be >avoidable with the elapse of time. E.g., the communication of moves will >take some time (with auto232 time, these spans are even of the order of >seconds). Some programs don't seem to account for the time which they >use to scan the opening book. And so on. >IMO, for these reasons you can't rely on the time book-keeping of one of >the competing chess programs. >Other points of doubt could be: have both of the programmers correctly >implemented the criteria for draws (50move-rule, ...) ? Discrepancies >are possible. Therefore I think that a clean solution for automatic play >must include a referee program, which should be a complete chess-machine >(to be able of keeping track of 50move rule, repetitions, ...). >Regards, Uli however all of this is trivial to address. Write a simple "server" process that understands chess and acts as a "referee". Both chess programs send moves to this referee which sends the moves to the other program. We run othello tournaments here for our AI class, and use exactly this. The referee takes less than 1 second of cpu time for the entire game, so there is no overhead to speak of. It can keep up with the time, the position, 50 move draws and all of that.. and it would be trivial to do...
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.