Author: John Merlino
Date: 12:21:54 03/13/00
Go up one level in this thread
On March 13, 2000 at 14:47:08, KarinsDad wrote: >On March 13, 2000 at 13:09:49, John Merlino wrote: > >>On March 10, 2000 at 15:42:25, KarinsDad wrote: >> >>>On March 10, 2000 at 13:39:13, John Merlino wrote: >>> >>>[snip] >>>> >>>>As it stands now, the games that you would play (once again, due to time >>>>constraints) would be Fischer 2/12. True, this is at the high end of blitz >>>>chess, but since there are SO MANY personalities that need to be tested, they >>>>would need as many games as possible. >>> >>>Why not just make it G15 (or even G10 for that matter)? It's approximately the >>>same amount of time for a reasonable length game and you do not get people who >>>have limited opening experience from losing on time (or via gross positional >>>blunders due to time pressure) in the opening. >>> >>>KarinsDad :) >> >>Admittedly, Fischer, for the most part, favors the computer (especially in the >>opening, when a personality with a deep opening book will gain an automatic 12 >>seconds with every book move). > > >So, why do a test that favors the computer? Do one that challenges the computer. > > > But the main reason that the development team >>chose that time control is that it allows for a much longer game than G10 >>(which, on ICC, is approximately "equal" in time to Fischer 2/12). > > >Yes, but not much longer than G15 (at the same rate, a 65 move game). This does >not seem to be that great of a reason. > > > It also tests >>the personalities' endgame abilities better, giving them a minimum of 12 seconds >>to think about every move, no matter how long they have had up to then. > > >How so? If the program is designed properly, it will estimate how long the game >will continue based on number of pieces, activity of pieces, number of moves >made, time left, maximum game length, etc., and will ensure that it has >sufficient time remaining (or not). Just like a human would. > >Your method does not test the program's ability to handle standard time controls >against humans and artificially ensures that there will be at least 12 seconds >per move for the endgame (i.e. the program can make sure that it does not lose >on time and makes a reasonable move at all times). The entire problem with >Fischer time control is that it is artificial (i.e. you will get skewed results >from a lot of people less than class A). There is a good letter to the editor on >this in this month's edition of the Chess Life. > >With your method, you will get skewed bad results from the humans in the opening >and skewed good results for the computer in the endgame. How does that attempt >to ensure that the personalities are "similar" to their human counterparts? > > > I see >>your point, but I think I agree with the team here. >> >>jm > > >Hmmm. To each their own. I have always been a purist when it comes to time >controls. > >KarinsDad :) You make your argument well, and I will pass it on to the development team. Are you USCF rated? Would you be willing to participate? jm
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.