Author: Robert Hyatt
Date: 20:55:25 09/10/04
Go up one level in this thread
On September 10, 2004 at 22:55:30, enrico carrisco wrote: >On September 10, 2004 at 17:14:44, Robert Hyatt wrote: > >>On September 10, 2004 at 15:56:45, Sandro Necchi wrote: >> >>>On September 09, 2004 at 10:50:49, Robert Hyatt wrote: >>> >>>>On September 08, 2004 at 19:12:56, Matthew White wrote: >>>> >>>>>On September 03, 2004 at 15:07:17, Graham Banks wrote: >>>>> >>>>>>On September 03, 2004 at 13:17:51, robert flesher wrote: >>>>>> >>>>>>>If you are going to waste your precious time and everyone else here then please >>>>>>>indicate that you have given unfair advantages to certain engines. >>>>>> >>>>>> >>>>>>I think people should read the setup details and maybe look through the whole >>>>>>range of games before going off half cocked! >>>>>>All engines are using the Fritz powerbook tournament settings. There is the odd >>>>>>strange opening due to the maximum variety setting used, but I think you'll find >>>>>>that this has equally affected all engines and that no particular engine has >>>>>>been disadvantaged. >>>>>>For the final of the tournament I intend to optimise the powerbook settings, so >>>>>>this should eliminate any unusual openings. >>>>>>Graham. >>>>> >>>>>Would it be a more equitable test to have each pair of opponents play both sides >>>>>of each oddball opening? >>>> >>>> >>>>No... >>> >>>OK, I try to find an example to show you what you are stating. Again Bob is 100% >>>correct. >>> >>>Now as you know the F1 cars do not use the same tyres; mainly there are 2 >>>company making them; let's call them X and Y. >>>Since everybody is asked to improve as much as possible the latest improvements >>>involve the tyres too. >>> >>>So if you state that ALL cars needs to use the same X tyres to eliminate >>>advantages, you are not doing that as you are favoring those who have been >>>working in cooperation with company X and penalizing those who have been >>>cooperating with company Y, so improving the cars with those tyres. >>> >>>In your case it is even more unfair as the car company could make changes to >>>reduce/eliminate the handicap, but you are chosing a chess program which is as >>>it is and will suffer from that. >>> >>>If you think that you know more than me in this field I give you some figures: >>> >>>1) I am testing/checking computer games since 1976 >>>2) I think I have seen/checked something like 140.000 games (about 50% played by >>>computers) >>>3) I have tested/own something like 250 chess programs/chess boards (including >>>experimenthal versions too). >>> >>>So, I can state that Bob is correct without any doubts. >>> >>>Sandro >> >> >>I don't even understand how the topic keeps coming up over and over. Games with >>ponder=off. Games with odd books. Games with random books. Games with both >>sides forced into the same opening positions. Games with no learning. Games >>with learning reset between games. And I don't see how any of that produces >>anything but excessive noise... >> >>But those of us that have done this a while understand the problem... >> >>Thanks... > >Well, let's not forget about John Nunn's positions. Certainly, testing two >engines and forcing them into these positions as either color is a useful >benchmark. No it isn't. Again, have you _ever_ seen a serious human tournament were players were forced to play a specific opening, regardless of whether GM consensus says the positions are equal or not? Of course not... And you can't expect the programs to be put in the same predicament either. A program _might_ play all positions well. It _might_ play some better. I'd say the author has a feel for that and assists via reasonable book line preparation to avoid the positions where the program plays poorly, and vice-versa. Najdorf lines come to mind as very dangerous and many avoid them completely for that reason... > >If not, then Hiarcs 9 was produced on about 50% excessive noise. Maybe we can >turn it into an MP3 player... :P Hiarcs might do well in _any_ position. But then I know human players that play any opening as well. But not _all_ humans do that. neither do all programs... > > >-elc.
This page took 0.02 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.