Author: Vincent Diepeveen
Date: 06:10:12 03/05/04
Go up one level in this thread
On March 05, 2004 at 09:02:41, Albert Silver wrote: >On March 05, 2004 at 08:55:05, Vincent Diepeveen wrote: > >>On March 05, 2004 at 08:49:14, Albert Silver wrote: >> >>Please ask SSDF at what hardware they *really* test, not what is put in the >>list. And/or whether they use time compensation. >> >>I really like to see all games they play and like to be able to see the moves >>played there show up at the screen here too. That is something SSDF can easily >>provide (only count for their rating list games that get published) and should >>do to start with to get more credibility. >> >>How can you test things without having the games? >> >>There is just 1 tester who publishes games AFAIK. > >And the games he publishes suggest different hardware is being used? Yes? You >don't need every game played. Just a few should be necessary to see whether the >times and depths published correspond to the announced hardware configuration >used. Or are you suggesting he uses different hardware than the other testers? > >If you can show that the games played and published by Tony Hedlund here do NOT >correspond to the announced hardware, you have a point, otherwise, you're just >making accusations with no facts of any kind to sustain your position. So either >show some kind of analytical evidence to back up your position, or state >outright that you believe Hedlund is rebelling against the SSDF and is using the >announced hardware contrary to the rest of the testers. > > Albert You are not reading what i wrote. I say 1 tester is transparant, the rest is not. That's my statement. If i keep a tournament without publishing results, but claim A, then even Uri posts he doesn't believe it until he sees proof. Why do you believe SSDF? My experience is that even the most experienced testers in testsets of games make major mistakes. I had a logfile from diep back, diep getting 20k nps. It appeared Fritz swallowed all system time at that machine. Other test, same thing, this was at a dual K7. Same problem now it was deepfritz which by default eats 2 cpu's when playing opponents at the same computer. First putting it to 1 cpu then starting match just does't work. There is a lot of pitfalls. If just 1 tester publishes things and the rest consequently do not, then that is reason for me to mention that testers in my experience make major mistakes. The ideal way to hide your bad testing is by not publishing results. Who says that they are not fake games? People pay Karlsson. Perhaps people who pay get fake games? Read me well, i do NOT accuse Karlsson from that. Not at all. I just say that in my experience the biggest problem of auto playing programs is the human factor. Misinterpretation of facts and wrong setups. Even the BEST testers have this problem. They also make mistakes. Usually they do not make the mistake for the interface they read the results from... ...it's always advantage for that single interface... There is not a single proof of the opposite here.
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.