Adjudication & engine testing ?
Moderators: hgm, Rebel, chrisw
-
- Posts: 234
- Joined: Sat Jan 17, 2015 11:54 pm
Adjudication & engine testing ?
How much would adjudication affect the result if at all , given that sometimes an engine maybe winning/drawing but about to run out of time to the point that without adjudication it would lose ?
-
- Posts: 2204
- Joined: Sat Jan 18, 2014 10:24 am
- Location: Andorra
Re: Adjudication & engine testing ?
Depends on how well the engine handles time. Many engines don't lose on time mostly never, so this will barely affect to them.MahmoudUthman wrote:How much would adjudication affect the result if at all , given that sometimes an engine maybe winning/drawing but about to run out of time to the point that without adjudication it would lose ?
Daniel José - http://www.andscacs.com
-
- Posts: 4052
- Joined: Thu May 15, 2008 9:57 pm
- Location: Berlin, Germany
- Full name: Sven Schüle
Re: Adjudication & engine testing ?
I think that time losses are very rare today, so they are not the key point regarding adjudication. For me the key point is how often the adjudication is "incorrect" or "inappropriate" with respect to the situation on the board.MahmoudUthman wrote:How much would adjudication affect the result if at all , given that sometimes an engine maybe winning/drawing but about to run out of time to the point that without adjudication it would lose ?
An adjudication could be "incorrect" if both engines agree on a winning score for one side but lack endgame knowledge to see that the position is in fact a draw. Example: an endgame KBP-KPP can't be won due to a "blind bishop" so that Black could lose both of its pawns and would still hold a draw.
And an adjudication could also be "inappropriate" if it is based on EGT knowledge but one or both engines are not configured to use an EGT so the actual result when playing on the board could very well be different from the theoretical result. Example: KQ-KR where the rook side uses an EGT but the queen side doesn't so it is certainly possible that the queen side is unable to win within 50 moves against perfect defense. Another example might be a position KRBPPP-KRPPP where it is difficult to find the winning plan and the side with the bishop does not make progress, but the evaluation of both engines says +5 for some reason.
My experience, though, is that both cases do not appear very often so they do not have any significant impact on the overall testing result when playing many test games. Also I think that such "incorrect" or "inappropriate" adjucation cases would occur equally often for both sides so that would be another reason for having a low statistical impact.
-
- Posts: 234
- Joined: Sat Jan 17, 2015 11:54 pm
Re: Adjudication & engine testing ?
thank you both , one more "unrelated" thing , I have a 5820k processor , do I need to disable hyper threading to get more accurate results "I intend to use 5 cores for testing"?
-
- Posts: 2204
- Joined: Sat Jan 18, 2014 10:24 am
- Location: Andorra
Re: Adjudication & engine testing ?
I have also an 5820K and I use the 6 cores, leaving hyper threading active. In my experience there is no problem at all. I don't see any special distortions.MahmoudUthman wrote:thank you both , one more "unrelated" thing , I have a 5820k processor , do I need to disable hyper threading to get more accurate results "I intend to use 5 cores for testing"?
Daniel José - http://www.andscacs.com