Author: Christophe Theron
Date: 18:33:36 02/01/00
Go up one level in this thread
On February 01, 2000 at 20:46:35, Chessfun wrote:
>On February 01, 2000 at 20:36:28, Dann Corbit wrote:
>
>>On February 01, 2000 at 20:27:01, Chessfun wrote:
>>>On February 01, 2000 at 19:30:29, Dann Corbit wrote:
>>>>On February 01, 2000 at 19:19:58, Côme wrote:
>>>>>On February 01, 2000 at 19:09:18, Enrique Irazoqui wrote:
>>>>>
>>>>>>Junior 6a won 12-8, +8 -4 =8. Games will be posted at
>>>>>>http://www.computerschach.de/tourn/cad00.htm
>>>>>>
>>>>>>Next match, Century - Hiarcs 7.32.
>>>>>>
>>>>>>Enrique
>>>>>
>>>>>Hello Enrique,
>>>>>Great Tournament !
>>>>>Few months ago CT destroy all this opponent, but now the new king seem to be
>>>>>Junior 6a !
>>>>>What do you think Enrique?
>>>>
>>>>I suspect that you are reading a lot more into this tournament than is possible.
>>>>Junior (by the SSDF) indeed looks very strong. But this tournament was at
>>>>fairly short time controls. Hence, the data is not at all commensurate with
>>>>SSDF testing [Though I am not sure you implied that].
>>>
>>>
>>>According to both Enrique and Christophe the 40"/40 40"/40 40 as used
>>>in this tourney is just as suitable a time control as those used by the
>>>SSDF, to measure the strength of a program. Christophe stated 60 game,
>>>therefore I am somewhat extracting that he would also see these times
>>>the same way.
>>
>>Any time control is suitable to measure the strength of a program. Blitz,
>>lightning, Postal (24 Hrs/move or even longer), whatever. But 40 moves in 40
>>minutes is a full ply shorter than 40/2. In other words, you are measuring
>>different things. It could be especially salient if extensions kick in at some
>>point in the analysis.
>>
>>The point I think I wanted to make, but failed completely to do so, was that a
>>1000:0 result at 40/2 and a 0:1000 result at 40/40 could be genuine aritfacts of
>>how the programs play (however unlikely).
>>
>>People sometimes see a short set of games at some time control very different
>>from the SSDF results and imagine that there is a refutation, confirmation or
>>otherwise in connection to the SSDF results. In reality, it's none of the
>>above. Apples compared to oranges.
>
>
>It is possible I could also expressed better. The point I was
>trying to make was the time control 40"/40 40"/40 40 is far from
>-short- and that the results produced at these time controls should
>produce results similar to the SSDF results after a similar # of games.
>Thanks.
I agree with you, Chessfun.
More generally, I don't care about the time controls used to test my program. I
accept any time control, and if I see that Tiger does not do well at a given
time control, I'll work to improve it. Even if it is 1mn/game. Even if it is
1mn/game on a 386.
Christophe
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.