Author: Christophe Theron
Date: 10:52:39 05/10/00
Go up one level in this thread
On May 10, 2000 at 10:52:19, Mark Andreoli wrote:
>
> I would like to _thank everyone_ for helping me try to solve the mystery
>of why a lot of engines i.e. Fritz and Hiarcs etc. didn't find the _BEST_ move
>Bf8.
>( posted yesterday)
>
> Here are some _not all_ of the results:
>
> 1. Ed Schroder found it on Rebel using an athlon 1000Mhz/EPD2DIAG?in 1:31
> Q. Congratulations and What is an "EPD2DIAG"?
>
> 2.His Century and Century cd rom version also had success!
>
> 3. John Stanback found the move in 9 ply on Zarkov "by luck?"using one extra
> ply in the capture/critical line!?
>
> 4. Dan Corbit had a nice conclusion that said "Bf8 was the GM move."
> Q. I hope you test other programs and let me know the results! Thanks
>
> 5. Enrique Irazoqui found that F6a,Tiger and Nimzo 7.32 will find the move
> after more then 10 minutes! more success!
>
> 6.Blass Uri states that Chessmaster will find it! although I am still a little
> confused (although not totally)on _null moves_!
>
> 7. I read __all__ the comments posted by everyone and all answers seem to be
>legitimate and in great detail! my compliments
>
>
> BUT HERE GOES: I NEED SOME _FEEDBACK_ON A FEW _NAIVE CONCLUSIONS_!? THANKS
>
> 1.I have heard time after time that "there is only one best move in any given
>position"(It makes sense /the one that wins the fastest) . In this case most of
>us agreed it was 19.Bf8! Ok my question is why
>didn't My Fritz 5 ,6 or Hiarcs find this move especially given the fact that F6
>is in the top 5 on the SSDF list?Are they weak analyzers?
>
> 2.How can it be that the above mentioned engines (F5/6/Hiarcs)manage to "Kick
>Butt" time and time again on other engines who found Bf8 but _they_ were unable
>to find it? So What engine is _really _ the strongest?
>
> 3. Given the fact that I do play a lot of tournament chess what _computer
>chess program_ (Will the truly best analyzing engine please _stand up_!) do I
>have to buy now to get a "second opinion" on my game positions. I still can't
>express how disappointed I am in F5/F6 and Hiarcs for not finding 19.Bf8!
>(remember Nenad Aleksic found it in only about 20 seconds over the board!)
>
> Thanks MJA //Comments will be greatly appreciated!!!
You have just discovered that it is a very bad idea to rely on a single position
to assess the strength of a player, or the strength of a program.
The idea of testing a player by letting him study a given set of interesting
positions is not bad. But you need to use several positions, preferably a lot of
positions, in order to make your mind.
In computer chess we call them "test suites". There are a number of well known
test suites, and some of them include a big number of positions.
One of the test suites I know is the LCT-II (Louguet Chess Test 2). It only has
35 positions. It was popular some years ago, because by relating the solution
times to a predefined set of points, it was possible to compute approximately
the ELO rating of a program in two or tree hours. But only 35 positions is not
enough.
There are also the well known WAC and BT-xxx test suites.
It is important to understand that chess (human or computer chess) is a matter
of statistics. There is no player that win all the time. But there are players
that win more often than others, and those are rated higher.
The same applies to computers. There is no program that will solve any tactical
problem, and no program that will understand any positional concept. But there
are programs that will solve more problems than others.
In statistics, the size of the sample is important. That's why it is unwise to
rely solely on the results you have got on this position.
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.