Author: Don Dailey
Date: 15:10:01 09/05/98
Go up one level in this thread
On September 03, 1998 at 01:53:20, Ed Schröder wrote:
>>>>(1) Ed Schröder's idea above sounds like it may be the key: Select the best
>>>>"anti-human" playing style.
>
>>>>(2) Now, what remains to be said is what, specifically, we mean by a
>>>>"playing
>>>>style," and to identify, define, and itemize the specific elements of the
>>>>best
>>>>"anti-human playing styles." To be complete, it also should be said: how
>>>>and
>>>>why those specific elements contribute to the strength of the engine against
>>>>humans.
>
>>>Impossible to give a definition. What's in Rebel is mainly balanced and well
>>>tuned stuff through the years. Rebel doesn't take huge risks except for some
>>>specific king attack patterns (when the king is forced to enter the 3th
>>>row or
>>>above). If you enter the well known 1.e4 Nf6 2.Bc4 Nxe4 3.Bxf7+ Kxf7 4.Qh5+
>>>Rebel will immediately pick 4..Kg8 as best move. I think this kind of
>>>knowledge is crucial playing strong human players.
>
>>>I also believe that the playing style of a chess program is strongly related
>>>to the goals of the programmer. Possibilities:
>>>a) Make the best program against computers;
>>>b) Make the best program against humans;
>>>c) Make the program the best finder in tactics;
>>>d) More?
>
>>>In case of (a) auto232 results are decisive. In case of (b) the "good
>>>feeling" of a chess programmer about a version is decisive. Case
>>>(c) is self-understood.
>
>>>Of course all programmers have their own priorities.
>>>For Rebel this since day one has been:
>>>(a) 20%
>>>(b) 60%
>>>(c) 20%
>
>>>- Ed -
>
>
>>But this is all just restating what you believe in. You do nothing to convince
>>someone who is undecided on this point. How do you propose to prove this ?
>
>>There are several programs who are about equal to Rebel in comp-comp
>>competition. Is Rebel significantly stronger than them against humans ? It
>>should be possible to demonstrate this.
>
>You can't be serious. I impossibly can contribute statistics that should prove
>that program X is +100 elo when playing humans and Program Y is -100 elo
>when playing humans based on their comp-comp (elo) results. I am not in
>the right position for that.
>
>All I can do is talk about my own program. In this respect I quote a recent
>example.
>
> "I can report a similar case on the last Aegon 1997. A pre-version of
> Rebel9 had a horrible deep bug (detected after the tournament when I
> suddenly got worse comp-comp results). Still this version gained a
> rating of 2619, scored 4.5 out of 6 playing against 4 x IM and 2 x GM."
>
>Here is another remarkable example.
>
>Have a look at:
>http://www.rebel.nl/jeroen2.htm (and)
>http://www.rebel.nl/jeroen3.htm
>
>Here the "Mephisto Chess Challenger" played two 30:00 games against
>Anatoly Karpov and 2 draws was the result. I would say this is a very
>good result more if you know that in one of the games Karpov was offered
>a draw while his flag was almost fallen. Very sportsmanship of the
>producers (Hegener & Glaser) who organized the event.
>
>The "Mephisto Chess Challenger" was a product for the german mass
>market basically Rebel8 outlook with the "Rebel Gold" chess engine. This
>Rebel Gold chess engine is estimated 50 elo points stronger than the old
>free Rebel Decade 1.2 and 150 elo points weaker than its big brother
>Rebel8.
>
>Coming to my point. The Rebel Gold engine will be crushed by all current
>top Pc chess programs but it drew 2 times against a super GM like Karpov!
>
>Like Rebel Decade the Rebel Gold engine is a lowered playing strength
>version of the real Rebel with about 150-200 elo points based on comp-
>comp auto232 results by removing tactical stuff from the real Rebel. So
>Rebel's playing style remains 100% intact!
>
>For me this was another strong indicator that "comp-comp" is a complete
>different story than "Man vs Machine" and it's the program's playing style
>that does the trick.
>
>Next I could add Rebel's best overall performance at Aegon in the period
>of 1991-1997. Then Rebel-Yusupov and Rebel-Anand.
>
>Now it is not so difficult to produce similar statistics for other programs. As
>said I am not in the right position to do this. But if some people do you will
>see remarkable fluctuations in respect to comp-comp rating lists.
>
>- Ed -
>
>
>>Amir
I have always been interested in this question of intransitivity
between humans and chess programs. A lot of people claim, almost
always based on some anecdotal evidence, that various programs
are much better or weaker against humans or other programs and
that huge intrasitivities exist. My own thinking is that this
is minor, but I know I'll get a lot of disagreement here.
I can't tell you how many times I have heard this type of
conversation: "I played a game with program
X and got crushed, then I played a game with program Y and
won easily. But when I played the two programs together,
program Y demolished program X." From a measly 3 games
can you come to the conclusion that program X is not very
strong against other programs but is "crushing" against
humans?
But once you form a conclusion, then you start noticing the
events that reinforce your conlusion and you minimize the
events that do not. Usually, the opinion propogates to other
people if it gets stated enough times.
Here are the ones I have heard but have serious doubts whether
they are true, or at least crystal clear:
1. Novag machines a much better against people but no good
against other programs.
2. Genius is not very good against people but crushes other
computers.
3. Genius is particularly good aginst people, but just so-so
against other programs (yes, I've heard both cases stated
as facts.)
4. Deep Blue will crush any computer but is not much better
that micro programs against humans.
5. Any kind of forward prunning or selectivity will help a
lot against humans but is not so good against computers.
6. Same as 5 but reversed.
- Don
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.