The Diep Home page (more correction needed)
Ed Schröder
Vincent,
On your home page I find the following text:
[ For all chessprograms worth buying search is tremendeously important.
The reason: the deeper your lookahead (ply depth = half move depth),
the more tactical things you see, and the deeper the positional
evaluation is done. As most programs (for example: Fritz, Kallisto,
Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
knowledge so heavily on piece-square tables and cheap knowledge, not
surprisingly a deeper search doesn't bring that much more positional
insight to these programs. ]
Speaking only for Rebel your claim that Rebel "depend for their knowledge
so heavily on piece-square tables and cheap knowledge" is *NOT* true!
I estimate that my pieces-square knowledge is about 2-3% of Rebel 8.0
total chess knowledge, so how can you say such silly things?
Did you ever asked me, or do you have my source code or so?
Can you please change this *untrue* text on your home page?
Thanks in advance.
- Ed Schroder -
Author of Rebel 8.0
Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
these programs do have a lot of chess knowledge too!
Vincent, your are talented programmer but you have a terrible image here.
Think before you write something in public!
Copy sent by email as well posted RGCC.
Ed Schröder
Ed Schröder
I don't know what's wrong here.
I am 100% sure I posted this message once, still it appears twice.
Sorry.
- Ed -
Vincent Diepeveen
>Vincent,
>
>On your home page I find the following text:
>
> [ For all chessprograms worth buying search is tremendeously important.
> The reason: the deeper your lookahead (ply depth = half move depth),
> the more tactical things you see, and the deeper the positional
> evaluation is done. As most programs (for example: Fritz, Kallisto,
> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
> knowledge so heavily on piece-square tables and cheap knowledge, not
> surprisingly a deeper search doesn't bring that much more positional
> insight to these programs. ]
>
>Speaking only for Rebel your claim that Rebel "depend for their knowledge
>so heavily on piece-square tables and cheap knowledge" is *NOT* true!
Sorry, i removed Rebel.
>I estimate that my pieces-square knowledge is about 2-3% of Rebel 8.0
>total chess knowledge, so how can you say such silly things?
I mean next:
There are in my view different ways to implement knowledge
a) not
b) knowledge in Preprocessor: parameter adjustment and filling piece square
tables (which does not slow down, but doesn't evaluate leaf nodes).
c) knowledge in evaluation function (which slows down program, but takes
care you evaluate leave nodes).
d) a combination of them all.
In Diep i mainly use c). the faster a program is, the more it usually will
depend on b. It is clear that Genius,Fritz,Nimzo,The King are type b.
The King as an exception however uses mobility extensively in its evaluation,
but no more than that.
An easy test to see how dependant your program is on piece square tables
is simply put them all on 0, then recompile.
>Did you ever asked me, or do you have my source code or so?
>Can you please change this *untrue* text on your home page?
>
>Thanks in advance.
I changed it. Sorry, i didn't know that Rebel was so quickly evaluating
all the positional factors. Why doesn't it know when to exchange
a bad bishop against a knight in order to get strong knight, or just to
get rid of the bad bishop?
Why doesn't it know when to exchange queens and when not?
Does it know a thing about bishop mobility?
I'm afraid this formed the reason i concluded the above.
I changed my homepage for this reason and removed Rebel from the list.
>- Ed Schroder -
>Author of Rebel 8.0
>
>Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
>these programs do have a lot of chess knowledge too!
No way. Hiarcs does dynamically evaluate, which is a good thing,
but positionally very little. Mchess
is mainly busy with code to select moves, not with positional factors.
It derives them from his tables.
>Vincent, your are talented programmer but you have a terrible image here.
>Think before you write something in public!
Ok, i removed Rebel from the list.
Perhaps you could tell something about Rebel, what kind of knowledge
it has? I wonder. I guess you wrote something about Rooks? Mobility?
do you have a special pawnendings evaluation-function?
A special rook-endings function?
>Copy sent by email as well posted RGCC.
(Thanks, if you email me the chance you get a reply i quite huge.
If you just post in RGCCC, and i miss article ( i cannot read all crap,
sorry), after a week it is deleted
on the University, so i cannot read it anymore, and never will know there
was an article someone wanted me to respond on.)
Vincent
--
+----------------------------------------------------+
| Vincent Diepeveen email: vdie...@cs.ruu.nl |
| http://www.students.cs.ruu.nl/~vdiepeve/ |
+----------------------------------------------------+
Vincent Diepeveen
>I don't know what's wrong here.
>
>I am 100% sure I posted this message once, still it appears twice.
>Sorry.
>- Ed -
A problem could be that you post your name like Ed Schr o(met puntjes) der
Non-PC programs don't have an extended ascii set. So if you spell
your name different: Ed Schroeder or Ed Schroder,
i guess this will nothappen again.
It hear you mail a lot in newsgroup, but most of the msg's i cannot read
as you use extended ascii.
i got a copy by email 2 times, so its better to send it 2 times than
none...
Mark Uniacke
In article <85388828...@gln01-12.dial.xs4all.nl>,
Ed Schröder <rebc...@xs4all.nl> wrote:
>Vincent,
>On your home page I find the following text:
> [ For all chessprograms worth buying search is tremendeously important.
> The reason: the deeper your lookahead (ply depth = half move depth),
> the more tactical things you see, and the deeper the positional
> evaluation is done. As most programs (for example: Fritz, Kallisto,
> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
> knowledge so heavily on piece-square tables and cheap knowledge, not
> surprisingly a deeper search doesn't bring that much more positional
> insight to these programs. ]
>
[some of Ed's response cut due to server rules]
>
>Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
>these programs do have a lot of chess knowledge too!
That is right. HIARCS relies *very* little on its piece-square tables.
It is maybe as low as 1% of the knowledge!
The rest of the knowledge is by no means cheap by comparision which is
primarily why Hiarcs 5.0 can only search about 5500 nps on a P100.
I too would like you to correct the text on your home page.
Best wishes,
Mark
Author of Hiarcs
--
The opinions and comments expressed herein are my own and do not in anyway
represent those of BNR Europe or Northern Telecom.
John Stanback
Ed Schröder wrote:
>
> Vincent,
>
> On your home page I find the following text:
>
> [ For all chessprograms worth buying search is tremendeously important.
> The reason: the deeper your lookahead (ply depth = half move depth),
> the more tactical things you see, and the deeper the positional
> evaluation is done. As most programs (for example: Fritz, Kallisto,
> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
> knowledge so heavily on piece-square tables and cheap knowledge, not
> surprisingly a deeper search doesn't bring that much more positional
> insight to these programs. ]
>
> so heavily on piece-square tables and cheap knowledge" is *NOT* true!
>
It is also true that Zarkov makes very little use of piece-square
tables.
Almost all of the evaluation is determined by static analysis done at
each node in the search where the score could potentially lie within
the alpha-beta window.
John
Stefan Meyer-Kahlen
On 22 Jan 1997 14:14:51 GMT, vdie...@cs.ruu.nl (Vincent Diepeveen)
wrote:
...
>I changed it. Sorry, i didn't know that Rebel was so quickly evaluating
>all the positional factors. Why doesn't it know when to exchange
>a bad bishop against a knight in order to get strong knight, or just to
>get rid of the bad bishop?
>Why doesn't it know when to exchange queens and when not?
>Does it know a thing about bishop mobility?
>Perhaps you could tell something about Rebel, what kind of knowledge
>it has? I wonder. I guess you wrote something about Rooks? Mobility?
>do you have a special pawnendings evaluation-function?
>A special rook-endings function?
Oh yes, I'd like to know more details about Rebel, too.
Come on Mr. Schroeder, send Vincent and me your source code,
so we can check how you do it :-)
My email address is : mey...@fmi.uni-passau.de :-)
Stefan
Vincent Diepeveen
>That is right. HIARCS relies *very* little on its piece-square tables.
>It is maybe as low as 1% of the knowledge!
You tested it?
>The rest of the knowledge is by no means cheap by comparision which is
>primarily why Hiarcs 5.0 can only search about 5500 nps on a P100.
I'll remove Hiarcs from the fast searching programs list!
Sorry,
i read something you wrote about your program, and i concluded
you mainly described tactical evaluation code and tactical extensions,
and very few and very rude positional
things (from which most had to do with mobility, very little points
for rook+bishop mobility and way too much in my opinion
correction for being the side to move), so i assumed you didn't
implement much more, after i ran probably positions that don't fall under
Class C at Hiarcs 5 (but as i don't own Hiarcs, i can not test it
night and day like Rebel), and furthermore based on games Diep autoplayed
against Hiarcs 5 (diep scoring 40%!, although it were very few games,
both on fast hardware PP200/16 - P166/32, Diep mainly managed the many
draws because Hiarcs misjudged positionally things in middlegame, and
usually came back in ending).
>I too would like you to correct the text on your home page.
Done.
Still i am wondering whether you improve the existing amount of knowledge;
how many positional patterns/phenomena's you add to the dynamic knowledge
(so the knowledge in your evaluation) every month?
>Best wishes,
> Mark
>Author of Hiarcs
>
>--
>The opinions and comments expressed herein are my own and do not in anyway
>represent those of BNR Europe or Northern Telecom.
Vincent Diepeveen
In <32E6E4...@verinet.com> John Stanback <j...@verinet.com> writes:
True, my definition was not really correct.
A lot of todays programs
depend more on static analysis of each node than previously, but if you take
a close look, this is basically very few and very rude calculated information,
which in order to be prevented to be too rude usually is given very
small bonusses and very small penalties.
Even though little and rude, this information seems still very important
for the programs, noticing the reactions here.
BTW Zarkov does reasonably well in middlegame, but seems to loose grip
as soon as pieces are exchanged, or seems to misperform if it cannot castle
in opening.
Vincent
>John
Vincent
Tom C. Kerrigan
Vincent Diepeveen (vdie...@cs.ruu.nl) wrote:
> There are in my view different ways to implement knowledge
> a) not
> b) knowledge in Preprocessor: parameter adjustment and filling piece square
> tables (which does not slow down, but doesn't evaluate leaf nodes).
> c) knowledge in evaluation function (which slows down program, but takes
> care you evaluate leave nodes).
> d) a combination of them all.
> In Diep i mainly use c). the faster a program is, the more it usually will
> depend on b. It is clear that Genius,Fritz,Nimzo,The King are type b.
> The King as an exception however uses mobility extensively in its evaluation,
> but no more than that.
Please tell me how you know this.
Placing Genius and The King in the "b" category runs against everything I
know about these programs.
I would like to think you asked the authors, but I have every reason to
believe they wouldn't answer such a question from you.
Also, Stobor is one of the faster programs, and it only has two static,
symmetric piece/square tables (for the knight and bishop). I doubt I'm the
only exception to your little rule of thumb.
Cheers,
Tom
MCHESS PRO
Vincent Diepeven wrote:
>Mchess is mainly busy with code to select moves, not with positional
factors. It >derives them from his tables.
This is, simply speaking, bullshit.
At leaf nodes MChess does a detailed analysis of passed pawns, pawn
structure and king safety and considers the mobility of every piece,
numerous positional factors, and "threat potential". This is further
supplemented by "expert system" evaluations of specific endgames.
M-Chess Pro 6.0 even allows the end-user to modify the weightings of the
evaluation factors, all of which are calculated at the leaf nodes.
Mr. Diepeven, please delete your misleading comments from all on-line
publications including your home page.
And please refrain from broadcasting such outlandish bullshit in the
future.
-Marty Hirsch
mclane
muni...@bnsgh724.bnr.co.uk (Mark Uniacke) wrote:
>In article <85388828...@gln01-12.dial.xs4all.nl>,
>Ed Schröder <rebc...@xs4all.nl> wrote:
>>Vincent,
>>On your home page I find the following text:
>> [ For all chessprograms worth buying search is tremendeously important.
>> The reason: the deeper your lookahead (ply depth = half move depth),
>> the more tactical things you see, and the deeper the positional
>> evaluation is done. As most programs (for example: Fritz, Kallisto,
>> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
>> knowledge so heavily on piece-square tables and cheap knowledge, not
>> surprisingly a deeper search doesn't bring that much more positional
>> insight to these programs. ]
>>
>[some of Ed's response cut due to server rules]
>>
>>Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
>>these programs do have a lot of chess knowledge too!
>That is right. HIARCS relies *very* little on its piece-square tables.
>It is maybe as low as 1% of the knowledge!
>The rest of the knowledge is by no means cheap by comparision which is
>primarily why Hiarcs 5.0 can only search about 5500 nps on a P100.
5500 nodes per second !!! Thanks for the info mark.
>I too would like you to correct the text on your home page.
I would like to see Vincent correct his page too concerning Hiarcs.
For me it looks as if he doesn't know much about hiarcs at all.
mclane
vdie...@cs.ruu.nl (Vincent Diepeveen) wrote:
>I changed my homepage for this reason and removed Rebel from the list.
Good done!
>>Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
>>these programs do have a lot of chess knowledge too!
>No way. Hiarcs does dynamically evaluate, which is a good thing,
>but positionally very little.
HAHAHA !
> Mchess
>is mainly busy with code to select moves, not with positional factors.
>It derives them from his tables.
>
The select-mechanism in Mchess and the tree-mechanism in Genius have
not changed much over the years. In both programs the 2 features are
the main-strengths of the programs.
They have tuned the features, and mchess today is much more
speculative than old-mchess. But still mchess has much knowledge.
I am not conform with your decision of putting hiarcs and mchess into
this group. We should ask marty and mark about how much knowledge is
in the programs.
>>Vincent, your are talented programmer but you have a terrible image here.
>>Think before you write something in public!
>Ok, i removed Rebel from the list.
>Perhaps you could tell something about Rebel, what kind of knowledge
>it has? I wonder. I guess you wrote something about Rooks? Mobility?
>do you have a special pawnendings evaluation-function?
>A special rook-endings function?
>>Copy sent by email as well posted RGCC.
>(Thanks, if you email me the chance you get a reply i quite huge.
>If you just post in RGCCC, and i miss article ( i cannot read all crap,
>sorry), after a week it is deleted
>on the University, so i cannot read it anymore, and never will know there
>was an article someone wanted me to respond on.)
>Vincent
mclane
"Ed Schröder" <rebc...@xs4all.nl> wrote:
>Vincent,
>On your home page I find the following text:
> [ For all chessprograms worth buying search is tremendeously important.
> The reason: the deeper your lookahead (ply depth = half move depth),
> the more tactical things you see, and the deeper the positional
> evaluation is done. As most programs (for example: Fritz, Kallisto,
> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
> knowledge so heavily on piece-square tables and cheap knowledge, not
> surprisingly a deeper search doesn't bring that much more positional
> insight to these programs. ]
Dear Vincent: Mchess, Hiarcs, The King, Rebel and Zarkov do not have a
cheap knowledge feature like you describe.
You are wrong.
>Speaking only for Rebel your claim that Rebel "depend for their knowledge
>so heavily on piece-square tables and cheap knowledge" is *NOT* true!
>I estimate that my pieces-square knowledge is about 2-3% of Rebel 8.0
>total chess knowledge, so how can you say such silly things?
Because he is louwmans loudspeaker he can only repeat what louwman
tells him to speak loud.
>Did you ever asked me, or do you have my source code or so?
>Can you please change this *untrue* text on your home page?
>Thanks in advance.
>- Ed Schroder -
>Author of Rebel 8.0
>Perhaps you can also remove this stuff for The King, Mchess, Hiarcs since
>these programs do have a lot of chess knowledge too!
Show us games between Hiarcs5 / Mchess6 / Rebel8 where Diep shows more
knowledge, but don't lie always the same lies in public.
>Vincent, your are talented programmer but you have a terrible image here.
>Think before you write something in public!
>Copy sent by email as well posted RGCC.
mclane
mey...@fmi.uni-passau.de (Stefan Meyer-Kahlen) wrote:
>On 22 Jan 1997 14:14:51 GMT, vdie...@cs.ruu.nl (Vincent Diepeveen)
>wrote:
>...
>>I changed it. Sorry, i didn't know that Rebel was so quickly evaluating
>>all the positional factors. Why doesn't it know when to exchange
>>a bad bishop against a knight in order to get strong knight, or just to
>>get rid of the bad bishop?
>>Why doesn't it know when to exchange queens and when not?
>>Does it know a thing about bishop mobility?
>...
>>Perhaps you could tell something about Rebel, what kind of knowledge
>>it has? I wonder. I guess you wrote something about Rooks? Mobility?
>>do you have a special pawnendings evaluation-function?
>>A special rook-endings function?
>Oh yes, I'd like to know more details about Rebel, too.
>Come on Mr. Schroeder, send Vincent and me your source code,
>so we can check how you do it :-)
>My email address is : mey...@fmi.uni-passau.de :-)
>Stefan
It is very difficult to find out if a program uses much knowledge or
only good tuned little knowledge.
The King has not much knowledge. But it used much time for finding out
about space/room and overprotection. Hiarcs has much knowledge.
Rebel has good tuned knowledge to decide what to prune.
Mchess-mechanism to select moves is very smart, genius
asymetric-search is very smart, ....
I am interested to find out how shredder does it.
Can you tell us more about Shredder, Stefan ???
How have YOU done it ????
Some small functions ? Good tuned ?
Or many functions , massive and many lines ?`
mclane
John Stanback <j...@verinet.com> wrote:
>Ed Schröder wrote:
>>
>> Vincent,
>>
>> On your home page I find the following text:
>>
>> [ For all chessprograms worth buying search is tremendeously important.
>> The reason: the deeper your lookahead (ply depth = half move depth),
>> the more tactical things you see, and the deeper the positional
>> evaluation is done. As most programs (for example: Fritz, Kallisto,
>> Nimzo, The King, Mchess, Zarkov, Hiarcs, Rebel) depend for their
>> knowledge so heavily on piece-square tables and cheap knowledge, not
>> surprisingly a deeper search doesn't bring that much more positional
>> insight to these programs. ]
>>
>> Speaking only for Rebel your claim that Rebel "depend for their knowledge
>> so heavily on piece-square tables and cheap knowledge" is *NOT* true!
>>
>It is also true that Zarkov makes very little use of piece-square
>tables.
>Almost all of the evaluation is determined by static analysis done at
>each node in the search where the score could potentially lie within
>the alpha-beta window.
>John
Hello John, I still have Zarkov3.
What is the latest Zarkov I could update or buy ?!
Is there a release sooner or later ?!
Is the program working with Saitek-chess-computers like
Leonardo/Renaissance ?!
Thanks in advance.
Komputer Korner
MCHESS PRO wrote:
>
> Vincent Diepeven wrote:
>
> >Mchess is mainly busy with code to select moves, not with positional
> factors. It >derives them from his tables.
>
>
> At leaf nodes MChess does a detailed analysis of passed pawns, pawn
> structure and king safety and considers the mobility of every piece,
> numerous positional factors, and "threat potential". This is further
> supplemented by "expert system" evaluations of specific endgames.
>
> M-Chess Pro 6.0 even allows the end-user to modify the weightings of the
> evaluation factors, all of which are calculated at the leaf nodes.
>
> Mr. Diepeven, please delete your misleading comments from all on-line
> publications including your home page.
>
> And please refrain from broadcasting such outlandish bullshit in the
> future.
>
> -Marty Hirsch
Marty, it seems that Vincent is being so outlandish just to get
you guys to divulge tidbits about your program. So far he has been
successful in this, but the information he learned is not worth the
cost to his reputation. BTW, am still waiting to hear from you about
your opening book editor.
--
Komputer Korner
The komputer that kouldn't keep a password safe from
prying eyes, kouldn't kompute the square root of 36^n,
kouldn't find the real Motive and variation tree in
ChessBase, missed the Hiarcs functionality in Extreme
and also misread the real learning feature of Nimzo.
Chris Whittington
--
http://www.demon.co.uk/oxford-soft
Stefan Meyer-Kahlen <mey...@fmi.uni-passau.de> wrote in article
<32e873fd...@nntpserver.uni-passau.de>...
>
> >Can you tell us more about Shredder, Stefan ???
> >
> >How have YOU done it ????
>
> >Some small functions ? Good tuned ?
> >Or many functions , massive and many lines ?`
>
> The source code of Shredder contains several rather big evaluation
> fuctions.
>
> Hope this helps
>
So its a quite big thing; made by putting some rather big things into it ?
Well, glad you made that clear then :)
Chris Whittington
Vincent Diepeveen
>Vincent Diepeveen (vdie...@cs.ruu.nl) wrote:
>
>> There are in my view different ways to implement knowledge
>> a) not
>> b) knowledge in Preprocessor: parameter adjustment and filling piece square
>> tables (which does not slow down, but doesn't evaluate leaf nodes).
>> c) knowledge in evaluation function (which slows down program, but takes
>> care you evaluate leave nodes).
>> d) a combination of them all.
>> In Diep i mainly use c). the faster a program is, the more it usually will
>> depend on b. It is clear that Genius,Fritz,Nimzo,The King are type b.
>> The King as an exception however uses mobility extensively in its evaluation,
>> but no more than that.
>
>Please tell me how you know this.
a) 100% for sure it will be never. It could of course be the case that
i myself simply tested positions where both programs mismatched.
A very clear example: the game Diep-The King DCCC 1995.
The King without any reason took a weak bishop of white from the board.
white pawn structure: c4,d5,e4,f2,g2,h2
black pawn structure: a5,b4,e5,f7,g7,h7
Black had a Knight on c5 and white a bishop on d3, which was covered by
the queen on d1.
>Placing Genius and The King in the "b" category runs against everything I
>know about these programs.
I don't place them in a category saying that the programs play weak.
To see their tournament performance just look at the SSDF list and you
will see that they both play very strong.
In case of Genius, it is really very good from the program that it
not quickly plays a bad move. For years Genius was the only program
that could do this. Genius always scores well, no matter the opposition.
That is of course a great thing of a program.
>I would like to think you asked the authors, but I have every reason to
>believe they wouldn't answer such a question from you.
Did they every tell you more than that there program could play reasonable
chess?
I say a lot, i know, a lot of the people don't think it good, but i
don't make too much secrets about what kind of search algorithms i have
implemented, nor do i make secrets about the way i implement knowledge,
and the size of that knowledge.
>Also, Stobor is one of the faster programs, and it only has two static,
>symmetric piece/square tables (for the knight and bishop). I doubt I'm the
>only exception to your little rule of thumb.
The only thing about piece-square tables i think will become true is that
it will at a certain depth (perhaps 2400) not that much anymore.
>Cheers,
>Tom
Vincent.
Vincent Diepeveen
>Vincent Diepeven wrote:
>
>>Mchess is mainly busy with code to select moves, not with positional
>factors. It >derives them from his tables.
>
>This is, simply speaking, bullshit.
Sorry, i corrected it.
>At leaf nodes MChess does a detailed analysis of passed pawns, pawn
>structure and king safety and considers the mobility of every piece,
>numerous positional factors, and "threat potential". This is further
>supplemented by "expert system" evaluations of specific endgames.
Does Mchess search brute force, does it use nullmove?
>M-Chess Pro 6.0 even allows the end-user to modify the weightings of the
>evaluation factors, all of which are calculated at the leaf nodes.
How much evaluation factors are there in Mchess for the dynamic evaluation?
>Mr. Diepeven, please delete your misleading comments from all on-line
You seems to spell my name like i told something about your program.
>publications including your home page.
>And please refrain from broadcasting such outlandish bullshit in the
>future.
>-Marty Hirsch
Vincent
Vincent Diepeveen
In <E4HrE...@news.prima.ruhr.de> mcl...@prima.ruhr.de (mclane) writes:
>Because he is .. loudspeaker he can only repeat what ..
>tells him to speak loud.
This is not true of course.
I NEVER INSULTED YOU PERSONALLY mclane. I never intended to insult
anyone personally from something. If i talk about a program, then i talk
about a program, and not about the person that programmed it.
If i say something, then i say something MYSELF.
Perhaps Rolf Tueschen could talk with you, he's having a profession
that is there especially made to help people like you.
mclane
vdie...@cs.ruu.nl (Vincent Diepeveen) wrote:
>In <5c59g2$e...@bcrkh13.bnr.ca> muni...@bnsgh724.bnr.co.uk (Mark Uniacke) writes:
>>That is right. HIARCS relies *very* little on its piece-square tables.
>>It is maybe as low as 1% of the knowledge!
>You tested it?
As the programmer he should KNOW it. He has the source code in mind.
>>The rest of the knowledge is by no means cheap by comparision which is
>>primarily why Hiarcs 5.0 can only search about 5500 nps on a P100.
>I'll remove Hiarcs from the fast searching programs list!
Brilliant decision !
>Sorry,
You see: also Vincent is able to react like a human beeing.
>i read something you wrote about your program, and i concluded
>you mainly described tactical evaluation code and tactical extensions,
>and very few and very rude positional
>things (from which most had to do with mobility, very little points
>for rook+bishop mobility and way too much in my opinion
>correction for being the side to move), so i assumed you didn't
>implement much more, after i ran probably positions that don't fall under
>Class C at Hiarcs 5 (but as i don't own Hiarcs, i can not test it
>night and day like Rebel), and furthermore based on games Diep autoplayed
>against Hiarcs 5 (diep scoring 40%!, although it were very few games,
>both on fast hardware PP200/16 - P166/32, Diep mainly managed the many
>draws because Hiarcs misjudged positionally things in middlegame, and
>usually came back in ending).
I am really interested in the games. I don't believe that DIEP plays
that strong against Hiarcs5 !
>>I too would like you to correct the text on your home page.
>Done.
>Still i am wondering whether you improve the existing amount of knowledge;
>how many positional patterns/phenomena's you add to the dynamic knowledge
>(so the knowledge in your evaluation) every month?
>>Best wishes,
>> Mark
>>Author of Hiarcs
>>
>>--
>>The opinions and comments expressed herein are my own and do not in anyway
>>represent those of BNR Europe or Northern Telecom.
John Stanback
mclane wrote:
>
>
> Hello John, I still have Zarkov3.
> What is the latest Zarkov I could update or buy ?!
> Is there a release sooner or later ?!
>
> Is the program working with Saitek-chess-computers like
> Leonardo/Renaissance ?!
>
> Thanks in advance.
Sorry, I haven't released a version since Zarkov3,
although I did make a "DDE server" version of Zarkov which
works with Bookup.
I've made some significant improvements in the engine and am gradually
putting together a new Windows95 interface. I don't know if I will
retain the capability to hook to the Saitek boards. I have no idea
when I'll actually have a version ready to sell...
The engine plays on ICC via WinBoard now (ZarkovX).
John
mclane
mey...@fmi.uni-passau.de (Stefan Meyer-Kahlen) wrote:
>>It is very difficult to find out if a program uses much knowledge or
>>only good tuned little knowledge.
>>
>>The King has not much knowledge. But it used much time for finding out
>>about space/room and overprotection. Hiarcs has much knowledge.
>>Rebel has good tuned knowledge to decide what to prune.
>>Mchess-mechanism to select moves is very smart, genius
>>asymetric-search is very smart, ....
>>
>>I am interested to find out how shredder does it.
>Shredder uses piece square tables, too, but also has a rather complex
>evaluation function. In average middlegame positions I spent more than
>50% of the CPU time in my leaf evaluation function, yes, I do lazy
>evaluation. On a PPro 200, the machine I used in Jakrata, Shredder
>searches about 40-60knps in the middlegame and up to 100knps in
>endgames.
>>Can you tell us more about Shredder, Stefan ???
>>
>>How have YOU done it ????
>>Some small functions ? Good tuned ?
>>Or many functions , massive and many lines ?`
>The source code of Shredder contains several rather big evaluation
>fuctions.
>
Aha ! An old flame of my school-time also contains several rather big
evaluation-functions.
>Hope this helps
Yes.
> Stefan
mclane
"Chris Whittington" <chr...@demon.co.uk> wrote:
>So its a quite big thing; made by putting some rather big things into it ?
Now you explained it to me, chris, it is much more clear to me.
I have to put quite big things into it !!! I will tell michaela about
this....
mclane
mche...@aol.com (MCHESS PRO) wrote:
>Vincent Diepeven wrote:
>>Mchess is mainly busy with code to select moves, not with positional
>factors. It >derives them from his tables.
>This is, simply speaking, bullshit.
>At leaf nodes MChess does a detailed analysis of passed pawns, pawn
>structure and king safety and considers the mobility of every piece,
>numerous positional factors, and "threat potential". This is further
>supplemented by "expert system" evaluations of specific endgames.
>M-Chess Pro 6.0 even allows the end-user to modify the weightings of the
>evaluation factors, all of which are calculated at the leaf nodes.
>Mr. Diepeven, please delete your misleading comments from all on-line
>publications including your home page.
>And please refrain from broadcasting such outlandish bullshit in the
>future.
>-Marty Hirsch
Maybe Vincent should concentrate on commenting or showing us facts
about HIS program instead of publishing strange NEWS about the
programs we know better than him.
mclane
vdie...@cs.ruu.nl (Vincent Diepeveen) wrote:
>In <19970123174...@ladder01.news.aol.com> mche...@aol.com (MCHESS PRO) writes:
>>Vincent Diepeven wrote:
>>
>>>Mchess is mainly busy with code to select moves, not with positional
>>factors. It >derives them from his tables.
>>
>>This is, simply speaking, bullshit.
>Sorry, i corrected it.
Brilliant. The guy learns. But how long does he remember ? 15 Minutes
?!
>>At leaf nodes MChess does a detailed analysis of passed pawns, pawn
>>structure and king safety and considers the mobility of every piece,
>>numerous positional factors, and "threat potential". This is further
>>supplemented by "expert system" evaluations of specific endgames.
>Does Mchess search brute force, does it use nullmove?
Hahahaha!!! You are brilliant naiv vincent. And you tell us that you
have analysed mchess ?!
What do you call analysis ? Reading the instruction booklet ?
>>M-Chess Pro 6.0 even allows the end-user to modify the weightings of the
>>evaluation factors, all of which are calculated at the leaf nodes.
>How much evaluation factors are there in Mchess for the dynamic evaluation?
MANY!!
Haven't you seen from the games against DIEP ?
mclane
vdie...@cs.ruu.nl (Vincent Diepeveen) wrote:
>In <E4HrE...@news.prima.ruhr.de> mcl...@prima.ruhr.de (mclane) writes:
>>Because he is .. loudspeaker he can only repeat what ..
>>tells him to speak loud.
>This is not true of course.
>I NEVER INSULTED YOU PERSONALLY mclane. I never intended to insult
>anyone personally from something. If i talk about a program, then i talk
>about a program, and not about the person that programmed it.
If you tell us that chess system tal I gave you in Aegon is bullshit
and I can play nice games with the same version against many
well-known programs, if you tell us Rebel8 has a bigger killer-book
than Mchess5, and don't show us any line... if you tell us from your
home-page that your program is the best and all the others have less
knowledge and later you have to retire because all these programmers
attack you because you have said something wrong...
but of course you never insult the people. Thats right.
But it does not count. As long as you say stupid things about the
programs you don't have to insult people. You insult programs.
>If i say something, then i say something MYSELF.
Brilliant. So do I.
>Perhaps Rolf Tueschen could talk with you, he's having a profession
>that is there especially made to help people like you.
>Vincent
The level of wrong stuff you posted and rolf has posted is
not to relate. Rolf made silly mistakes and attacked a little
paranoic.
You post wrong information : desinformation. And you are a programmer.
You should know it better. Rolf is just a newbie.
If you sent me a copy of your DIEP, I will give you the money later
when I see you in Den Haag again.
I will than let this DIEP play against
Hiarcs5
Chess System Tal
Stobor
Shredder
Virtual Chess Win95
Colossus Chess X
and post the results here in public, that anybody can see what we can
learn from your claims.
we will see if you (sorry - your program!) are able to win ONE SINGLE
GAME.
I don't think so.
Tom C. Kerrigan
Vincent Diepeveen (vdie...@cs.ruu.nl) wrote:
> >> In Diep i mainly use c). the faster a program is, the more it usually will
> >> depend on b. It is clear that Genius,Fritz,Nimzo,The King are type b.
> >> The King as an exception however uses mobility extensively in its evaluation,
> >> but no more than that.
> >Please tell me how you know this.
> a) 100% for sure it will be never. It could of course be the case that
> i myself simply tested positions where both programs mismatched.
I have no idea what this sentence is supposed to mean. Please repeat it
with different words.
> >Placing Genius and The King in the "b" category runs against everything I
> >know about these programs.
> I don't place them in a category saying that the programs play weak.
And I never said they were weak, either.
> In case of Genius, it is really very good from the program that it
> not quickly plays a bad move. For years Genius was the only program
> that could do this. Genius always scores well, no matter the opposition.
This does not mean Genius has a high NPS rate and therefore little
knowledge (as you imply). It is possible (and more likely, from what I've
observed) that Genius uses an excellent selective search.
It seems you've taken every tactically strong program, assumed they are
fast, and told the world they have no knowledge. Sounds a bit silly when I
put it that way, now doesn't it?
> >I would like to think you asked the authors, but I have every reason to
> >believe they wouldn't answer such a question from you.
> Did they every tell you more than that there program could play reasonable
> chess?
Actually, yes, but I don't think I have a "right" to repeat what they said
in a public forum. Sorry.
Cheers,
Tom
mclane
John Stanback <j...@verinet.com> wrote:
Thanks for answering. I am looking forward to see it, I will instantly
buy it then.
>John
Komputer Korner
John Stanback wrote:
>
> mclane wrote:
> >
> >
> > Hello John, I still have Zarkov3.
> > What is the latest Zarkov I could update or buy ?!
> > Is there a release sooner or later ?!
> >
> > Is the program working with Saitek-chess-computers like
> > Leonardo/Renaissance ?!
> >
> > Thanks in advance.
>
> Sorry, I haven't released a version since Zarkov3,
> although I did make a "DDE server" version of Zarkov which
> works with Bookup.
>
> I've made some significant improvements in the engine and am gradually
> putting together a new Windows95 interface. I don't know if I will
> retain the capability to hook to the Saitek boards. I have no idea
> when I'll actually have a version ready to sell...
>
> The engine plays on ICC via WinBoard now (ZarkovX).
>
Hi John,
Can you tell us whether the DDE version that works with
Bookup has any changes to it or is it basically the same as
Zarkov 3?
--
Komputer Korner
The komputer that kouldn't keep a password safe from
prying eyes, kouldn't kompute the square root of 36^n,
kouldn't find the real Motive and variation tree in
ChessBase, kouldn't compute the proper time in 2 variation
mode, missed the Hiarcs functionality in Extreme
mclane
kerr...@merlin.pn.org (Tom C. Kerrigan) wrote:
>It seems you've taken every tactically strong program, assumed they are
>fast, and told the world they have no knowledge. Sounds a bit silly when I
>put it that way, now doesn't it?
Yes - this was, what he has done. But he got answer from the authors.
For heavens sake.
>> >I would like to think you asked the authors, but I have every reason to
>> >believe they wouldn't answer such a question from you.
>> Did they every tell you more than that there program could play reasonable
>> chess?
>Actually, yes, but I don't think I have a "right" to repeat what they said
>in a public forum. Sorry.
>Cheers,
>Tom
Right again.
Chris Whittington
--
http://www.demon.co.uk/oxford-soft
Robert Hyatt <hy...@cis.uab.edu> wrote in article
<5cc20m$b...@juniper.cis.uab.edu>...
> mclane (mcl...@prima.ruhr.de) wrote:
>
> : I would like to see Vincent correct his page too concerning Hiarcs.
> : For me it looks as if he doesn't know much about hiarcs at all.
>
>
>
> In fact, probably the only program most anybody really knows anything
about is
> Crafty, because you can get the source and see what I'm doing. All of
the stuff
> about what the others do is nothing more than idle speculation. If Marty
or
> Ed or whomever wants to discuss their program, more power to 'em... But
I don't
> see why anyone else has to start speculating about how "they must be
doing this
> or that" because only the authors know.
>
> I feel qualified to comment on the quality of play of most of the
programs, but
> I won't pretend to understand how/what they are doing, what is search,
and what
> is evaluation. There's nothing to gain from speculation except
misinformation...
Come on, there's a lot to be gained by speculation:
1. you share mad ideas with crazy people - this may elicit some truth
2. you wind up some people who know things, so they post what they know -
this also elicits some truth.
Long may idle speculation continue .....
(except when it comes from Vincent)
Chris Whittington
Harald Faber
quoting a mail from mclane # prima.ruhr.de
Hello mclane,
m> I would like to see Vincent correct his page too concerning Hiarcs.
m> For me it looks as if he doesn't know much about hiarcs at all.
It seems that he doesn't know about anything except for his own program.
:-)
Harald
--