Tournament 40/40
Enrique Irazoqui
sense to keep playing in my tournament with the old and rather disappointing
one. Dropping it is not ideal, but I think it's more interesting to play all
programs at their best. The first four places in this tournament were
already decided and wouldn't have changed after playing the last 2 Crafty
matches:
1 Junior 5
2 Mchess 8
3 Hiarcs 7b
4 Tiger 11.7.5
This is the final crosstable of the not completely finished tournament:
PII-400/256, 40/40
J5 M8 H7b T11 R10 F5 N99 CR TOTAL % Elo
J5 --- 6 6 6.5 5 4.5 6 8.5 42.5/70 60.7% 70
M8 4 --- 6.5 5.5 5.5 5 8 7 41.5/70 59.3% 60
H7b 4 3.5 --- 4 7 8 7 7.5 41.0/70 58.6% 55
Ti 3.5 4.5 6 --- 6 5.5 6.5 7 39.0/70 55.7% 37
R10 5 4.5 3 4 --- 5.5 4.5 26.5/60 44.2% -16
F5 5.5 5 2 4.5 4.5 --- 5 26.5/60 44.2% -16
N99 4 2 3 3.5 5.5 5 --- 23.0/60 38.3% -53
C161 1.5 3 2.5 3 --- 10.0/40 25.0% -139
From now on, I will keep playing it with the new Nimzo 99. This is the
current situation, including the match being played between Junior 5 and
Fritz 532:
J5 H7b M8 T11 R10 F5 CR F532 TOTAL % Elo
J5 --- 6 6 6.5 5 4.5 8.5 3 39.5/65 60.8% 69
H7b 4 --- 3.5 4 7 8 7.5 34.0/60 56.7% 44
M8 4 6.5 --- 5.5 5.5 5 7 33.5/60 55.8% 39
Ti 3.5 6 4.5 --- 6 5.5 7 32.5/60 54.2% 30
R10 5 3 4.5 4 --- 5.5 22.0/50 44.0% -11
F5 5.5 2 5 4.5 4.5 --- 21.5/50 43.0% -17
C161 1.5 2.5 3 3 --- 10.0/40 25.0% -146
F532 2 --- 2.0/5 40.0% -10
The life of a computer chess freak is so difficult... :)
Enrique
Robert Hyatt
: Since there is a new chess engine for Nimzo 99, it doesn't seem to make much
: sense to keep playing in my tournament with the old and rather disappointing
: one. Dropping it is not ideal, but I think it's more interesting to play all
: programs at their best. The first four places in this tournament were
: already decided and wouldn't have changed after playing the last 2 Crafty
: matches:
One quick note... I just discovered that the "fritz interface" is really
disruptive to other engines. IE it apparently sends "new" followed by
*all* the moves in the game, for *each* move played. For crafty this is
a killer, because I depend on hash stuff carrying over from one search to
another, which it won't here. And I depend on move ordering stuff staying
around (killers/history counts/etc) and they get killed by the "new" as
well.
IE if you are using the engine-vs-engine mode, this looks to be not so
reliable... I had not noticed this until someone sent me some log
files that showed it quite clearly. Blitz games are even worse because
I end up clearing the hash before each move which can take time as well...
Bob
: 1 Junior 5
: PII-400/256, 40/40
: Enrique
--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
RDavis101
>From: Robert Hyatt <hy...@crafty.cis.uab.edu>
>Date: 12/15/98 4:32 PM Central Standard Time
>Message-id: <756o2r$oho$1...@juniper.cis.uab.edu>
Well...this is certainly unwelcome news, because it potentially means that the
playing field is not level for all engines in a tournament.
Worse, what are the implications for using multiple engines to analyze a game,
or other training functions of Fritz?
What if I buy the Junior 5 engine and play it in the Fritz interface... Is this
the same as getting the Junior 5 program? If not, why not?
I guess this means that each author needs to determine how destructive the
interface is to his engine, otherwise, we'll never know whether its analysis is
useful from inside the Fritz interface.
Isn't this a big issue for anyone who wants to use engine analysis to study the
game?
Roger
Robert Hyatt
:>Subject: Re: Tournament 40/40
: Roger
It is probably ok if it does this when you are only analyzing a game. The
real issue is when it does this when two engines are playing each other thru
the interface, which is becoming quite common. I can't speak for other
programmers, but *my* program isn't happy getting a new command before each
and every move. IE the interface is *far to intrusive* into the game and
breaks lots of assumptions I have made...
Amir Ban
This is how all versions and brands of Fritz5 and Junior5 work and have
always worked with all engines. The engine is given the entire game
history from scratch before every move, so it doesn't need to remember
anything. Same behavior for all interfaces.
This doesn't affect analysis at all, or practically anything else.
Remember the interface is used by ChessBase and four different
professional programmers, all of whom want to provide their users the
best playing strength they can, more often than not know what they are
doing, and have no tendency to shoot themselves in the foot.
The message you read is just part of something the we in Computer Chess
like to do even more than write programs or play games: find reasons why
the result of a game/match doesn't count/is not fair/is not
significant/whatever. This one is reason 46 or so, and has just now been
invented. The objective of this game is not to improve your play or your
results but to prove conclusively that any result any program scores
anywhere doesn't count. While this game makes for interesting debate you
should not confuse it with anything actually happening in the real
world.
Amir
Enrique Irazoqui
>One quick note... I just discovered that the "fritz interface" is really
>disruptive to other engines. IE it apparently sends "new" followed by
>*all* the moves in the game, for *each* move played. For crafty this is
>a killer, because I depend on hash stuff carrying over from one search to
>another, which it won't here. And I depend on move ordering stuff staying
>around (killers/history counts/etc) and they get killed by the "new" as
>well.
>
>IE if you are using the engine-vs-engine mode, this looks to be not so
>reliable... I had not noticed this until someone sent me some log
>files that showed it quite clearly. Blitz games are even worse because
>I end up clearing the hash before each move which can take time as well...
I am not using eng-eng. All my matches are run on 2 identical PII-400 with
256 MB RAM. I think that what you describe above may apply to eng-eng on one
single machine. Do you think it also happens when playing on separate
machines?
Enrique
>Bob
mongrel
Amir Ban wrote:
If YOUR program performed worse than expected, would you want us to know why? YES.
ChessBase believes people are interested in engine-engine testing, a part of this
hobby.
ChessBase promotes the Junior 5 interface as a host for automated matches.
I spend real-world money for Junior 5, (partly) for this reason.
I learn in a newsgroup that the results are unreliable.
I warn others.
A chess newsgroup is the appropriate forum.
mongrel
mongrel
Dear Mr. Ban:
I am a customer of yours. Junior 5 is an excellent program. Thank you.
But about this engine-engine feature, programmed by ChessBase...
I invite you to inform us now, that the host program's results are unrealistically
inflated by the way the interface communicates with the engine.
Thank you
mongrel
Henri H. Arsenault
<car...@bellsouth.net> wrote:
> If YOUR program performed worse than expected, would you want us to know
why? YES.
>
> ChessBase believes people are interested in engine-engine testing, a
part of this
> hobby.
> ChessBase promotes the Junior 5 interface as a host for automated matches.
> I spend real-world money for Junior 5, (partly) for this reason.
> I learn in a newsgroup that the results are unreliable.
> I warn others.
> A chess newsgroup is the appropriate forum.
But isn't this to be expected? If you have more than one program running
in the same memory space, dhouldn't you expect the program to hand over
the memory to the other program when it passes the move? How could this be
avoided without rewriting the code for the programs? I supose that if a
computer has enough memory, it could divide up the memory space between
computers, but I don't see how this could be done without reprogramming
both engines. And of course that would require twice as much memory as for
one engine playing against a human.
Does anyone see any solution?
Henri
Enrique Irazoqui
>disruptive to other engines. IE it apparently sends "new" followed by
>*all* the moves in the game, for *each* move played. For crafty this is
>a killer, because I depend on hash stuff carrying over from one search to
>another, which it won't here. And I depend on move ordering stuff staying
>around (killers/history counts/etc) and they get killed by the "new" as
>well.
>
>IE if you are using the engine-vs-engine mode, this looks to be not so
>reliable... I had not noticed this until someone sent me some log
>files that showed it quite clearly. Blitz games are even worse because
>I end up clearing the hash before each move which can take time as well...
>
>Bob
One more thing: what you describe above applies also to Hiarcs 6 when
running as an engine in Fritz, but only in engine-engine games on the same
machine and never when the game is played on 2 machines, as all matches in
my tournament are. I don't see how Fritz's interface can affect Crafty in
the matches I play.
Enrique
Enrique Irazoqui
>
>This is how all versions and brands of Fritz5 and Junior5 work and have
>always worked with all engines. The engine is given the entire game
>history from scratch before every move, so it doesn't need to remember
>anything. Same behavior for all interfaces.
>
>This doesn't affect analysis at all, or practically anything else.
>Remember the interface is used by ChessBase and four different
>professional programmers, all of whom want to provide their users the
>best playing strength they can, more often than not know what they are
>doing, and have no tendency to shoot themselves in the foot.
>
>The message you read is just part of something the we in Computer Chess
>like to do even more than write programs or play games: find reasons why
>the result of a game/match doesn't count/is not fair/is not
>significant/whatever. This one is reason 46 or so, and has just now been
>invented. The objective of this game is not to improve your play or your
>results but to prove conclusively that any result any program scores
>anywhere doesn't count. While this game makes for interesting debate you
>should not confuse it with anything actually happening in the real
>world.
Bob would have had a point if hashtables were reset after the new move
command as it happens in engine-engine form when running on the same
machine. Only that as I said before, hashtables are not reset after every
new move when autoplaying with 2 machines. I just got confirmation from
ChessBase: "The native Crafty is working as designed, no hashtable reset
between setting up new positions and sending moves." Which means that Crafty
is not being disrupted by CB's interface.
I want to congratulate you for Junior 5. It is not only very strong, but it
plays a really fine game of chess. It's as if it would have got rid of the
astigmatism that affects other fast searchers, playing with a more focused,
sharper understanding of the game. And Junior is also very, very quick.
Enrique
>Amir
Enrique Irazoqui
>
>If YOUR program performed worse than expected, would you want us to know
why? YES.
This was not the case with Crafty. See my reply to Amir.
>ChessBase believes people are interested in engine-engine testing, a part
of this
>hobby.
>ChessBase promotes the Junior 5 interface as a host for automated matches.
>I spend real-world money for Junior 5, (partly) for this reason.
>I learn in a newsgroup that the results are unreliable.
Engine-engine matches running on the same machine are always unreliable to
some extent: no permanent brain, resetting of hashtables... This has been
said before a number of times. I still think that this kind of matches are
fun and informative about playing style. But for accurate results it is much
better to run the games on different machines, as some of us always do.
Enrique
Enrique Irazoqui
>why? YES.
>>
>> ChessBase believes people are interested in engine-engine testing, a
>part of this
>> hobby.
>> ChessBase promotes the Junior 5 interface as a host for automated
matches.
>> I spend real-world money for Junior 5, (partly) for this reason.
>> I learn in a newsgroup that the results are unreliable.
>> I warn others.
>> A chess newsgroup is the appropriate forum.
>
>in the same memory space, dhouldn't you expect the program to hand over
>the memory to the other program when it passes the move? How could this be
>avoided without rewriting the code for the programs? I supose that if a
>computer has enough memory, it could divide up the memory space between
>computers, but I don't see how this could be done without reprogramming
>both engines. And of course that would require twice as much memory as for
>one engine playing against a human.
>
>Does anyone see any solution?
Parallel processors. But who has them?
Enrique
>Henri
Andreas Schwartmann
>
> One quick note... I just discovered that the "fritz interface" is really
> disruptive to other engines. IE it apparently sends "new" followed by
> *all* the moves in the game, for *each* move played. For crafty this is
> a killer, because I depend on hash stuff carrying over from one search to
> another, which it won't here.
If it's an engine/engine tournament on a single processor machine, there will not
be any hash carrying over anyway because there is no pondering. Right or wrong?
Andreas
--
schwa...@netcologne.de
ICQ # 24939276
Visit my homepage: http://www.netcologne.de/~nc-schwaran2/
View my latest Fritz 5.32 engine tournament results at
http://www.netcologne.de/~nc-
schwaran2/chess.htm
"The man in black fled across the desert, and the gunslinger followed." - Stephen
King, The Gunslinger
mongrel
> In article <3677A320...@bellsouth.net>, mongrel
> <car...@bellsouth.net> wrote:
>
> > If YOUR program performed worse than expected, would you want us to know
> why? YES.
> >
> > ChessBase believes people are interested in engine-engine testing, a
> part of this
> > hobby.
> > ChessBase promotes the Junior 5 interface as a host for automated matches.
> > I spend real-world money for Junior 5, (partly) for this reason.
> > I learn in a newsgroup that the results are unreliable.
> > I warn others.
> > A chess newsgroup is the appropriate forum.
>
> But isn't this to be expected? If you have more than one program running
> in the same memory space,
They are running in different blocks of the whole memory.
> dhouldn't you expect the program to hand over
> the memory to the other program when it passes the move?
Each program retains possession of its own block of memory.
> How could this be
> avoided without rewriting the code for the programs? I supose that if a
> computer has enough memory, it could divide up the memory space between
> computers, but I don't see how this could be done without reprogramming
> both engines.
The operating system divides the memory between the programs. The program
requests a block of memory when it starts, then it owns this memory while it is
running. One engine puts its hash memory in one place, and the other uses
another place.
> And of course that would require twice as much memory as for
> one engine playing against a human.
Yes. Each program still desires a large block of memory for itself.
> Does anyone see any solution?
>
> Henri
But this is why I replied to Mr. Ban:
Some people posted about a problem with the way Fritz 5.32 communicates with
client engines.
Mr. Ban's post was contemptuous toward those whose interest is finding such
bugs. (This is a hobby in itself.) He does not care whether or not the owners of
the hiarcs.eng know WHY it does not do as well as expected (in engine-engine
matches on one computer).
I do care.
I replied with the rhetorical question: "If YOUR program were being affected,
would you want people to know why?"
So, Henri, if YOUR program were being affected, would you want people to know
why?
My post was not EVEN about the technical issues involved.
Do YOU think that a COMPUTER chess newsgroup is an appropriate forum for bug
reports?
mongrel
Under a multi-threading operating system, with more than one process competing for
I/O,
sometimes the order of events evident from the output can be misleading.
Oh, yes. You are right about that. Thank you. Let everyone know that my argument was
based on a false assumption.
mongrel
mongrel
For example, suppose a small factor caused Junior to be in fourth place in a
very close engine-engine tournament, instead of first. Would you want people
to know why?
In an engine-engine match on one computer, the clearing of hash tables of ONE
engine can be decisive.
In order to have more realistic engine-engine matches under Junior 5, did
the ChessBase programmers command Junior 5 to clear ITS hash tables also?
Finally, what if *I* ran a computer-computer tournament in which I caused Junior 5
to clear its hash tables between moves, but not the other engines, and because of
the loss of just two or three points, Junior finishes fourth instead of first.
Should I tell anyone what I did to Junior?
What is happening in the real world is that people pay money for these programs,
and have a right to know everything they can. And this, a COMPUTER chess
newsgroup, is an appropriate forum for bug reports.
mongrel
Did ChessBase tell the programmers how their engines would be treated?
An expert, Dr. Hyatt, only JUST NOW discovered this!
So ChessBase did not tell some of the programmers.
And they did not tell their potential customers.
Is this why there will not be a hiarcs 7 engine, I wonder?
>
>
> The message you read is just part of something the we in Computer Chess
> like to do even more than write programs or play games: find reasons why
> the result of a game/match doesn't count/is not fair/is not
> significant/whatever. This one is reason 46 or so, and has just now been
> invented. The objective of this game is not to improve your play or your
> results but to prove conclusively that any result any program scores
> anywhere doesn't count. While this game makes for interesting debate you
> should not confuse it with anything actually happening in the real
> world.
>
> Amir
In the real world ChessBase is advertising its "Engine Research Operating System"
as if it is a valid tool.
Will you post here, that the host engine has an advantage over all other engines
under
this system?
Don't you think the buyers should know?
mongrel
Komputer Korner
ordering info between each move which would affect strength?
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Amir Ban wrote in message <367788...@m-sys.com>...
Komputer Korner
The concerns that Bob raised about the engine vs engine matches are
still valid.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Enrique Irazoqui wrote in message
<758fs4$hcm$1...@diana.bcn.ibernet.es>...
Komputer Korner
will see that both engines are grabbing cpu slices at the same time
(theoretically not the same time with only 1 processor but you know
what I mean).
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Andreas Schwartmann wrote in message
<3677F404...@netcologne.de>...
>Robert Hyatt wrote:
>
>>
>> One quick note... I just discovered that the "fritz interface" is
really
>> disruptive to other engines. IE it apparently sends "new" followed
by
>> *all* the moves in the game, for *each* move played. For crafty
this is
>> a killer, because I depend on hash stuff carrying over from one
search to
>> another, which it won't here.
>
mongrel
> like to do even more than write programs or play games: find reasons why
> the result of a game/match doesn't count/is not fair/is not
> significant/whatever. This one is reason 46 or so, and has just now been
> invented. The objective of this game is not to improve your play or your
> results but to prove conclusively that any result any program scores
> anywhere doesn't count. While this game makes for interesting debate you
> should not confuse it with anything actually happening in the real
> world.
If I had known that H6 was not running right under Junior, I would have
bought the full version of H6.
This was a real world decision based on incomplete information.
Now, those contemplating the purchase of Junior 5 or Fritz 5.32 should
be aware of these facts. What is wrong with that?
Does anybody else understand this:
If Mr. Ban's program were being affected, he would probably post that.
But because someone ELSE's program is affected, he ridicules those who
report these things.
If the programmer of Hiarcs posted the above paragraph, one might
consider it a generous attitude.
But since Mr. Ban, a beneficiary, posted it, it is asinine.
My reply to Mr. Ban is not about the technical details, or the money.
Simply, I am enraged by his attitude.
I'll get over it.
But I won't be buying an engine for a ChessBase interface. And as a simple
courtesy to others, I post my opinion here.
What is wrong with that?
Jeremiah Penery
>
> Robert Hyatt escribió en mensaje <756o2r$oho$1...@juniper.cis.uab.edu>...
> >disruptive to other engines. IE it apparently sends "new" followed by
> >*all* the moves in the game, for *each* move played. For crafty this is
> >a killer, because I depend on hash stuff carrying over from one search to
> >around (killers/history counts/etc) and they get killed by the "new" as
> >well.
> >
> >IE if you are using the engine-vs-engine mode, this looks to be not so
> >reliable... I had not noticed this until someone sent me some log
> >files that showed it quite clearly. Blitz games are even worse because
> >I end up clearing the hash before each move which can take time as well...
> >
> >Bob
>
> running as an engine in Fritz, but only in engine-engine games on the same
> machine and never when the game is played on 2 machines, as all matches in
> my tournament are. I don't see how Fritz's interface can affect Crafty in
> the matches I play.
>
> Enrique
The issue is with the Fritz interface and *Winboard* programs *only*.
Whether there is one engine running or two, the Fritz-Winboard interface
does not do things the same way as the true Winboard interface does.
I.E. Winboard only sends a 'new' command when you tell Winboard to
restart the game. The reason is that the 'new' command gets rid of all
the hash tables, move ordering, etc. info that is important to the
running engine. The Fritz interface does not use 'standard' Winboard
command protocol, and so engines are broken. This is the case in
eng-eng matches on one machine, or when running a Winboard engine in the
Fritz interface on its own machine.
One solution, although it'd be difficult to implement, is that when the
command 'Fritz' is received (this is sent by Fritz at the beginning, is
it not?), the engine could parse commands differently, so that a 'new'
under Fritz won't act like a true 'new'. *shrug* Either that or
Chessbase makes Fritz-Winboard behave properly.
Jeremiah
Robert Hyatt
: This is how all versions and brands of Fritz5 and Junior5 work and have
: always worked with all engines. The engine is given the entire game
: history from scratch before every move, so it doesn't need to remember
: anything. Same behavior for all interfaces.
Ugh... I thought this was only the "winboard interface" stuff, *not* the
real fritz interface. If the *real* interface behaves like this it is
a really poor design. For reasons I have already given. IE this is sort
of like letting a human play one move, then blanking his mind, then
giving him the next position to search. He has no continuity between
moves, no remembered key positions, etc...
: This doesn't affect analysis at all, or practically anything else.
It certainly effects engine strength. In a blitz match, clearing the
hash table takes time if it is of decent size. But losing the info
from one search to another is easily a factor of 2x in places. IE
If I complete a 12 ply search, and make a move, I would normally *start*
searching at 11 plies, since after making the 12 ply search move, and
then the opponent made the expected move, I still have a perfectly valid
10 ply variation and score and don't need to start all over again.
That is a *huge* penalty on each move that is correctly predicted, which
is at *least* 50% of the total moves played and usually (for me) much
higher.
: Remember the interface is used by ChessBase and four different
: professional programmers, all of whom want to provide their users the
: best playing strength they can, more often than not know what they are
: doing, and have no tendency to shoot themselves in the foot.
Perhaps they "shoot others in the foot instead"???
: The message you read is just part of something the we in Computer Chess
: like to do even more than write programs or play games: find reasons why
: the result of a game/match doesn't count/is not fair/is not
: significant/whatever. This one is reason 46 or so, and has just now been
: invented. The objective of this game is not to improve your play or your
: results but to prove conclusively that any result any program scores
: anywhere doesn't count. While this game makes for interesting debate you
: should not confuse it with anything actually happening in the real
: world.
that's baloney. Heiko pointed out a problem I had not noticed, because I
don't use the chessbase stuff at all. Someone else sent me a couple of
log files from games played by crafty doing this. Whether you want to
believe it or not, it is a huge penalty to my program. I didn't say that
such matches shouldn't be played, but I did say that if someone plays a
20 game match like this, then plays a 20 game match on ICC, they are going
to get *far* different results...
: Amir
Robert Hyatt
: In article <3677A320...@bellsouth.net>, mongrel
: <car...@bellsouth.net> wrote:
:> If YOUR program performed worse than expected, would you want us to know
: why? YES.
:>
:> ChessBase believes people are interested in engine-engine testing, a
: part of this
:> hobby.
:> ChessBase promotes the Junior 5 interface as a host for automated matches.
:> I spend real-world money for Junior 5, (partly) for this reason.
:> I learn in a newsgroup that the results are unreliable.
:> I warn others.
:> A chess newsgroup is the appropriate forum.
: But isn't this to be expected? If you have more than one program running
: in the same memory space, dhouldn't you expect the program to hand over
: the memory to the other program when it passes the move? How could this be
: avoided without rewriting the code for the programs? I supose that if a
: computer has enough memory, it could divide up the memory space between
: computers, but I don't see how this could be done without reprogramming
: both engines. And of course that would require twice as much memory as for
: one engine playing against a human.
The above isn't a problem. IE if you turn "pondering off" then you don't get
any interference between the engines when one is running and one is waiting
for the move. And memory shouldn't be "shared" by swapping one out as that
will really add overhead to *both* engines and effectively slow the speed
of the machine by a significant margin. But you *should* be able to expect
that the engine itself isn't going to be completely reset after each move
so that nothing can be carried across searches, for those of us that spent
a lot of time making this effective.
: Does anyone see any solution?
Yes... it has to do with a grenade. :) A large magnet. And a tender
body part. :)
Robert Hyatt
: Henri H. Arsenault escribió en mensaje ...
:>In article <3677A320...@bellsouth.net>, mongrel
:><car...@bellsouth.net> wrote:
:>
:>> If YOUR program performed worse than expected, would you want us to know
:>why? YES.
:>>
:>> ChessBase believes people are interested in engine-engine testing, a
:>part of this
:>> hobby.
:>> ChessBase promotes the Junior 5 interface as a host for automated
: matches.
:>> I spend real-world money for Junior 5, (partly) for this reason.
:>> I learn in a newsgroup that the results are unreliable.
:>> I warn others.
:>> A chess newsgroup is the appropriate forum.
:>But isn't this to be expected? If you have more than one program running
:>in the same memory space, dhouldn't you expect the program to hand over
:>the memory to the other program when it passes the move? How could this be
:>avoided without rewriting the code for the programs? I supose that if a
:>computers, but I don't see how this could be done without reprogramming
:>both engines. And of course that would require twice as much memory as for
:>one engine playing against a human.
:>Does anyone see any solution?
: Parallel processors. But who has them?
: Enrique
*many* do. I have a student that I just discovered has a dual PII/400 in
his dorm room.
But that isn't the issue... that would cure the ponder=off stuff, but
it wouldn't help at all as you can't ponder if the interface says "new"
before each move, that resets and tosses out *everything*...
:>Henri
Robert Hyatt
: Robert Hyatt wrote:
:>
:> One quick note... I just discovered that the "fritz interface" is really
:> disruptive to other engines. IE it apparently sends "new" followed by
:> *all* the moves in the game, for *each* move played. For crafty this is
:> a killer, because I depend on hash stuff carrying over from one search to
:> another, which it won't here.
: If it's an engine/engine tournament on a single processor machine, there will not
: be any hash carrying over anyway because there is no pondering. Right or wrong?
: Andreas
Wrong. When I complete a search (say 12 plies) and make that move, I then sit
and wait. If you make the right move, I remove the first 2 moves from the
12 ply PV, trim it to say "10 plies" (because it is still a valid 10 ply
search result with the first 2 plies removed) and then start searching at
11 plies, avoiding the wasted time of plies 1-10 again. If you type "new"
and then force the moves into crafty, you wipe this out, so that when I
start searching I don't have the hash from the last search to help with
move ordering and to provide scoring information, and I don't get to start
the search at ply=11, I have to start over at 1...
*big* penalty...
: --
: schwa...@netcologne.de
: ICQ # 24939276
: Visit my homepage: http://www.netcologne.de/~nc-schwaran2/
: View my latest Fritz 5.32 engine tournament results at
: http://www.netcologne.de/~nc-
: schwaran2/chess.htm
: "The man in black fled across the desert, and the gunslinger followed." - Stephen
: King, The Gunslinger
--
Robert Hyatt
: Robert Hyatt escribió en mensaje <756o2r$oho$1...@juniper.cis.uab.edu>...
:>disruptive to other engines. IE it apparently sends "new" followed by
:>*all* the moves in the game, for *each* move played. For crafty this is
:>a killer, because I depend on hash stuff carrying over from one search to
:>around (killers/history counts/etc) and they get killed by the "new" as
:>well.
:>
:>IE if you are using the engine-vs-engine mode, this looks to be not so
:>reliable... I had not noticed this until someone sent me some log
:>files that showed it quite clearly. Blitz games are even worse because
:>I end up clearing the hash before each move which can take time as well...
: I am not using eng-eng. All my matches are run on 2 identical PII-400 with
: 256 MB RAM. I think that what you describe above may apply to eng-eng on one
: single machine. Do you think it also happens when playing on separate
: machines?
I wouldn't think so, so long as I don't get "new" commands before each move.
The "new" is the killer issue here. I was sent several log files that for
each move had "new", "force", then a long series of moves, then a "level xxx"
and then a "go". No continuity across moves at all.
: Enrique
:>Bob
Robert Hyatt
: Robert Hyatt escribió en mensaje <756o2r$oho$1...@juniper.cis.uab.edu>...
:>One quick note... I just discovered that the "fritz interface" is really
:>disruptive to other engines. IE it apparently sends "new" followed by
:>*all* the moves in the game, for *each* move played. For crafty this is
:>a killer, because I depend on hash stuff carrying over from one search to
:>another, which it won't here. And I depend on move ordering stuff staying
:>around (killers/history counts/etc) and they get killed by the "new" as
:>well.
:>
:>IE if you are using the engine-vs-engine mode, this looks to be not so
:>reliable... I had not noticed this until someone sent me some log
:>files that showed it quite clearly. Blitz games are even worse because
:>I end up clearing the hash before each move which can take time as well...
:>
: One more thing: what you describe above applies also to Hiarcs 6 when
: running as an engine in Fritz, but only in engine-engine games on the same
: machine and never when the game is played on 2 machines, as all matches in
: my tournament are. I don't see how Fritz's interface can affect Crafty in
: the matches I play.
: Enrique
I agree that your setup seems perfectly reasonable. One good sanity check
is to check the crafty directory after one game. If there are many log.nnn
files, there is big trouble, because "new" closes the old file and opens a
new one.
Enrique Irazoqui
>
>The issue is with the Fritz interface and *Winboard* programs *only*.
True. In the matches I play I don't use Winboard. Crafty 16.1 runs as
the chess engine of Nimzo 99 and it is not disturbed by resetting hashtables
with the new move command.
Enrique
Enrique Irazoqui
>I agree that your setup seems perfectly reasonable. One good sanity check
>is to check the crafty directory after one game. If there are many log.nnn
>files, there is big trouble, because "new" closes the old file and opens a
>new one.
No log.nnn files at all.
Enrique
Enrique Irazoqui
>Okay comp vs comp matches are okay, but not engine vs engine matches.
>The concerns that Bob raised about the engine vs engine matches are
>still valid.
Not for the original issue, which was Crafty's performance in my tournament.
Enrique
Robert Hyatt
: Komputer Korner escribió en mensaje ...
:>Okay comp vs comp matches are okay, but not engine vs engine matches.
:>The concerns that Bob raised about the engine vs engine matches are
:>still valid.
: Not for the original issue, which was Crafty's performance in my tournament.
: Enrique
Sorry... didn't mean to make your tournament an issue. I had just found out
about this "glitch" and wanted to make it known...
Bob
--
Komputer Korner
engine vs engine matches within the Fritz/Junior/Nimzo GUI are sending
signals to guest engines to clear all info between each move so that
both environments are acting this way 1)Winboard adaptor environment
and the regular Fritz/Junior/Nimzo engine- engine environment?
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Robert Hyatt wrote in message <7590ar$aq9$2...@juniper.cis.uab.edu>...
>Amir Ban <ami...@m-sys.com> wrote:
>
>: This is how all versions and brands of Fritz5 and Junior5 work and
have
>: always worked with all engines. The engine is given the entire game
>: history from scratch before every move, so it doesn't need to
remember
>: anything. Same behavior for all interfaces.
>
>
>Ugh... I thought this was only the "winboard interface" stuff, *not*
the
>real fritz interface. If the *real* interface behaves like this it
is
>a really poor design. For reasons I have already given. IE this is
sort
>of like letting a human play one move, then blanking his mind, then
>giving him the next position to search. He has no continuity between
>moves, no remembered key positions, etc...
>
>: This doesn't affect analysis at all, or practically anything else.
>
>It certainly effects engine strength.
snipped
>That is a *huge* penalty on each move that is correctly predicted,
which
>is at *least* 50% of the total moves played and usually (for me) much
>higher.
>: Amir
Komputer Korner
same as the Fritz/Winboard matches. Bob, will the Crafty engine that
is supplied for the 32 bit Junior environment make a log file? If it
does I will run an engine vs engine match within Junior 5 and check to
see if more than 1 log file is created.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Jeremiah Penery wrote in message <36781CE5...@lords.com>...
>
>The issue is with the Fritz interface and *Winboard* programs *only*.
Komputer Korner
and Junior 5. Crafty got killed and I stopped the slaughter after 3
games. The problem is BOB, there were no log files created by Crafty
nor Junior.
You should make it mandatory that (if anybody modifies Crafty to
produce an engine for engine vs engine tests) Crafty always produce a
log file. So we can't tell if the Winboard adaptor problem is being
repeated for Junior environment engine vs engine matches or not.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Komputer Korner wrote in message ...
Amir Ban
> Dear Mr. Ban:
>
> I am a customer of yours. Junior 5 is an excellent program. Thank you.
> But about this engine-engine feature, programmed by ChessBase...
> I invite you to inform us now, that the host program's results are unrealistically
> inflated by the way the interface communicates with the engine.
>
> Thank you
> mongrel
Do I get a choice in what to inform you ?
I don't remember seeing anything in the discussion that says any engine
is treated differently than any other. The only post I made about it
said the opposite quite plainly, and I didn't read anything posted by
anyone else that would make you think otherwise.
Seems to me you have no reason to suspect what you say above, except
your default assumption that I and ChessBase are dishonest. I'm sorry,
you don't suspect it, you know it, and I can only confess now.
The facts here are that J5 clears its hash tables before every move, but
not because it is told by the interface (there is no such command). If
you believe everything that you readg here, this makes J5 much weaker
than it can be. This is the opposite of what you are claiming, but I'm
sure it's possible to build a conspriacy theory out of this too.
Amir
Robert Hyatt
: mongrel wrote:
: Amir
It does make fritz weaker. But that's a decision Frans made when he chose
to be a "root processor". no problem. It is *not* a decision I chose to make,
however. I spent a *lot* of time (a) making it possible to carry the hash
table from move to move without having to clear it and lost valuable infor-
mation; (b) carrying the PV from search to search so that I don't have to
duplicate work from the last search when my opponent plays the predicted
move; (c) carry move ordering information from search to search thru the
killer moves and history counts; (d) carry timing information from move to
move (no, stuffing a "level" command into an engine at the front of each move
is *not* the correct way to set the time limits, the xboard protocol has a
simple time/otim command pair that works without destroying existing time-
control information).
Losing all of the above is *bad* "for my program." YMMV. And I don't think
there's any dishonesty involved. But perhaps a bit of *very* sloppy programming
is certainly involved. And, like it or not, it *does* affect each engine in
a probably different way. And it does affect the outcomes of at least a few
games/matches. Whether it makes any overall difference is unknown. But I do
know that it will definitely hurt Crafty.
--
Robert Hyatt Computer and Information Sciences
Robert Hyatt
: However, Amir seemed to suggest that all Fritz GUI matches act the
: same as the Fritz/Winboard matches. Bob, will the Crafty engine that
: is supplied for the 32 bit Junior environment make a log file? If it
: does I will run an engine vs engine match within Junior 5 and check to
: see if more than 1 log file is created.
I believe that Enrique (or someone) said that crafty doesn't make log files
when used as a fritz engine, for reasons unknown to me... I suggested this
"check" since a "new" rolls over onto a new log.nnn file.
Robert Hyatt
: I just ran a 2 minute blitz match on a 166 pentium between Crafy 15.8
: and Junior 5. Crafty got killed and I stopped the slaughter after 3
: games. The problem is BOB, there were no log files created by Crafty
: nor Junior.
: You should make it mandatory that (if anybody modifies Crafty to
: produce an engine for engine vs engine tests) Crafty always produce a
: log file. So we can't tell if the Winboard adaptor problem is being
: repeated for Junior environment engine vs engine matches or not.
That is probably the worst possible test case scenario. *if* "new" is
sent before each move. Because it is going to burn a lot of cpu time to
clear the hash table before each move... And there is no easy way to
actually know if this is happening without the logs.
And no, I'm not particularly happy that the log files have been removed,
because *I* get the questions about why it does this or that, and now my
main debugging tool is missing.
Amir Ban
>
> Amir what about the clearing of hash tables and the loss of move
> ordering info between each move which would affect strength?
>
I didn't say it wouldn't affect anything. I said this is just one more
argument out of many that always seem to be available to discredit any
result. I think it's minor, but even if it's not Fritz and Junior suffer
from it as much as anyone else.
By now this issue has taken an ugly twist, which in hindsight was
predictable. You should realize that I wrote my original post before
that. As things are, count me out of this debate.
Amir
mongrel
MY reply to your post was not even technical.
It was about your ATTITUDE toward your customers.
Robert Hyatt
: Komputer Korner wrote:
:>
:> Amir what about the clearing of hash tables and the loss of move
:> ordering info between each move which would affect strength?
:>
: I didn't say it wouldn't affect anything. I said this is just one more
: argument out of many that always seem to be available to discredit any
: result. I think it's minor, but even if it's not Fritz and Junior suffer
: from it as much as anyone else.
From your earlier statement, this isn't true. IE you said that you/fritz
always clear the hash between searches. I don't. You didn't mention whether
you always start at iteration 1 on the next move either. I don't, but I am
forced to in this scheme. you also didn't mention whether you/fritz keep
things like killer moves, history counts and so forth across searches. I
do, but I don't under this scheme.
It might be that I get hurt more than some, less than others, but that
wasn't the issue, it was that this doesn't need to happen *period* to
any program. The xboard interface doesn't do this. Tim spent the time
to make it work correctly and consistently, without doing things that might
well disrupt an engine's "capabilities/features."
: By now this issue has taken an ugly twist, which in hindsight was
: predictable. You should realize that I wrote my original post before
: that. As things are, count me out of this debate.
: Amir
I hate to see you bail out. I'm not accusing you of anything. I am
interested in how this affects your program, as well as fritz and the
other commercial engines, *plus* how it affects the non-commercial
engines as well. I can only explain what it does to my program. Based
on the logs sent to me by a user that was playing with this. There will
always be those that believe in "conspiracies". I don't here. I only
believe that there is a sloppy bit of programming going on somewhere that
can be repaired now that it has been noticed...
mclane
>By now this issue has taken an ugly twist, which in hindsight was
>predictable. You should realize that I wrote my original post before
>that. As things are, count me out of this debate.
>Amir
the THINGS always get an UGLY TWIST, especially when somebody finds
out something about fritz/chessBase methods.
isn't it amir ? one should forbid that people find out about
chessBase.
it seems most of the computerchess area is in fact a minefield. and
leutenant amir likes to walk somebody else instead :-)))
best wishes
Komputer Korner
with the next version of crafty and all future versions requiring that
log files be kept?
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Robert Hyatt wrote in message <75avi3$pah$3...@juniper.cis.uab.edu>...
>Komputer Korner <kor...@netcom.ca> wrote:
>: I just ran a 2 minute blitz match on a 166 pentium between Crafy
15.8
>: and Junior 5. Crafty got killed and I stopped the slaughter after 3
>: games. The problem is BOB, there were no log files created by
Crafty
>: nor Junior.
>: You should make it mandatory that (if anybody modifies Crafty to
>: produce an engine for engine vs engine tests) Crafty always
produce a
>: log file. So we can't tell if the Winboard adaptor problem is
being
>: repeated for Junior environment engine vs engine matches or not.
>
>That is probably the worst possible test case scenario. *if* "new"
is
>sent before each move. Because it is going to burn a lot of cpu time
to
>clear the hash table before each move... And there is no easy way to
>actually know if this is happening without the logs.
>
>And no, I'm not particularly happy that the log files have been
removed,
>because *I* get the questions about why it does this or that, and now
my
>main debugging tool is missing.
>
>
>
Komputer Korner
terms) that all the PV info, hash etc. is getting cleared between each
move in the guest engines in all engine vs engine matches in the
regular Fritz GUI? It seems to me that if Fritz and Junior hash is
getting cleared in engine vs engine matches then ChessBase wouldn't
allow guest engines to keep their info. So I am guessing the
probability is close to 100% that everything is cleared and this would
help to explain some results in engine vs engine matches within the
Fritz GUI. Am I right Bob?
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Robert Hyatt wrote in message <75bdhb$sgl$1...@juniper.cis.uab.edu>...
>Amir Ban <ami...@m-sys.com> wrote:
>: I didn't say it wouldn't affect anything. I said this is just one
more
>: argument out of many that always seem to be available to discredit
any
>: result. I think it's minor, but even if it's not Fritz and Junior
suffer
>: from it as much as anyone else.
>
>From your earlier statement, this isn't true. IE you said that
you/fritz
>always clear the hash between searches. I don't. You didn't mention
whether
>you always start at iteration 1 on the next move either. I don't,
but I am
>forced to in this scheme. you also didn't mention whether you/fritz
keep
>things like killer moves, history counts and so forth across
searches. I
>do, but I don't under this scheme.
>
snipped
>
>I hate to see you bail out. I'm not accusing you of anything. I am
>interested in how this affects your program, as well as fritz and the
>other commercial engines, *plus* how it affects the non-commercial
>engines as well. I can only explain what it does to my program.
Based
>on the logs sent to me by a user that was playing with this. There
will
>always be those that believe in "conspiracies". I don't here. I
only
>believe that there is a sloppy bit of programming going on somewhere
that
>can be repaired now that it has been noticed...
>
>
RDavis101
altered without his prior permission where the engine is used to enhance the
value of a commercial product.
When I ordered Fritz 5.32 from ICD a couple of weeks ago, one reason I did so
was that it would provide a nice interface for Crafty, because I wanted a
program whose development featured optimization for anti-human play (not SSDF
tweaking)...
Roger
>Subject: Re: Crafty vs Junior in engine vs engine within Junior 5 environment
>From: "Komputer Korner" <kor...@netcom.ca>
>Date: 12/17/98 11:56 PM Central Standard Time
>Message-id: <zfme2.1122$7l6....@tor-nn1.netcom.ca>
>
>Can't you stop this and have the log files put back in or at least
>with the next version of crafty and all future versions requiring that
>log files be kept?
>
>--
>--
>Komputer Korner
>The inkompetent komputer
>
>To send email take the 1 out of my address. My email address is
>kor...@netcom.ca but take the 1 out before sending the email.
Enrique Irazoqui
>Based on what you know now Bob, how likely is it (in probability
>terms) that all the PV info, hash etc. is getting cleared between each
>move in the guest engines in all engine vs engine matches in the
>regular Fritz GUI? It seems to me that if Fritz and Junior hash is
>getting cleared in engine vs engine matches then ChessBase wouldn't
>allow guest engines to keep their info. So I am guessing the
>probability is close to 100% that everything is cleared and this would
>help to explain some results in engine vs engine matches within the
>Fritz GUI. Am I right Bob?
As far as I can tell, in eng-eng games hashtables are reset with new moves.
I know this happens with Hiarcs 6, which is supposed to retain information
in hashtables from one move to the next. And yes, this contributes to
explain why Hiarcs 6 does better in DOS than in eng-eng within Fritz. I
don't know about Crafty in eng-eng.
Enrique
>Komputer Korner
>The inkompetent komputer
>
>To send email take the 1 out of my address. My email address is
>kor...@netcom.ca but take the 1 out before sending the email.
>Robert Hyatt wrote in message <75bdhb$sgl$1...@juniper.cis.uab.edu>...
Robert Hyatt
: Based on what you know now Bob, how likely is it (in probability
: terms) that all the PV info, hash etc. is getting cleared between each
: move in the guest engines in all engine vs engine matches in the
: regular Fritz GUI? It seems to me that if Fritz and Junior hash is
: getting cleared in engine vs engine matches then ChessBase wouldn't
: allow guest engines to keep their info. So I am guessing the
: probability is close to 100% that everything is cleared and this would
: help to explain some results in engine vs engine matches within the
: Fritz GUI. Am I right Bob?
I really don't know. *all* i have to go on is some log files sent
to me by someone chatting about this on ICC. And in the log files,
it is quite clear that there is a big problem. But then Enrique is
running Crafty as a normal fritz engine and ran a couple of tests for
me yesterday and they seemed perfectly normal.
At this point, all I can say is that there is definitely a problem
somewhere, but I am not anywhere near certain where/when this problem
arises...
I'll try to figure this out and at least post a summary...
Robert Hyatt
: Can't you stop this and have the log files put back in or at least
: with the next version of crafty and all future versions requiring that
: log files be kept?
I may simply consider stopping this "engine" stuff totally. IE I get a
lot of questions, and don't have much idea of what was changed. I feel
a little foolish when I can't answer a question about my own program at
times.. :)
Enrique Irazoqui
>I really don't know. *all* i have to go on is some log files sent
>to me by someone chatting about this on ICC. And in the log files,
>it is quite clear that there is a big problem. But then Enrique is
>running Crafty as a normal fritz engine and ran a couple of tests for
>me yesterday and they seemed perfectly normal.
In comp-comp (auto232) they are normal, but in eng-eng on the same machine I
have no idea. I do know, as I posted before, that the engine Hiarcs 6 for
Fritz clears hashtables in eng-eng and Hiarcs doesn't like that, but this is
not neccessarily the same for all engines, I suppose. Is this clearing
imposed by the interface or is it up to the engine's programmer?
In any case, I wouldn't say this is such a big issue, taking into
consideration that engine-engine games on the same machine are a bit
crippled to begin with, even if some engines are more handicapped than
others. Resetting hashtables when using Winboard/Fritz seems more of a
problem to me.
Enrique
>At this point, all I can say is that there is definitely a problem
>somewhere, but I am not anywhere near certain where/when this problem
>arises...
>
>I'll try to figure this out and at least post a summary...
>
>
>
Seifriz
Robert Hyatt schrieb:
> Komputer Korner <kor...@netcom.ca> wrote:
> : Based on what you know now Bob, how likely is it (in probability
> : terms) that all the PV info, hash etc. is getting cleared between each
> : move in the guest engines in all engine vs engine matches in the
> SNIP
> I really don't know. *all* i have to go on is some log files sent
> to me by someone chatting about this on ICC. And in the log files,
>
SNIP
> I'll try to figure this out and at least post a summary...
>
Maybe I am too late in this thread, butI found a letter to the editor in
last CSS where
Dr. Dieter Bauer found out that the HIARCS engine
in Fritz 5 shows hashtables but in fact uses
no hashtables at all (in WIN 95 certainly, maybe 98 as well).
Fritz 5 and Junior show correct hashtable use.
Bauer used a program called SPEICHER.EXE,
which seems to work very well even under Fritz which likes
the whole screen. We do not have it, but it can be
found easily I suppose on the internet.
Bert/gambitsoft.com
Ernst A. Heinz
> Maybe I am too late in this thread, butI found a letter to the editor in
> last CSS where
> Dr. Dieter Bauer found out that the HIARCS engine
> in Fritz 5 shows hashtables but in fact uses
> no hashtables at all (in WIN 95 certainly, maybe 98 as well).
> Fritz 5 and Junior show correct hashtable use.
> Bauer used a program called SPEICHER.EXE,
> which seems to work very well even under Fritz which likes
> the whole screen. We do not have it, but it can be
> found easily I suppose on the internet.
> Bert/gambitsoft.com
According to my inspections with the vanilla Windows/NT "Task Manager",
the "Hiarcs 6" engine in 16bit "Fritz 5" seems to allocate about 16MB
of memory *regardless* of the hash size that you set. Whether these
16MB are hash tables and if they are actually used, I do not know.
=Ernst=
Enrique Irazoqui
>
>Maybe I am too late in this thread, butI found a letter to the editor in
>last CSS where
>Dr. Dieter Bauer found out that the HIARCS engine
>in Fritz 5 shows hashtables but in fact uses
>no hashtables at all (in WIN 95 certainly, maybe 98 as well).
>Fritz 5 and Junior show correct hashtable use.
>Bauer used a program called SPEICHER.EXE,
>which seems to work very well even under Fritz which likes
>the whole screen. We do not have it, but it can be
>found easily I suppose on the internet.
>Bert/gambitsoft.com
According to Mark Uniacke, the programmer of Hiarcs, the engine Hiarcs 6 for
Fritz uses up to 8MB RAM for hashtables.
Enrique
Komputer Korner
future but if the matches aren't fair then all bets are off.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Enrique Irazoqui wrote in message
<75dm7j$k6j$1...@diana.bcn.ibernet.es>...
>In any case, I wouldn't say this is such a big issue, taking into
>consideration that engine-engine games on the same machine are a bit
>crippled to begin with, even if some engines are more handicapped
than
>others. Resetting hashtables when using Winboard/Fritz seems more of
a
>problem to me.
>
>Enrique
>
>>At this point, all I can say is that there is definitely a problem
>>somewhere, but I am not anywhere near certain where/when this
problem
>>arises...
>>
>>I'll try to figure this out and at least post a summary...
>>
>>
>>
Komputer Korner
will indeed play weaker in the Fritz GUI.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Enrique Irazoqui wrote in message
<75dv36$qpg$1...@diana.bcn.ibernet.es>...
Enrique Irazoqui
>I don't agree. Engine vs engine in the Fritz GUI is the wave of the
>future but if the matches aren't fair then all bets are off.
Hmmm. No salvation for Rebel, Mchess, the King, Genius... ? :)
Anyway, I was talking about the present.
Enrique
>Komputer Korner
mclane
>In comp-comp (auto232) they are normal, but in eng-eng on the same machine I
>have no idea. I do know, as I posted before, that the engine Hiarcs 6 for
>Fritz clears hashtables in eng-eng and Hiarcs doesn't like that, but this is
>not neccessarily the same for all engines, I suppose. Is this clearing
>imposed by the interface or is it up to the engine's programmer?
>In any case, I wouldn't say this is such a big issue, taking into
>consideration that engine-engine games on the same machine are a bit
>crippled to begin with, even if some engines are more handicapped than
>others. Resetting hashtables when using Winboard/Fritz seems more of a
>problem to me.
I have always said that my hiarcs plays much stronger than the hiarcs6
engine for fritz when in eng-eng autoplay.
this was a feeling.
i had no evidence.
again feelings seem to me much more precise than bean-counting !
>Enrique
best wishes
Seifriz
Enrique Irazoqui schrieb:
> Seifriz escribió en mensaje <367A7D4F...@gambitsoft.com>...
> >
> >Maybe I am too late in this thread, butI found a letter to the editor in
> >last CSS where
> >Dr. Dieter Bauer found out that the HIARCS engine
> >in Fritz 5 shows hashtables but in fact uses
> >no hashtables at all (in WIN 95 certainly, maybe 98 as well).
> >Fritz 5 and Junior show correct hashtable use.
> >Bauer used a program called SPEICHER.EXE,
> >which seems to work very well even under Fritz which likes
> >the whole screen. We do not have it, but it can be
> >found easily I suppose on the internet.
> >Bert/gambitsoft.com
>
> According to Mark Uniacke, the programmer of Hiarcs, the engine Hiarcs 6 for
> Fritz uses up to 8MB RAM for hashtables.
>
> Enrique
Well, Dr Bauer's letter said that you look at the hashtable usage
in Fritz and it says, so many MBs are used for Hiarcs' hashtables, but in
fact nothing happens in the Random Access Memory or on your
harddisk.
The user is mislead. HIARCS works without one single MB
hash, no matter what the FRITZ GUI shows to you.
This is the problem, Dr.Bauer described! So the Hiarcs engine
seems to have a disadvantage!
Bert
Enrique Irazoqui
The Hiarcs 6 engine for Fritz has a disadvantage compared to the DOS version
and to other engines for Fritz, I agree. But I find hard to believe that
Mark is wrong about his own engine, and he claims it uses 8MB hashtables.
Mark, can you tell us?
Enrique
Rolf Tueschen
<758fs4$hcm$1...@diana.bcn.ibernet.es>:
>I want to congratulate you for Junior 5. It is not only very strong, but it
>plays a really fine game of chess. It's as if it would have got rid of the
>astigmatism that affects other fast searchers, playing with a more focused,
>sharper understanding of the game. And Junior is also very, very quick.
Perhaps rgcc leaves enough room for us to ask trivial questions ...
What always interested me is this, how could someone who's only an
operator for chessprograms or who was in the ooperating scene for - say
- 30 years, understand the chessic side of the whole field if he himself
is just a low levelled chessplayer. Look, I'm not talking about JUNIOR,
I'm not talking about the details of engine vs engine "matches", this
time I don't even want to talk about the experimental design of such a
thing, --- -- -- all I want is to understand better how a lower chessic
level of understanding could be ameliorated by experience with
chessprograms or engines. More so, if I read the declarations of a man
who couldn't send me a *single* (in other words "not one game") game as
an example of his own chess record. Perhaps you remember our friendly
email exchange in '96.
Let me give a little analogy of motorsports. For me all this looks as if
the mechanicians in the technical team of a pilot would start to talk
about the details of the race itself. For me technique is the base of
the car itself, also depending of the specific profile of the local
details of the site, but only the pilot himself and perhaps his closest
manager will understand what a special course required.
Tennis is another example. Somewhere in Dejas is an older post of mine
containing my belief that I'm a real "TV" expert of the game. I know a
lot about my favorite players. All about their habits. Their tricks. But
would you believe me if I told you that I really *knew* what's *really*
going on down there on center-court?
I understand that you're smart enough to make the difference between two
completely seperated spheres of reality. Virtual the one (as TV watcher
or operator in computerchess) and real the other in RL. You follow me?
So, how could you talk about nice games and such things? Perhaps through
'morphic fields'? :)
>Enrique
mclane
>Perhaps rgcc leaves enough room for us to ask trivial questions ...
>What always interested me is this, how could someone who's only an
>operator for chessprograms or who was in the ooperating scene for - say
>- 30 years, understand the chessic side of the whole field if he himself
>is just a low levelled chessplayer.
how can harry valerien moderate a sport magazine, although he is not a
sport-guy ? how can marcel reich ranicki report about books, although
he is not a writer himself ? how can karl may write books about
winnetou and old shatterhand, and was never in us ? how can sombody be
against the german army, without ever having been a soldier ? how can
somebody taste different kind of wine, without beeing a wine-farmer
himself? how can somebody differenciate restaurants in quality,
without beeing a cook himself ? how can somebody be actor, without
ever having learnt acting ?
and - at least - how can an asshole like you understand how somebody
learns about a field, without having the capacity to find out yourself
?
> Look, I'm not talking about JUNIOR,
no - you talk about something else. you want to attack enrique. and
this is shit. fuck off rolf.
>I'm not talking about the details of engine vs engine "matches", this
>time I don't even want to talk about the experimental design of such a
>thing, --- -- -- all I want is to understand better how a lower chessic
>level of understanding could be ameliorated by experience with
>chessprograms or engines. More so, if I read the declarations of a man
>who couldn't send me a *single* (in other words "not one game") game as
>an example of his own chess record. Perhaps you remember our friendly
>email exchange in '96.
piss off rolf. we don't want your boring spamming here.
beat your wife. or your children. but let us alone do our hobby.
>Let me give a little analogy of motorsports.
another field you have plenty of knowledge ? :-))))
> For me all this looks as if
>the mechanicians in the technical team of a pilot would start to talk
>about the details of the race itself.
in fact, this is what happens. because the technicians are a major
part of the team. and schumacher is only the driver. he could not
repair his car.
an operator is - in your image - somebody who watches all kind of
races for years, and gets experienced. like a soccer-reporter knows
about the teams and seasons without playing soccer himself.
whats wrong with this ?
>For me technique is the base of
>the car itself, also depending of the specific profile of the local
>details of the site, but only the pilot himself and perhaps his closest
>manager will understand what a special course required.
of course the manager knows the most :-)))))
you seem to have much insider knowledge about managers :-)))
>Tennis is another example. Somewhere in Dejas is an older post of mine
>containing my belief that I'm a real "TV" expert of the game. I know a
>lot about my favorite players. All about their habits. Their tricks. But
>would you believe me if I told you that I really *knew* what's *really*
>going on down there on center-court?
nobody beliefs that YOU know something. thats right.
>I understand that you're smart enough to make the difference between two
>completely seperated spheres of reality. Virtual the one (as TV watcher
>or operator in computerchess) and real the other in RL. You follow me?
no-
>So, how could you talk about nice games and such things? Perhaps through
>'morphic fields'? :)
maybe he knows more about computerchess than YOU do.
you know nothing. all you can do is produce a lot of insults against
all kinds of people. this is not difficult. an ape can do it much
faster than you, and even more elegant.
best wishes
mclane
> Well, Dr Bauer's letter said that you look at the hashtable usage
>in Fritz and it says, so many MBs are used for Hiarcs' hashtables, but in
>fact nothing happens in the Random Access Memory or on your
>harddisk.
>The user is mislead. HIARCS works without one single MB
>hash, no matter what the FRITZ GUI shows to you.
>This is the problem, Dr.Bauer described! So the Hiarcs engine
>seems to have a disadvantage!
>Bert
But if this is true, and if this has been all the time, than i
understand why my dos-hiarcs in x-mode produced much better results
than those stupid chessbase engine vs. engines matches that
computerschach and spiele posted years over years to push the selling
of fritz against hiarcs. this would mean a great betrayal has
happened. again - nice style !!!
and of course, illegal in germany. if the program says: it is using
xyz hash, and it IS NOT using xyz hash, than this is illegal to claim.
best wishes
Rolf Tueschen
>This is how all versions and brands of Fritz5 and Junior5 work and have
>always worked with all engines. The engine is given the entire game
>history from scratch before every move, so it doesn't need to remember
>anything. Same behavior for all interfaces.
>This doesn't affect analysis at all, or practically anything else.
>Remember the interface is used by ChessBase and four different
>professional programmers, all of whom want to provide their users the
>best playing strength they can, more often than not know what they are
>doing, and have no tendency to shoot themselves in the foot.
>The message you read is just part of something the we in Computer Chess
>like to do even more than write programs or play games: find reasons why
>the result of a game/match doesn't count/is not fair/is not
>significant/whatever. This one is reason 46 or so, and has just now been
>invented. The objective of this game is not to improve your play or your
>results but to prove conclusively that any result any program scores
>anywhere doesn't count. While this game makes for interesting debate you
>should not confuse it with anything actually happening in the real
>world.
Hope you don't adapt to that cynism. There's no reason at all. Critic's
a trivial consequence if you are someone who achieved something and now
sell your product. Your little tolerance of frustration is most
unwanted. Of course people want to criticise what you do, hair style and
all. Shirt always open down to 2 buttons ... :)
What is not so funny that is your painting here with the broad brush. Of
course it's nonsense to criticise without even giving it a try to enjoy
the good sides of the products, without even playing against it, BTW a
defamation of Hyatt against me personally (but I think I'm one of the
very few here who's able to play the actual machines without losing at
move 12 with a piece down), but it must be possible to look a bit closer
at the usually weak methods of our operators. Bob claims that it's good
if it was already used in older ACM events. Well, I'm just in train to
demonstrate the stupidiness of that theory. It would be good if you
could write a little bit more carefully what you mean. Otherwise people
will continue their own style of defaming justifying it with quotes out
of your post here. While I know for sure that you didn't want to stop
any critic at all. A critic is even better if it's proven wrong than the
false piece of a cemetary because a critic is the evidence that people
try to think for themselves. We have enough censorship. And I think
Steve Lopez spoke it out, in CCC it's more about posting "lovely"
acclamations without deeper sense. It seems (from the behavior of the
programmers also here on rgcc) that they prefer such more "brown-nosing"
[TM by ChrisW and Rolf] posts to any form of critic, justified in itself
or not ...
I hope that you are not in that camp yet.
>Amir
Rolf Tueschen
<75bdhb$sgl$1...@juniper.cis.uab.edu>:
>I hate to see you bail out. I'm not accusing you of anything.
This is a lie. A lie if one takes for granted that the one who said so
is a university professor.
Here is my personal proof of the lie.
I want to make understandable the technique of Bob's defamation. A
technique he already used many times before-. (Only for the evidence, I
take the post 7590ar$aq9$2...@juniper.cis.uab.edu as most recent example.)
To define that technique with one sentence, let me express it like this:
Bob insinuates without any justification a certain allegedly made
statement, which was never made in reality, and then he goes over this
(foolish, narrow-minded) invention of a position with all his limited
lingual tools to prove that his collegue (here no one less than the
actual World Champion) is a stupid beginner who believes in nonsense.
That's indirectly the meaning of Bob Hyatt's defamation.
No surprise that he then denies any thing spectacular , or something
close to accusation, not to even think of insults.
The truth is however, that the whole style of Mr. Hyatt is the insult
itself. Because he treats the actual Champion as a nutcase. BTW in the
same style he defamed Kasparov as jackass, not enough, as an asshole.
This is so low, so mean, that Bob is a single insult for his own actual
stae of mind. A state who could best be characterized with mental
confusion ... For me it seems as if Bob still has an axe to grind with
Amir. So, he's losing his contenance and he starts to shoot from the
hip.
(For the sceptical reader, please check out for Bob's latest post on
that topic where he suddenly states that he doesn't know for sure what's
going on there. Well, in former posts he had already declared the
ChessBase programming as very poor and crippled. BTW a typical 'Hyatt'
who like to condemn before even a fair trial has begun.)
mclane
>This is a lie. A lie if one takes for granted that the one who said so
>is a university professor.
from a liar like you, claiming somebody else is a liar looks very
funny, isn't it ?
>Here is my personal proof of the lie.
your personal lie :-)))
> Because he treats the actual Champion as a nutcase. BTW in the
>same style he defamed Kasparov as jackass, not enough, as an asshole.
And hyatt is right. kasparov is an idiot, and an arrogant asshole.
what's wrong with this ?
> Well, in former posts he had already declared the
>ChessBase programming as very poor and crippled. BTW a typical 'Hyatt'
>who like to condemn before even a fair trial has begun.)
but he is right too. any idiot can see that this product is programmed
very crippled.
best wishes
Komputer Korner
programmers to conform to an engine vs engine format on one machine.
Whether it is a Fritz GUI or somebody's elses, it will happen and is
happening now with all the new engines that run in the Fritz GUI.
However the engine engine interface must be fair and transparent to
everyone. Since most people have only 1 computer or do not have 2
computers that run at the same speed, one computer engine vs engine
will be the popular form of testing. In fact it is a good way to learn
chess as you are watching IMs play against each other.
--
--
Komputer Korner
The inkompetent komputer
To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Enrique Irazoqui wrote in message
<75ehpo$bo4$1...@diana.bcn.ibernet.es>...
Robert Hyatt
: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote in
: <75bdhb$sgl$1...@juniper.cis.uab.edu>:
:>I hate to see you bail out. I'm not accusing you of anything.
: This is a lie. A lie if one takes for granted that the one who said so
: is a university professor.
Rolf, shut the hell up. You are just about as full of crap as a
Christmas turkey. If you look at what I wrote, I *never* accused
anyone of doing anything at all, other than being guilty of very
sloppy programming. And I know that Amir didn't write the interface
code anyway.
So please *buzz off* with your bullshit logic.
: Here is my personal proof of the lie.
The only thing this proves is that you are a complete idiot... and
that didn't need any proof at all..
: I want to make understandable the technique of Bob's defamation. A
: technique he already used many times before-. (Only for the evidence, I
: take the post 7590ar$aq9$2...@juniper.cis.uab.edu as most recent example.)
: To define that technique with one sentence, let me express it like this:
: Bob insinuates without any justification a certain allegedly made
: statement, which was never made in reality, and then he goes over this
: (foolish, narrow-minded) invention of a position with all his limited
: lingual tools to prove that his collegue (here no one less than the
: actual World Champion) is a stupid beginner who believes in nonsense.
: That's indirectly the meaning of Bob Hyatt's defamation.
: No surprise that he then denies any thing spectacular , or something
: close to accusation, not to even think of insults.
: The truth is however, that the whole style of Mr. Hyatt is the insult
: itself. Because he treats the actual Champion as a nutcase. BTW in the
: same style he defamed Kasparov as jackass, not enough, as an asshole.
: This is so low, so mean, that Bob is a single insult for his own actual
: stae of mind. A state who could best be characterized with mental
: confusion ... For me it seems as if Bob still has an axe to grind with
: Amir. So, he's losing his contenance and he starts to shoot from the
: hip.
: (For the sceptical reader, please check out for Bob's latest post on
: that topic where he suddenly states that he doesn't know for sure what's
: going on there. Well, in former posts he had already declared the
: ChessBase programming as very poor and crippled. BTW a typical 'Hyatt'
: who like to condemn before even a fair trial has begun.)
I certainly did declare their programming to be "poor". And there
is ample evidence of this, as I have been sent log files that support
this, as well as comments by Heiko and many others that investigated
this quite carefully. I did *not* say they intentionally did this to
obtain an advantage in engine vs engine matches.
Your statements are just as bogus as always... My posts are in deja
for those wanting to see the original, rather than the "mangled"
interpretation offered by our mentally deficient critic...
Chris Whittington
Robert Hyatt wrote in message <75hjo6$sr4$1...@juniper.cis.uab.edu>...
>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
>: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote in
>: <75bdhb$sgl$1...@juniper.cis.uab.edu>:
>
>:>I hate to see you bail out. I'm not accusing you of anything.
>
>: This is a lie. A lie if one takes for granted that the one who said so
>: is a university professor.
>
>Rolf, shut the hell up. You are just about as full of crap as a
>Christmas turkey.
Well, I don't know about your Christmas turkeys out in Alabama, but mine
will be full of sage and onion stuffing at one end and chestnut stuffing at
the other. My wife will be up all night performing the stuffing operation,
like she does every year.
If my wife stuffed our turkey full of crap, well, I'ld assume she was trying
to tell me something :)
Chris Whittington
mclane
>If my wife stuffed our turkey full of crap, well, I'ld assume she was trying
>to tell me something :)
>Chris Whittington
! :-))))))) now i know why it is better not to be married at all.
best wishes
Robert Hyatt
: Robert Hyatt wrote in message <75hjo6$sr4$1...@juniper.cis.uab.edu>...
:>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
:>: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote in
:>: <75bdhb$sgl$1...@juniper.cis.uab.edu>:
:>
:>:>I hate to see you bail out. I'm not accusing you of anything.
:>
:>: This is a lie. A lie if one takes for granted that the one who said so
:>: is a university professor.
:>
:>Rolf, shut the hell up. You are just about as full of crap as a
:>Christmas turkey.
: Well, I don't know about your Christmas turkeys out in Alabama, but mine
: will be full of sage and onion stuffing at one end and chestnut stuffing at
: the other. My wife will be up all night performing the stuffing operation,
: like she does every year.
: If my wife stuffed our turkey full of crap, well, I'ld assume she was trying
: to tell me something :)
: Chris Whittington
I was thinking of "live" turkeys. :) they do get stuffed, but mainly
with "turkey feed" to make 'em fat. And as a result of all that feed,
they are pretty "full"...
:)