Auto232
Chris Whittington
As a response to Stefan's post to CCC (sorry, but I'm banned from posting
there) I can confirm, despite nonsensical allegations to the contrary from
those who should know better:
1. the auto232 code supplied by Stefan works just fine.
2. with some few mods and a little time by a gui programmer, cstal windows
(version) ii, had had all this code installed.
3. we added a secret feature which makes havoc with autoplayer learning
(more properly called cheating, I suspect) of certain programs. this sceret
feature wil be revealed on release.
4. auto232 allows also lay by network, as well as by serial cable.
5. by serial cable you can play against any program supporting the standard.
And not just if the opponent is on windows, it also works with the old DOS
programs. Giving win95 to DOS autoplay capability.
6. You can play auto games between any or all of the following (probably not
a complete list)
Shredder
CSTal
Genius
Fritz (before 5)
Mchess
Rebel (not 10 apparently)
Nimzo
Gandalf
Hiarcs
Comet
and loads of others.
Specifically not supported:
Crafty. Too difficult allegedly.
Rebel 10. Ed gives his reasons elsewhere.
Fritz5. They only want to be able to play secretly.
Chessmaster series. Not for the freaks ?
But, for sure, these is enought support for auto232 for enough games between
enough programs; and any new programmer can get the source and drop it in
withina few days. Thus openign up development opportunities by test and
game analysis.
Don't knock it.
Chris Whittington
Posted by Stefan Meyer-Kahlen on September 12, 1998 at 05:11:14:
I have posted this here a couple of times now, but still I get the
impression
that is was not properly understood:
I have released the source code for an autoplayer for Win32 free of charge,
everybody can add it to his program, even though he/she wants to sell it
(the
program, not my source code).
The autoplayer is compatible with the standard auto232 interface and also
with
the chess232 chessboard. (That means if you want to play autoplayer games
with
your program against eg. Hiarcs6 you can do so if you have a serial cable).
If
you use Win32 you don't have to take care about any timing problems or
strange
move formats. All of this is handled by my code.
If you don't have the auto232 cable and want to play on two computers which
are
connected in a network, fine, this can also be done with my code. (Obviously
you
can't play against eg. MChessPro, but against any program that comes with my
code). You can even play autoplayer games on one machine. (For the last two
options you need at least one WindowsNT operating system).
Example:
If somebody adds this code to Crafty you can play autoplayer games in a
network
(no serial cable) against Shredder or against Genius5 on another machine
connect
with a serial cable. All with the same code.
It is really not rocket science to add this autoplayer to your program, it
takes
two or three hours maximum.
If you are complaining that it is Windows only:
It should be no(!) problem to write a similiar driver for Unix or whatever,
that
can be used by all programs running on this operating system, but please
don't
expect me to do this or blame me for not doing it.
I don't know why I have to promote my code so much. I don't get any money
from
this, get it, it's free!!!!!!!
Stefan
Robert Hyatt
: As a response to Stefan's post to CCC (sorry, but I'm banned from posting
: there) I can confirm, despite nonsensical allegations to the contrary from
: those who should know better:
As the one who originally complained about auto232, I'll respond. Note
that I *specifically* addressed "auto232" and not "winauto232". auto232
is poorly designed, from the ground up. The protocol is unnecessarily
complicated, requiring a tab here, a space there, etc. It does *not*
allow underpromotion of any kind, which is rediculous. It has timing
bugs that make moving too quickly hang it. Or moving too slowly causes
it to time out. Just like Goldilocks and the three bears, it has to be
"just right" to work.
: 1. the auto232 code supplied by Stefan works just fine.
Never said it didn't, although it is based on a protocol that is flawed
badly. Stefan did fix the underpromotion problem, but only if you play
a win version vs a win version, not a dos version. But the protocol is
still byzantine and unnecessarily complicated. Winboard/Xboard interface
protocol is very easy to understand, with keyword commands, and specific
(and easy to find) commands to the engine, and from the engine. Including
information about the opponent's time and so forth...
: 2. with some few mods and a little time by a gui programmer, cstal windows
: (version) ii, had had all this code installed.
: 3. we added a secret feature which makes havoc with autoplayer learning
: (more properly called cheating, I suspect) of certain programs. this sceret
: feature wil be revealed on release.
I wouldn't call this a feature that wrecks havoc. I'd call it a feature that
exposes bugs in the other engine. IE *I* don't depend on a game-end signal
to do my learning, because players on ICC learned they could disconnect and
avoid the learning stuff happening. Not any longer...
: 4. auto232 allows also lay by network, as well as by serial cable.
The code I've seen doesn't... it uses serial-port specific programming
which won't fly on a network connection...
: 5. by serial cable you can play against any program supporting the standard.
: And not just if the opponent is on windows, it also works with the old DOS
: programs. Giving win95 to DOS autoplay capability.
: 6. You can play auto games between any or all of the following (probably not
: a complete list)
: Shredder
: CSTal
: Genius
: Fritz (before 5)
: Mchess
: Rebel (not 10 apparently)
: Nimzo
: Gandalf
: Hiarcs
: Comet
: and loads of others.
: Specifically not supported:
: Crafty. Too difficult allegedly.
Crafty supports the old auto232 protocol. I have been looking at Stefan's
code, but the commentary is in German, and it is all windows-based, which
makes it impossible for me to test at all...
: Rebel 10. Ed gives his reasons elsewhere.
: Fritz5. They only want to be able to play secretly.
: Chessmaster series. Not for the freaks ?
: But, for sure, these is enought support for auto232 for enough games between
: enough programs; and any new programmer can get the source and drop it in
: withina few days. Thus openign up development opportunities by test and
: game analysis.
: Don't knock it.
: Chris Whittington
I knock what is lousy design. I doubt you'll find *anyone* that will
say "what a clean elegant design this is, easy to understand, easy to
figure out exactly what you have to send, easy to figure out how to
introduce delays so it won't hang, etc."
It is useful. It is also a lousy piece of software from the design to
the implementation. Yes I understand that it was originally done to
make programs that wouldn't talk to each other play with each other without
knowing. And that was clever. But the "protocol" is bad. And it has
enough bugs to keep a flock of birds fed...
: Posted by Stefan Meyer-Kahlen on September 12, 1998 at 05:11:14:
: Stefan
--
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
Chris Whittington
Robert Hyatt wrote in message <6te1o7$dma$3...@juniper.cis.uab.edu>...
>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>
>: As a response to Stefan's post to CCC (sorry, but I'm banned from posting
>: there) I can confirm, despite nonsensical allegations to the contrary
from
>: those who should know better:
>
>As the one who originally complained about auto232, I'll respond. Note
>that I *specifically* addressed "auto232" and not "winauto232". auto232
>is poorly designed, from the ground up. The protocol is unnecessarily
>complicated, requiring a tab here, a space there, etc. It does *not*
>allow underpromotion of any kind, which is rediculous. It has timing
>bugs that make moving too quickly hang it. Or moving too slowly causes
>it to time out. Just like Goldilocks and the three bears, it has to be
>"just right" to work.
Doesn't matter. It works. For everybody except you.
>
>: 1. the auto232 code supplied by Stefan works just fine.
>
>Never said it didn't, although it is based on a protocol that is flawed
>badly. Stefan did fix the underpromotion problem, but only if you play
>a win version vs a win version, not a dos version. But the protocol is
>still byzantine and unnecessarily complicated.
Doesn't matter. It works. You get the source code. Its in C. You drop it in.
Magic.
Anybody can do it. Except ....
> Winboard/Xboard interface
>protocol is very easy to understand, with keyword commands, and specific
>(and easy to find) commands to the engine, and from the engine. Including
>information about the opponent's time and so forth...
>
>
>: 2. with some few mods and a little time by a gui programmer, cstal
windows
>: (version) ii, had had all this code installed.
>
>: 3. we added a secret feature which makes havoc with autoplayer learning
>: (more properly called cheating, I suspect) of certain programs. this
sceret
>: feature wil be revealed on release.
>
>
>I wouldn't call this a feature that wrecks havoc. I'd call it a feature
that
>exposes bugs in the other engine. IE *I* don't depend on a game-end signal
>to do my learning, because players on ICC learned they could disconnect and
>avoid the learning stuff happening. Not any longer...
Sorry, you didn't understand me. All will become clear on prg release.
>
>: 4. auto232 allows also lay by network, as well as by serial cable.
>
>The code I've seen doesn't... it uses serial-port specific programming
>which won't fly on a network connection...
Who cares ? Bullet is irelevant. For blitz and up 'not flying' is
unimportant. Anyway, you could always ask Cray for a 64,000,000X speed
harddrive and machine time for the comunications if you wanted.
>
>: 5. by serial cable you can play against any program supporting the
standard.
>: And not just if the opponent is on windows, it also works with the old
DOS
>: programs. Giving win95 to DOS autoplay capability.
>
>: 6. You can play auto games between any or all of the following (probably
not
>: a complete list)
>
>: Shredder
>: CSTal
>: Genius
>: Fritz (before 5)
>: Mchess
>: Rebel (not 10 apparently)
>: Nimzo
>: Gandalf
>: Hiarcs
>: Comet
>
>: and loads of others.
>
>: Specifically not supported:
>
>: Crafty. Too difficult allegedly.
>
>Crafty supports the old auto232 protocol. I have been looking at Stefan's
>code, but the commentary is in German,
The gui rogrammer here doesn't speak german either, and he managed just
fine.
>and it is all windows-based, which
>makes it impossible for me to test at all...
You mean you have NO win95 machines ? Oh sorry I forgot *you* run the
department :)))
>
>: Rebel 10. Ed gives his reasons elsewhere.
>: Fritz5. They only want to be able to play secretly.
>: Chessmaster series. Not for the freaks ?
>
>
>: But, for sure, these is enought support for auto232 for enough games
between
>: enough programs; and any new programmer can get the source and drop it in
>: withina few days. Thus openign up development opportunities by test and
>: game analysis.
>
>: Don't knock it.
>
>: Chris Whittington
>
>
>
>
>I knock what is lousy design. I doubt you'll find *anyone* that will
>say "what a clean elegant design this is, easy to understand, easy to
>figure out exactly what you have to send, easy to figure out how to
>introduce delays so it won't hang, etc."
Who cares ? Just drop in the code and go .....
>
>It is useful. It is also a lousy piece of software from the design to
>the implementation.
God this is hopeless .......
>Yes I understand that it was originally done to
>make programs that wouldn't talk to each other play with each other without
>knowing. And that was clever. But the "protocol" is bad. And it has
>enough bugs to keep a flock of birds fed...
Well, we got it working just fine. Autoplay to our little hearts content
with all manner of opponents.
But then I'm not a computer scientist.
Chris Whittington
Robert Hyatt
: Doesn't matter. It works. For everybody except you.
No it doesn't, not perfectly. It works fine for crafty if the game
is played with no endgame databases. If they are used, it will hang.
It is hanging for *other* programs that use tablebases *in* the
search as well...
:>: 1. the auto232 code supplied by Stefan works just fine.
:>
:>Never said it didn't, although it is based on a protocol that is flawed
:>badly. Stefan did fix the underpromotion problem, but only if you play
:>a win version vs a win version, not a dos version. But the protocol is
:>still byzantine and unnecessarily complicated.
: Doesn't matter. It works. You get the source code. Its in C. You drop it in.
: Magic.
: Anybody can do it. Except ....
Yes, you can drop in "crap" and it sort of works. But it is *still* poorly
conceived... some examples:
underpromotion?
offering/accepting a draw?
changing the "protocol system" to work with ethernet? Have to recompile
*all* the engines...
What about time? can you find out how much time your opponent actually
has?
Can you find out the name of your opponent?
Can you find the rating of your opponent to set the contempt factor
appropriately?
etc...
etc...
:>
:>: 4. auto232 allows also lay by network, as well as by serial cable.
:>
:>The code I've seen doesn't... it uses serial-port specific programming
:>which won't fly on a network connection...
: Who cares ? Bullet is irelevant. For blitz and up 'not flying' is
: unimportant. Anyway, you could always ask Cray for a 64,000,000X speed
: harddrive and machine time for the comunications if you wanted.
I'd just as soon spend 50 bucks apiece for two 100mbit/sec ethernet
interfaces and connect them together and use TCP/IP...
But you misunderstood "won't fly". change that to "won't work" and
you got it. IE I'd like to separate the interface from the chess
program so that the interface can be modified to add something new,
without any other pieces of code having to be modified, distributed,
etc..
: You mean you have NO win95 machines ? Oh sorry I forgot *you* run the
: department :)))
Yep. we are *unix*.
:>
:>: Rebel 10. Ed gives his reasons elsewhere.
: Chris Whittington
Chris Whittington
Robert Hyatt wrote in message <6te9vt$fsp$1...@juniper.cis.uab.edu>...
>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>
>: Doesn't matter. It works. For everybody except you.
>
>
>No it doesn't, not perfectly. It works fine for crafty if the game
>is played with no endgame databases. If they are used, it will hang.
>It is hanging for *other* programs that use tablebases *in* the
>search as well...
Algorithm:
In autoplayer mode and table base move got then insert 500 ms delay.
>
>:>: 1. the auto232 code supplied by Stefan works just fine.
>:>
>:>Never said it didn't, although it is based on a protocol that is flawed
>:>badly. Stefan did fix the underpromotion problem, but only if you play
>:>a win version vs a win version, not a dos version. But the protocol is
>:>still byzantine and unnecessarily complicated.
>
>: Doesn't matter. It works. You get the source code. Its in C. You drop it
in.
>: Magic.
>: Anybody can do it. Except ....
>
>Yes, you can drop in "crap" and it sort of works. But it is *still* poorly
>conceived... some examples:
>
>underpromotion?
Each DOS prg had different keyboard action. Hiarcs was menus. Genius had
shortcuts. Mchess had other shortcuts etc.
So a driver was written for each of the standard prgs, and a noname driver
for the others.
Just be compatible with a noname prg. Thorsten can tell you how.
>
>offering/accepting a draw?
Big deal for autoplayer games between comps. Play it out.
>
>changing the "protocol system" to work with ethernet? Have to recompile
>*all* the engines...
Get a standard system, like the rest of us.
>
>What about time? can you find out how much time your opponent actually
>has?
Unless you want to mess around with bullet, this don't matter.
>
>Can you find out the name of your opponent?
No. It would allow serious cheating by programmers, wouldn't it ?
>
>Can you find the rating of your opponent to set the contempt factor
>appropriately?
If I set up the autoplayer, then I know the (goddam) name of the opponent.
Doh !
Adjust your prg as need be. Or do we want SSDF lists where each prg knows
who it is playing ? Even worse state then, no ?
>
>etc...
>
>
>etc...
>
Normally we sack negative and pessimistic no-can-do programmers.
>
>
>:>
>:>: 4. auto232 allows also lay by network, as well as by serial cable.
>:>
>:>The code I've seen doesn't... it uses serial-port specific programming
>:>which won't fly on a network connection...
>
>: Who cares ? Bullet is irelevant. For blitz and up 'not flying' is
>: unimportant. Anyway, you could always ask Cray for a 64,000,000X speed
>: harddrive and machine time for the comunications if you wanted.
>
>I'd just as soon spend 50 bucks apiece for two 100mbit/sec ethernet
>interfaces and connect them together and use TCP/IP...
>But you misunderstood "won't fly". change that to "won't work" and
>you got it. IE I'd like to separate the interface from the chess
>program so that the interface can be modified to add something new,
>without any other pieces of code having to be modified, distributed,
>etc..
>
Then wait for ever for an ideal world where we worked it all out beforehand.
The rest of us will get on on a can-do will-do basis.
Chris Whittington
>
>
>: You mean you have NO win95 machines ? Oh sorry I forgot *you* run the
>: department :)))
>
>Yep. we are *unix*.
>
>
>:>
>:>: Rebel 10. Ed gives his reasons elsewhere.
Robert Hyatt
: Robert Hyatt wrote in message <6te9vt$fsp$1...@juniper.cis.uab.edu>...
:>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
:>
:>: Doesn't matter. It works. For everybody except you.
:>
:>
:>No it doesn't, not perfectly. It works fine for crafty if the game
:>is played with no endgame databases. If they are used, it will hang.
:>It is hanging for *other* programs that use tablebases *in* the
:>search as well...
: Algorithm:
: In autoplayer mode and table base move got then insert 500 ms delay.
Algorithm *fails*. I have a delay command in Crafty, now, to allow
users to adjust this. And it has to be dinked around with on *every*
different machine is it run on. IE one value for a P5/90mhz, another
for a P5/166mhz, another for a P5/166mhz/MMX machine...
Does that sound reasonable? Not to moi...
: Each DOS prg had different keyboard action. Hiarcs was menus. Genius had
: shortcuts. Mchess had other shortcuts etc.
: So a driver was written for each of the standard prgs, and a noname driver
: for the others.
: Just be compatible with a noname prg. Thorsten can tell you how.
:>
:>offering/accepting a draw?
: Big deal for autoplayer games between comps. Play it out.
It's part of "the rules of chess". It ought to be handled
properly. Under winboard/xboard it is...
: Get a standard system, like the rest of us.
"standard" nowadays *is* ethernet-based. For those that have more than
one machine. I have a two-machine network at Home. I believe Bruce has
done the same... and many of these "matches" are done by people with
idle machines at work, all of which are generally networked.
This is the "future" not the "past"...
:>
:>What about time? can you find out how much time your opponent actually
:>has?
: Unless you want to mess around with bullet, this don't matter.
It does near the end of time controls...
:>
:>Can you find out the name of your opponent?
: No. It would allow serious cheating by programmers, wouldn't it ?
Again, it is part of "the rules of chess". I've *never* played in a
human tournament without knowing who I was playing...
:>
:>Can you find the rating of your opponent to set the contempt factor
:>appropriately?
: If I set up the autoplayer, then I know the (goddam) name of the opponent.
: Doh !
It is not *me* I am working on. It is the thousands of other folks that
will do this... and won't know how to set the name, nor the rating, nor
even know that it is important...
: Adjust your prg as need be. Or do we want SSDF lists where each prg knows
: who it is playing ? Even worse state then, no ?
Not sure, but it would be reasonable to know your opponent...
:>
:>etc...
:>
:>
:>etc...
:>
: Normally we sack negative and pessimistic no-can-do programmers.
Good idea... particularly since *I* can do *all* of the above with the
interface many of us have worked out with Tim Mann. So *we* don't have
the problem, it is the auto232 users that do...
:>
:>
:>:>
:>:>: 4. auto232 allows also lay by network, as well as by serial cable.
:>:>
:>:>The code I've seen doesn't... it uses serial-port specific programming
:>:>which won't fly on a network connection...
:>
:>: Who cares ? Bullet is irelevant. For blitz and up 'not flying' is
:>: unimportant. Anyway, you could always ask Cray for a 64,000,000X speed
:>: harddrive and machine time for the comunications if you wanted.
:>
:>I'd just as soon spend 50 bucks apiece for two 100mbit/sec ethernet
:>interfaces and connect them together and use TCP/IP...
:>But you misunderstood "won't fly". change that to "won't work" and
:>you got it. IE I'd like to separate the interface from the chess
:>program so that the interface can be modified to add something new,
:>without any other pieces of code having to be modified, distributed,
:>etc..
:>
: Then wait for ever for an ideal world where we worked it all out beforehand.
: The rest of us will get on on a can-do will-do basis.
: Chris Whittington
Rolf Tueschen
>Again, it is part of "the rules of chess". I've *never* played in a
>human tournament without knowing who I was playing...
Kasparov was not so lucky against Deep Blue. He was lost in Ghoast City.
Sorry, I give correction. He also knew the name of his opponent and the
street where he had to play him ...
Robert Hyatt
: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
He knew *exactly* "what" he was playing. There were hundreds of deep
thought (and older deep blue) games available. Just nothing from the
very latest version. Of course, he didn't exactly make *his* pre-match
plans public either, did he?
Chris Whittington
Robert Hyatt wrote in message <6tf0s5$mem$2...@juniper.cis.uab.edu>...
>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>
>: Robert Hyatt wrote in message <6te9vt$fsp$1...@juniper.cis.uab.edu>...
>:>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>:>
>:>: Doesn't matter. It works. For everybody except you.
>:>
>:>
>:>No it doesn't, not perfectly. It works fine for crafty if the game
>:>is played with no endgame databases. If they are used, it will hang.
>:>It is hanging for *other* programs that use tablebases *in* the
>:>search as well...
>
>: Algorithm:
>
>: In autoplayer mode and table base move got then insert 500 ms delay.
>
>
>users to adjust this. And it has to be dinked around with on *every*
>different machine is it run on. IE one value for a P5/90mhz, another
>for a P5/166mhz, another for a P5/166mhz/MMX machine...
>
>Does that sound reasonable? Not to moi...
>
>
>: shortcuts. Mchess had other shortcuts etc.
>
>: So a driver was written for each of the standard prgs, and a noname
driver
>: for the others.
>
>: Just be compatible with a noname prg. Thorsten can tell you how.
>
>:>
>:>offering/accepting a draw?
>
>: Big deal for autoplayer games between comps. Play it out.
>
>properly. Under winboard/xboard it is...
>
>
>: Get a standard system, like the rest of us.
>
>"standard" nowadays *is* ethernet-based. For those that have more than
>one machine. I have a two-machine network at Home. I believe Bruce has
>done the same... and many of these "matches" are done by people with
>idle machines at work, all of which are generally networked.
>
>This is the "future" not the "past"...
>
>:>
>:>has?
>
>: Unless you want to mess around with bullet, this don't matter.
>
>
>:>
>
>: No. It would allow serious cheating by programmers, wouldn't it ?
>
>human tournament without knowing who I was playing...
>
>:>
>:>appropriately?
>
>: If I set up the autoplayer, then I know the (goddam) name of the
opponent.
>: Doh !
>
>
>will do this... and won't know how to set the name, nor the rating, nor
>even know that it is important...
>
>: who it is playing ? Even worse state then, no ?
>
>
>
>:>
>:>
>:>
>:>etc...
>:>
>
>: Normally we sack negative and pessimistic no-can-do programmers.
>
>interface many of us have worked out with Tim Mann. So *we* don't have
>the problem, it is the auto232 users that do...
Let's look it this from another level .....
There's an interface that kind of works. Or works enough for N thousand ssdf
games, or N 100's of Ed Schroder games, or N 100s of chrisw autoplayer
games, or, or, or. And links over 20 programs together in a mass testing (we
won't discuss the testing methods).
For you there are 10 million little niggly reasons to not like it, campaign
against it on ccc and rgcc, generally knock it, make dumb arguments with
just about anybody why we shouldn't use it, propose some other standard
(named after yourself no doubt) and more bla-bla.
So we have something concrete, accurate, comp scientistically correct,
simple, according to correct software engineering principles. The Hyatt
Autoplayer (Vapourware Inc.).
And we have a kind of fuzzy thing that works just about. The auto232 (tried,
tested and shipping now).
Interesting dichotomy, no ? Shall we look for parallels ?
Something human and something mechanical and controllable ?
The bean-counter paradigm and a fuzzy-knowledge, dynamic approach ?
The security and safety of the Censored-Chicken Coup against freedom and
speech and dangerous dissenters ?
Chris Whittington (not putting it all into compartments)
>
>
>:>
>:>
>:>:>
>:>:>: 4. auto232 allows also lay by network, as well as by serial cable.
>:>:>
>:>:>The code I've seen doesn't... it uses serial-port specific programming
>:>:>which won't fly on a network connection...
>:>
>:>: Who cares ? Bullet is irelevant. For blitz and up 'not flying' is
>:>: unimportant. Anyway, you could always ask Cray for a 64,000,000X speed
>:>: harddrive and machine time for the comunications if you wanted.
>:>
>:>I'd just as soon spend 50 bucks apiece for two 100mbit/sec ethernet
>:>interfaces and connect them together and use TCP/IP...
>:>But you misunderstood "won't fly". change that to "won't work" and
>:>you got it. IE I'd like to separate the interface from the chess
>:>program so that the interface can be modified to add something new,
>:>without any other pieces of code having to be modified, distributed,
>:>etc..
>:>
>
>: Then wait for ever for an ideal world where we worked it all out
beforehand.
>: The rest of us will get on on a can-do will-do basis.
>
>
>: Chris Whittington
>
>
Robert Hyatt
: Let's look it this from another level .....
: There's an interface that kind of works. Or works enough for N thousand ssdf
: games, or N 100's of Ed Schroder games, or N 100s of chrisw autoplayer
: games, or, or, or. And links over 20 programs together in a mass testing (we
: won't discuss the testing methods).
Never said it doesn't "sort of work". Just that the original auto232, as
used in the dos engines, has a couple of really substantial bugs, namely
no underpromotion and some timing problems that still cause even the
commercial programs to hang at times (hence the auto232 time-out option).
: For you there are 10 million little niggly reasons to not like it, campaign
: against it on ccc and rgcc, generally knock it, make dumb arguments with
: just about anybody why we shouldn't use it, propose some other standard
: (named after yourself no doubt) and more bla-bla.
Nope. Don't do the "name it after myself stuff.." Will leave that to
you. But I certainly reserve the right to point out serious flaws in
something that I'd like to use. But it doesn't matter. Just wait for a
year or two and see what happens. I already know of *one* commercial
engine that is in the process of implementing the winboard/xboard
protocol standard. Since it is used by many already. And since it is
missing the problems that exist in the dos auto232. And since it removes
the "interface" from the program. No special code to call or anything, just
some specific messages to parse that come from stdin.
: So we have something concrete, accurate, comp scientistically correct,
: simple, according to correct software engineering principles. The Hyatt
: Autoplayer (Vapourware Inc.).
Not fuzzy at all. Something supported by at least a dozen programs, *not*
named after me, *not* created by me, but at least *complete* enough to make
a reasonable program-to-program communication protocol that is capable of
supplying all the necessary information to allow the chess engines to play
chess automatically with the same information they would have if played
in a manual mode...
: And we have a kind of fuzzy thing that works just about. The auto232 (tried,
: tested and shipping now).
: Interesting dichotomy, no ? Shall we look for parallels ?
Yes... something well-defined and complete vs something ill-defined,
incomplete except that it comes complete with bugs...
: Something human and something mechanical and controllable ?
: The bean-counter paradigm and a fuzzy-knowledge, dynamic approach ?
Aha. back to that. I'll bite... "The bean-counter paradigm that
*works* the last time I looked..." You like to keep omitting that
"it works" aspect, which is rather important. It's easy to propose
all sorts of word-games. It's harder to propose something that is
demonstrable as "working better than bean-counting." I'm waiting
for evidence that this happens... when it does...
: The security and safety of the Censored-Chicken Coup against freedom and
: speech and dangerous dissenters ?
Why does everything have to be tied to CCC? *very* poor inductive
reasoning there, unless you want to tie this to Einstein's unified field
theory that says *everything* is related.
This started on CCC, because I had been asked several times to look into
the windows of auto232 as a possible solution to the timing problems in the
dos version...
: Chris Whittington (not putting it all into compartments)
Rolf Tueschen
>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
>: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
>:>Again, it is part of "the rules of chess". I've *never* played in a
>:>human tournament without knowing who I was playing...
>: Kasparov was not so lucky against Deep Blue. He was lost in Ghoast City.
>: Sorry, I give correction. He also knew the name of his opponent and the
>: street where he had to play him ...
>He knew *exactly* "what" he was playing. There were hundreds of deep
>thought (and older deep blue) games available. Just nothing from the
>very latest version. Of course, he didn't exactly make *his* pre-match
>plans public either, did he?
But then we're back in questions of Cold War. When I wanted to become a
communist in my early youth (with 6 or 7 years old I guess), it was the CIA
and all the western philosophers of capitalism who told me a clear "No
no!". Because, and now it comes, man will always stay man, and the commies
with their "new man" could not succeed.
Although I couldn't understand at the time why we westerners put *all* our
money on the evil side in man (because my priest told me the complete
opposite and he didn't give up to make a better man out of me...).
It's true. Kasparov could develop the most astonishing plans about his new
play against a computer (and you bet, alas, he did it under the spiritual
influence of Frederic) still in the very last night before the show, but
the main chessthinking of Kasparov as 'man' will always remain the same
after two decades of top chess. Even his little tricks with the necktie,
the watch, will remain the same till the end of his days.
This is *not* the case with Deep Blue. What "they" created with the machine
is simply not possible with the human brain. Let me give you a rough
summary. They took the eyes and the same mouth of course (you remember, the
window for the cryptic output of the machine) seperated the lower parts of
the hard(ware)brain from its upper higher part, threw the upper part into
the garbage and connencted a brandnew high fidelity part with the help of a
*very* hot needle to the basis.
SPLISH! SPLASH! BANG! (I love these ancient classics.)
Then they gave the stethoscope to 'Joe McBunji' and told him, "make'm play
like a real GM". Hehehe.
Of course this was nothing else than our buddy Joe B e n j a m i n, but he
had sort of CIA undercover new name and identity. Thus he could for example
lie without becoming red (sic!) in his face with the help of a new secret
drug of course. So he could tell the audience that he would publish the
scores of his many matches against DB. This was a joke of the CIA of
course. Because would you tell the world how your brain would *really*
really function (quoting of Prof Hyatt)? (;-)
So, please stop the continuous claim that to train a human brain is even
"similar" to fabricating a new machine. Man will always stay the oftenly
failing, lying animal (they taught me to believe this way). But a machine
is different. He/she/it will be better than man ...
That's why I'm in this computerchess group BTW.
"Scully, don't say that this ain't true? Oh, Scully!"
Robert Hyatt
: But then we're back in questions of Cold War. When I wanted to become a
I didn't make such a claim. I simply said this (for the hundredth time,
no doubt):
1. Kasparov wanted complete printouts for the games he played. That
would reveal a *wealth of information about DB*. IE how deep did it
search? Where does it extend? How does it evaluate positions with
respect to material values? ETC...
2. Kasparov would *never* sit down with Hsu and company and say "OK,
I am going to try to play into closed, blocked positions. I'm not going
to pay a lot of attention to deep tactics, rather, I am going to try to
pay attention to short-term positional safety to keep the game under
control. BTW, I believe that doubled rooks on the 7th is worth more than
a pawn if the king is on the 8th rank and there are pawns on the 7th,
etc."
So why does the deep blue team have to provide Kasparov with output that
would certainly help him understand the beast and produce insight into how
it can be beaten, when Kasparov would *never* provide the same information
in a reciprocal manner? It is a rediculous concept, from the beginning.
Rolf Tueschen
>I didn't make such a claim. I simply said this (for the hundredth time,
>no doubt):
>1. Kasparov wanted complete printouts for the games he played. That
>would reveal a *wealth of information about DB*. IE how deep did it
>search? Where does it extend? How does it evaluate positions with
>respect to material values? ETC...
Oh Lord, I could imagine quite easily a method where a clever guy could see
that *all* was ok with DB and nevertheless one couldn't see what you think
(and I'm game here with you) impossible to tell the whole world. But stay
tuned for a minute. I think I've a nice surprise for you.
Here we go:
>2. Kasparov would *never* sit down with Hsu and company and say "OK,
>I am going to try to play into closed, blocked positions. I'm not going
>to pay a lot of attention to deep tactics, rather, I am going to try to
>pay attention to short-term positional safety to keep the game under
>control. BTW, I believe that doubled rooks on the 7th is worth more than
>a pawn if the king is on the 8th rank and there are pawns on the 7th,
>etc."
You really must be joking, Bob. why couldn't he do that? All that is
content of each decent reader about chess for beginners ...
And even you know about the 7th rank. Me too. Methinks you overlooked that
you're almost into Phil's thread about the psychology of the human brain,
because that's it what you needed to know for *real* useful stuff. I mean
if we talk about Kasparovs secrets. But then you pointed out that "we" are
not that far actually, didn't you?
I think that "no", with that theory you can't explain/justify the cheat the
did to Kasparov.
>So why does the deep blue team have to provide Kasparov with output that
>would certainly help him understand the beast
Come down to earth, Bob. You have forgotten my claim, that I needed 6
months to study a fixed version and then be able to hold 'm down? Fixed
means that there would be no tweaking and twisting. Period.
>and produce insight into how
>it can be beaten, when Kasparov would *never* provide the same information
>in a reciprocal manner? It is a rediculous concept, from the beginning.
This is totally wrong, Bob. You must be missing the real difficulty. That
makes me laugh.
Let me explain. It's NOT that Kasparov has 1 million secrets and that when
you were told the 1 million secrets (by himself) that then you were able to
play like Kasparov. Now replace DB for "you". Then DB could NOT play like
Kasparov.
One little example. If you knew for sure that Kasparov could calculate a
dangerous line for 40 plies meaning that he "sees" a danger for his King,
then it wouldn't mean nothing for the play of DB, no? BTW a good GM like
Joel could well tell them about such factors. I mean, Kasparov isn't
playing different chess, I beg you. He's only, or was only, a bit faster
and deeper. And most important, he was n a s t y. But don't try to explain
that one to your computer. Or could you define what nasty means to him or
to me? :)
I fear you just are making the biggest mistake in the history of computer
sciences. I'm sure you didn't mean it this way. Please elaborate that.
Rolf Tueschen
>I didn't make such a claim. I simply said this (for the hundredth time,
>no doubt):
>1. Kasparov wanted complete printouts for the games he played. That
>would reveal
Another little observation about your method of posting politics.
Didn't you claim for years that Garry did in fact get what he had asked
for??
And now suddenly you want to tell me as if you'd never told me different
that he had asked for more and then didn't get it?
Nice little observation, Bob. Noted.
Robert Hyatt
: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
:>I didn't make such a claim. I simply said this (for the hundredth time,
:>no doubt):
:>1. Kasparov wanted complete printouts for the games he played. That
:>would reveal a *wealth of information about DB*. IE how deep did it
:>search? Where does it extend? How does it evaluate positions with
:>respect to material values? ETC...
: Oh Lord, I could imagine quite easily a method where a clever guy could see
: that *all* was ok with DB and nevertheless one couldn't see what you think
: (and I'm game here with you) impossible to tell the whole world. But stay
: tuned for a minute. I think I've a nice surprise for you.
: Here we go:
Think about playing a *game* where you managed to get DB's king somewhat
exposed, or DB got your king somewhat exposed. But nothing came of it.
Then, what if you could see the analysis which showed what DB thought
about it and you notice: "hey, its king was in mild danger here but it
had *no* clue at all and thought it was well ahead." That would reveal
that it was subject to long-term attacks, done correctly. Or suppose
you noticed that it evaluated a passed pawn *very* high (or very low,
doesn't matter here). Then you could adjust your play to let it
create a passed pawn that could run all the way to the 7th rank before
it is stopped, while you shred its king-side as compensation. And since it
evaluates the passed pawn more highly than king safety, you have yet another
angle to beat on it with. (of course, I am *not* saying that DB evaluates
passed pawns higher or lower than king safety here... but you should get the
idea that the output would be *most* revealing and could give a strong player
like Kasparov just the "edge" he needs, while he has given DB *nothing* in
return.)
Game scores are one thing, but "thought processes" are something else
entirely. He knew it was stupid to ask for them. Everyone there knew it
was stupid for him to ask for them. Almost everyone *here* could figure
out that it was stupid for him to ask for them. Because he wouldn't/couldn't
return the favor.
And chess *is* a "zero-sum" game...
:>2. Kasparov would *never* sit down with Hsu and company and say "OK,
:>I am going to try to play into closed, blocked positions. I'm not going
:>to pay a lot of attention to deep tactics, rather, I am going to try to
:>pay attention to short-term positional safety to keep the game under
:>control. BTW, I believe that doubled rooks on the 7th is worth more than
:>a pawn if the king is on the 8th rank and there are pawns on the 7th,
:>etc."
: You really must be joking, Bob. why couldn't he do that? All that is
: content of each decent reader about chess for beginners ...
: And even you know about the 7th rank. Me too. Methinks you overlooked that
: you're almost into Phil's thread about the psychology of the human brain,
: because that's it what you needed to know for *real* useful stuff. I mean
: if we talk about Kasparovs secrets. But then you pointed out that "we" are
: not that far actually, didn't you?
: I think that "no", with that theory you can't explain/justify the cheat the
: did to Kasparov.
What I'm saying is that Kasparov would *not* teach the DB guys how to beat
him. He wouldn't go into enough detail about what he was thinking *during*
the game, so that they could figure out what he had overlooked and modify DB
to take advantage of that. He simply wouldn't do it. He wouldn't even reveal
which opening he was going to play, or that he was going to try to take the
machine out of book very early. Because that would give up an "edge" that
might help him.
:>So why does the deep blue team have to provide Kasparov with output that
:>would certainly help him understand the beast
: Come down to earth, Bob. You have forgotten my claim, that I needed 6
: months to study a fixed version and then be able to hold 'm down? Fixed
: means that there would be no tweaking and twisting. Period.
I'd bet that Kasparov could take two game printouts and find some significant
weakness that he could exploit, *if* he knew exactly how it was evaluating
every position, how deeply it searched, how it was extending, how it balances
"A" vs "B" (above example) and so forth.
:>and produce insight into how
:>it can be beaten, when Kasparov would *never* provide the same information
:>in a reciprocal manner? It is a rediculous concept, from the beginning.
: This is totally wrong, Bob. You must be missing the real difficulty. That
: makes me laugh.
: Let me explain. It's NOT that Kasparov has 1 million secrets and that when
: you were told the 1 million secrets (by himself) that then you were able to
: play like Kasparov. Now replace DB for "you". Then DB could NOT play like
: Kasparov.
No... but Kasparov would give DB *zero* of his secrets. I've *never* met a
GM that would say "I am going to spring this novelty the next time I am black
and my opponent allows me to play the French. He just wouldn't give away a
novelty it took him hours or days to work out, because he'd *know* that I would
try it first chance I got and expose it to the world...
: One little example. If you knew for sure that Kasparov could calculate a
: dangerous line for 40 plies meaning that he "sees" a danger for his King,
: then it wouldn't mean nothing for the play of DB, no? BTW a good GM like
: Joel could well tell them about such factors. I mean, Kasparov isn't
: playing different chess, I beg you. He's only, or was only, a bit faster
: and deeper. And most important, he was n a s t y. But don't try to explain
: that one to your computer. Or could you define what nasty means to him or
: to me? :)
: I fear you just are making the biggest mistake in the history of computer
: sciences. I'm sure you didn't mean it this way. Please elaborate that.
I don't see any mistake. Football teams have closed practices to install
new offenses. Basketball teams have closed practices so that their opponents
can't see their offensive patterns until an important game. In ACM computer
chess events, Bert and I used to spend days looking at our openings so that we
could modify the book and not repeat games from previous years. This idea of
"closed practice" is quite common *everywhere*. When is the last time you
saw two GM's sit down either before or during a match they were playing, and
discuss their strategy together, opening choices, types of endgames to steer the
game toward, whether to play wild attacking chess or slow positional chess, etc?
Answer: *never*.
I support DB's right to not have its opponent "probe into its mind and see
what it was thinking" just as I support *my* right to do the same thing in
any competition. Remember my old karate tournament discussion a while back?
Same thing went on back then, too. Some guys have a propensity to depend more
on kicks, others on momentum moves, others use different attacking postures.
I was not about to tell them what *I* am going to use when we entered the circle
because that was my "surprise". They generally had some worked out for me, too.
But we certainly did *not* sit down before the match and discuss what we were
going to try to do to each other...
Robert Hyatt
: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
:>I didn't make such a claim. I simply said this (for the hundredth time,
:>no doubt):
:>1. Kasparov wanted complete printouts for the games he played. That
:>would reveal
: Another little observation about your method of posting politics.
: Didn't you claim for years that Garry did in fact get what he had asked
: for??
: And now suddenly you want to tell me as if you'd never told me different
: that he had asked for more and then didn't get it?
Nope. Kasparov *originally* asked to see the output for three specific
moves, one where the program didn't play Qb6 to win a pawn, but played
Be4 instead. He received that. So I have *not* changed the story at
all. He should *never* have received *any* printout of any kind. For
reasons I have already given. He certainly shouldn't have (and didn't)
receive complete printouts for the game.
But for the move he most protested, Be4, the analysis was published around
the world...
: Nice little observation, Bob. Noted.
Serge Desmarais
But if DB is never to return, in my opinion it is of no importance
if they show him (Kasparov) the printouts, since he will never be able
to take advantage from them. Of course, if they plan a come back, that
is another story... But one has to question the goal of the match : was
it to know/discover how good the machine was, or was it just to beat
Kasparov at all costs and then disappear, leaving the "impression" that
no other man and machine could beat it? Or worse, that it had kinda
"solved" the game. Since I think that a 6 game match is not enough to
really see what the machine is capable of (and the tweakings between ga,es
makes it even more difficult to evaluate, the answer seems to be the second
one...
Serge Desmarais
Rolf Tueschen
>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
>: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
>:>no doubt):
>:>1. Kasparov wanted complete printouts for the games he played. That
>:>would reveal a *wealth of information about DB*. IE how deep did it
>:>search? Where does it extend? How does it evaluate positions with
>:>respect to material values? ETC...
>: Oh Lord, I could imagine quite easily a method where a clever guy could
see
>: that *all* was ok with DB and nevertheless one couldn't see what you
think
>: (and I'm game here with you) impossible to tell the whole world. But
stay
>: tuned for a minute. I think I've a nice surprise for you.
>: Here we go:
>Think about playing a *game* where you managed to get DB's king somewhat
>exposed, or DB got your king somewhat exposed.
What are you doing here? I think I have understood your "7th rank" stuff.
You seem to miss my main point.
>But nothing came of it.
>Then, what if you could see the analysis which showed what DB thought
>about it and you notice: "hey, its king was in mild danger here but it
>had *no* clue at all and thought it was well ahead." That would reveal
>that it was subject to long-term attacks, done correctly. Or suppose
>you noticed that it evaluated a passed pawn *very* high (or very low,
>doesn't matter here).
Please. I thought I had told you my point with irony. But that seemed to be
a failure. Therefore here in capitals. You don't need to add more examples
on the same level. I understood it, mainly agreed with you, but then tried
to say that this is not the sort of stuff Kasparov needed to exploit
weaknesses of a machine. All these trivial examples are true but also
totally wrong. At least in the sense you tried to use them. To make this
quite clear. For you as a programmer of a specific computerchess program
these details are useful and important. But in human chess the GMs don't
need such petitesses. Because if they had to rely on such tricks or if they
had opponents who could be busted regularly with such trivialities, it
would be no longer GM-chess.
Let me repeat. A real GM doesn't need to wait for the moment when the
"opponent" finally falls into that sort of traps. A GM plays his game no
matter if it's a weak or a strong opponent. If the pressure finally is
rising it's normal that more and more of such tactical tricks become
important. But it would mean nothing to know before the game that opponent
A was a victim of a double attack in zeitnot. A GM doesn't think "well,
let's find a good situation for a double attack for this guy." Why? Simply
because double attacks are the daily bread for each chessplayer, not only a
GM. At first comes the game. The pressure must be created. So, what I'm
tryinfg to tell you is the non-importance of such knowledge.
I see another weak point however in your theory. We know that strong GMs
don't lose to a 7th rank attack because they didn't know the danger of it.
No, the game finally escalates into a weak 7th for one side. But in today's
chess this is never a singular pattern but a complexity of several
patterns. Both players "know" all about the single items. The art is to
find a complex strategy that goes a little step farther the opponent had
thought of.
I'm talking about human chess as Rolf is fantasysing it. How do you think
Kasparov could explain to (whom??) how to beat him theoretically? Let me
help you. I would take his queen at first and then mate him. Period.
>Then you could adjust your play to let it
>create a passed pawn that could run all the way to the 7th rank before
>it is stopped, while you shred its king-side as compensation. And since
it
>evaluates the passed pawn more highly than king safety, you have yet
another
>angle to beat on it with.
Go ahead with crafty this way but don't dream that you could ever come to
an end with such successive steps. With such descriptions you show me that
you still don't think about real GM chess. I fear you talk too often with
GMRoman. Of course he could talk for months about such points. He surely
knows many more than me for example. But I still don't understand what he's
really doing. Keeping you busy for the next years? Or is he trying to
explain his thought process? Watch out, Bob. He's a chess addict.
Why don't you start a cooperation of some sort with Chris? He could tell
you the same about isolated pawns and open files. But he has the links to
the future world of "knowledge". To give you my little warning: Don't
believe that you could add all the steps and hope for an automatic birth of
intelligent chess. It won't function this way.
> (of course, I am *not* saying that DB evaluates
>passed pawns higher or lower than king safety here... but you should get
the
>idea that the output would be *most* revealing and could give a strong
player
>like Kasparov just the "edge" he needs, while he has given DB *nothing*
in
>return.)
Yes, I already bought that. But I also said that one could hide the secret
parts and still leaving the parts open that could prove that all what
happened was kosher. Because that was the point for the asking. NOT to
detect some spooky tricks to exploit. Bob, we're talking about real chess
and not just you and Ed playing 24 h/move games. :)
>Game scores are one thing, but "thought processes" are something else
>entirely.
I knew since long that we had something in common ...
> He knew it was stupid to ask for them. Everyone there knew it
>was stupid for him to ask for them. Almost everyone *here* could figure
>out that it was stupid for him to ask for them. Because he
wouldn't/couldn't
>return the favor.
>And chess *is* a "zero-sum" game...
But now, Bob, it looks this way:
Kasparov, one of thze best players of the world was busted by a team of
mediocre, not even this, by complete chessic ignorants with the exception
of some advising GMs. Busted with the help of a machine that is NEVER of
real GM strength. That is as if Michael Johnson would be beaten by a
Vietnam veteran in his wheel chair...
And then you come and try to argue that it's all Garry's own fault because
he had to know better about the American business. How could he after the
first meetings were held in all sophisticated friendship? It was as if you
was friendly enough to let your bathroom-window open for a guest whom you
didn't want to salute at your front entrance. You can't understand what
unwritten rules the IBM /DB teamsters have violated? Could you please leave
aside your buddy-motivation for a moment and to better concentrate on your
task as scientist? I don't doubt for a second that you're man enough to
fight for "them" but then you pay a too high price in my eyes. You come in
danger with your reputation as a scientist. The same I told you with Chris
and his ideas. I think you have modified your politics a little bit and
that's great. But as far as DB is concerned you should simply try to let
live both sides. Both sides have their pro's and con's. It would honour you
if you would also defend the side you personally don't appreciate at all.
Get the idea?
This wouldn't weaken your position but would only shed a brighter light on
your reputattion as an expert.
Again this is too low stuff. Too trivial. Too average. Tomorrow you tell us
that he won't speak about his wish to win the game ...
>:>So why does the deep blue team have to provide Kasparov with output
that
>:>would certainly help him understand the beast
>: Come down to earth, Bob. You have forgotten my claim, that I needed 6
>: months to study a fixed version and then be able to hold 'm down? Fixed
>: means that there would be no tweaking and twisting. Period.
>I'd bet that Kasparov could take two game printouts and find some
significant
>weakness that he could exploit, *if* he knew exactly how it was
evaluating
>every position, how deeply it searched, how it was extending, how it
balances
>"A" vs "B" (above example) and so forth.
I can't bet because I think and wrote the same. But you sound as if you
would say that Kasparov did ask for the stuff to get an unfair advantage.
That's ridiculous. That's how *I* try to bust FRITZ but Garry has other
talents to win against a machine.
You generally don't refuse to repeat always the same good arguments. So let
me repeat mine. It's ridiculous to make a match about the question mankind
(!) or machine and then behind the curtains transform, better pervert it
into a mystery show where nobody really knew what was real and what was
fake. Garry is still a human being, no? Already your assumption of a clean
"thought-process" one could get from a human player was very superficial. A
human being has emotions, feelings. An invisible artificial opponent alone
is
sufficient to provoke feelings of anxiety. If then -- justified or not --
the impression grows up that something is wrong with the whole setting,
especially for someone who already is believing in supernatural events,
then the asking for the output is not at all a naive nonsense but a last
try to save the situation. History will speak the verdict about the details
of the event.
I have already made my decision. A team of highly intelligent technocrats
with a machine had a meeting with an artist, a great sportsman, a genius. A
human being still. With many faults and many good sides. What I have wished
for a fair show would have been sort of chairman for the machine's team who
at least had a glimpse of insight into the world of a chess genius. Someone
from Russia. From Aserbeidjan. The first ridiculous factor is when a man
called "The Operator" is placed in front of the genius, as if *he* would
play the game against him. Sure the operator has the monitor at his side.
But then in ther meantime he also stares at the board as if he was trying
to think ... This must be changed. It's troubling the basic concentration
of a player. Because normaly he has his real opponent in front. Here he has
to abstract from the operator and the display to the real machine some
kilometers away. That has to be changed. The machine has to be in the same
room where the board and the audience are. The operator has not the right
to be seated at the board but he should have his place in front of the
monitor. He then stands up makes a move on the board and goes away again.
Even better would be a monitor solution next to the board and a neutral
referee would move for the machine. There must not be an operator from the
team itself. This way the situation becomes clearer a fight man vs machine
with the teamster totally in the background.
I described all that not because all that had to be processed like that.
But I wanted to demonstrate a sample of the hundreds of factors that could
be important for a human player and his perception. Focussing on the human
thought process is a good research project but not a sufficient description
of the complete story. This is difficult to explain to someone who never
was himself in such an extreme psychological tension of a chess genius.
>:>and produce insight into how
>:>it can be beaten, when Kasparov would *never* provide the same
information
>:>in a reciprocal manner? It is a rediculous concept, from the
beginning.
Yes, because a chess genius isn't automatically a psychological scientist
of his own self.
>: This is totally wrong, Bob. You must be missing the real difficulty.
That
>: makes me laugh.
>: Let me explain. It's NOT that Kasparov has 1 million secrets and that
when
>: you were told the 1 million secrets (by himself) that then you were
able to
>: play like Kasparov. Now replace DB for "you". Then DB could NOT play
like
>: Kasparov.
>No... but Kasparov would give DB *zero* of his secrets. I've *never* met
a
>GM that would say "I am going to spring this novelty the next time I am
black
>and my opponent allows me to play the French. He just wouldn't give away
a
>novelty it took him hours or days to work out, because he'd *know* that I
would
>try it first chance I got and expose it to the world...
Ah, you're so easily to be instructed as agent provocateur? :)
I grant you that right, Bob.
But then I reserve me the right to tell you that you absolutely missed my
point.
And I must state this with all seriousity. We're not in Las Vegas. We're
not in gambling. And if we play against a machine we want a high
probability that it was the machine who made the moves. Was that clear
enough? We as scientists should support such a triviality. And to make
another thing clear. Because you always repeat that Kasparov had all the
cards to define whatever he wanted ...
We're not in boxing, Bob, where you had better a company of the Green
Barrets at your side in case your opponent would show appetite for your
ears ...
> Remember my old karate tournament discussion a while back?
>Same thing went on back then, too. Some guys have a propensity to depend
more
>on kicks, others on momentum moves, others use different attacking
postures.
>I was not about to tell them what *I* am going to use when we entered the
circle
>because that was my "surprise". They generally had some worked out for
me, too.
>But we certainly did *not* sit down before the match and discuss what we
were
>going to try to do to each other...
But you made a weak statement. Because I don't care, and Kasparov doesn't
care about what (now please pay attention) the *m a c h i n e* is
doing by itself. That's the deal in my eyes. But the machine has to be set
into the playing hall and all the operators, allegators and master(ator)s
have to be placed in the audience like Garry's mother. Far away from the
machine. It's a special misbehaviour that the teamsters tried to pretend
that their cryptic output of the machine fell completely into the
competence of the team itself. What would you say if (I repeat an older
idea) an athlet would claim the right to run 200 miles away and then
transmitted by a TV satellite ...? The moves of a chessplaying machine must
come visible and understandable for a really neutral referee.
And finally all the possible tweaking and personality twisting has to be
done before the match. I mean the human input. The machine itself might
change itself whatever whenever it wanted. Please not the stuff about Garry
talking to his mother, looking TV and eating lobsters again. Because you
once told me that therefore also the machine had to have chats with Joel
Benjamin ... Or with Hsu even better. :)
Robert Hyatt
: Please. I thought I had told you my point with irony. But that seemed to be
: a failure. Therefore here in capitals. You don't need to add more examples
: on the same level. I understood it, mainly agreed with you, but then tried
: to say that this is not the sort of stuff Kasparov needed to exploit
: weaknesses of a machine. All these trivial examples are true but also
: totally wrong. At least in the sense you tried to use them. To make this
: quite clear. For you as a programmer of a specific computerchess program
: these details are useful and important. But in human chess the GMs don't
: need such petitesses. Because if they had to rely on such tricks or if they
: had opponents who could be busted regularly with such trivialities, it
: would be no longer GM-chess.
I totally disagree. Why do you think he wanted to see *recent* games of
the program? For recreation, or because he thought that he could study them
to learn something about how it evaluates/searches?
: Let me repeat. A real GM doesn't need to wait for the moment when the
: "opponent" finally falls into that sort of traps. A GM plays his game no
: matter if it's a weak or a strong opponent. If the pressure finally is
: rising it's normal that more and more of such tactical tricks become
: important. But it would mean nothing to know before the game that opponent
: A was a victim of a double attack in zeitnot. A GM doesn't think "well,
: let's find a good situation for a double attack for this guy." Why? Simply
: because double attacks are the daily bread for each chessplayer, not only a
: GM. At first comes the game. The pressure must be created. So, what I'm
: tryinfg to tell you is the non-importance of such knowledge.
This is also wrong. I can lead you to a couple of GM's that will give you
some *eye-opening* input on this subject. But they *all* consider their
opponent, and the openings their opponents do well with and the openings they
do poorly against, etc. When I was in the midst of getting ready for the
proposed Yermolinsky match (which fell through) I asked Roman for suggestions.
Boy did I get suggestions... he pulled up his trusty database, checked to see
what he liked to play, and gave me some ideas of what I might try to lead him
away from his normal openings...
So I simply disagree... GM's consider the opponent all the time. And they
adjust their game plan to the opponent, whether they need a win or draw,
possible fatigue issues that might lead to mistakes, etc...
If you believe it isn't important, then you have to answer the question "why
did Kasparov want to see games played by DB. Why, after he lost the match,
did he start making demands about the next match, even when one was not on the
drawing board? IE I can quote his remarks, as relayed by Yasser in the ICCA
Journal, where he demanded "N" games played against known opponents before the
next match.
If he couldn't/wouldn't alter his style or strategy based on these games, what
good could they *possibly* do?
: I see another weak point however in your theory. We know that strong GMs
: don't lose to a 7th rank attack because they didn't know the danger of it.
: No, the game finally escalates into a weak 7th for one side. But in today's
: chess this is never a singular pattern but a complexity of several
: patterns. Both players "know" all about the single items. The art is to
: find a complex strategy that goes a little step farther the opponent had
: thought of.
I wasn't suggesting that the 7th rank was even worth discussing. Just that
Kasparov would *never* sit down with an opponent he was playing in a match
next week and discuss his strategy, or give lessons in positional understanding
to that opponent...
: I'm talking about human chess as Rolf is fantasysing it. How do you think
: Kasparov could explain to (whom??) how to beat him theoretically? Let me
: help you. I would take his queen at first and then mate him. Period.
: Go ahead with crafty this way but don't dream that you could ever come to
: an end with such successive steps. With such descriptions you show me that
: you still don't think about real GM chess. I fear you talk too often with
: GMRoman. Of course he could talk for months about such points. He surely
: knows many more than me for example. But I still don't understand what he's
: really doing. Keeping you busy for the next years? Or is he trying to
: explain his thought process? Watch out, Bob. He's a chess addict.
I know that. There are several such "addicts" around. But you seem to think
in terms of "revolution" which is ok. I think in terms of "evolution" which
nature has proven to work quite well, if somewhat slowly. I'll keep patching
the holes I find, and the holes that IM/GM players point out to me, while you
make the revolutionary jump. But somehow I suspect that evolution will win.
It took Deep Blue a long time and a lot of feedback from GM players to reach its
current level. There's plenty left that can be done without throwing out the
current approach that is playing pretty well for lots of us...
: Why don't you start a cooperation of some sort with Chris? He could tell
: you the same about isolated pawns and open files. But he has the links to
: the future world of "knowledge". To give you my little warning: Don't
: believe that you could add all the steps and hope for an automatic birth of
: intelligent chess. It won't function this way.
Of course not. No program of mine has ever beaten a GM player, not in blitz,
not in game/30, not in 40/2hrs. Was all a dream. Other programs based on the
same sort of paradigm, including Rebel, Junior, etc. are not beating IM/GM
players with enough regularity to raise warning flags either. It's totally
hopeless to think we could ever do this, right?
: Yes, I already bought that. But I also said that one could hide the secret
: parts and still leaving the parts open that could prove that all what
: happened was kosher. Because that was the point for the asking. NOT to
: detect some spooky tricks to exploit. Bob, we're talking about real chess
: and not just you and Ed playing 24 h/move games. :)
There never was any doubt about "kosher" behavior. The printout published
in the NY Times was clear enough in its meaning... The "cheating" was just
pure baloney, conjured up by a sore loser with a petty attitude making excuses
for his own stupidity...
: I knew since long that we had something in common ...
:> He knew it was stupid to ask for them. Everyone there knew it
: >was stupid for him to ask for them. Almost everyone *here* could figure
: >out that it was stupid for him to ask for them. Because he
: wouldn't/couldn't
: >return the favor.
:
: >And chess *is* a "zero-sum" game...
: But now, Bob, it looks this way:
: Kasparov, one of thze best players of the world was busted by a team of
: mediocre, not even this, by complete chessic ignorants with the exception
: of some advising GMs. Busted with the help of a machine that is NEVER of
: real GM strength. That is as if Michael Johnson would be beaten by a
: Vietnam veteran in his wheel chair...
I don't know what you mean. It *must* be of GM strength to beat Kasparov,
as he has *never* lost a match to a non-GM in his career. In fact, he hasn't
lost enough matches to count... no matter what the opposition...
: And then you come and try to argue that it's all Garry's own fault because
: he had to know better about the American business. How could he after the
: first meetings were held in all sophisticated friendship? It was as if you
: was friendly enough to let your bathroom-window open for a guest whom you
: didn't want to salute at your front entrance. You can't understand what
: unwritten rules the IBM /DB teamsters have violated? Could you please leave
: aside your buddy-motivation for a moment and to better concentrate on your
: task as scientist? I don't doubt for a second that you're man enough to
: fight for "them" but then you pay a too high price in my eyes. You come in
: danger with your reputation as a scientist. The same I told you with Chris
: and his ideas. I think you have modified your politics a little bit and
: that's great. But as far as DB is concerned you should simply try to let
: live both sides. Both sides have their pro's and con's. It would honour you
: if you would also defend the side you personally don't appreciate at all.
: Get the idea?
Nope. I have *zero* respect for Kasparov after the way he acted. He should
have walked to the stage, said "I screwed up, I listened to bad advice, I
blundered in the final game, I deserved to lose, and I'd like to have a rematch
since we are now 1-1 in match results." He didn't. He acted like a complete
jackass.
: This wouldn't weaken your position but would only shed a brighter light on
: your reputattion as an expert.
:
I don't do this for a "reputation as an expert." You want my opinion, you
get *my* opinion. Nothing more. Nothing less.
: Again this is too low stuff. Too trivial. Too average. Tomorrow you tell us
: that he won't speak about his wish to win the game ...
Again, if he wouldn't use the DB info to adjust his "plan" why would he
want games and output for anything *else*? It would be useless if he didn't
study and adjust...
: I can't bet because I think and wrote the same. But you sound as if you
: would say that Kasparov did ask for the stuff to get an unfair advantage.
: That's ridiculous. That's how *I* try to bust FRITZ but Garry has other
: talents to win against a machine.
Not enough, apparently. Remember, 3.5-2.5... 3.5 is greater than 2.5.
: You generally don't refuse to repeat always the same good arguments. So let
: me repeat mine. It's ridiculous to make a match about the question mankind
: (!) or machine and then behind the curtains transform, better pervert it
: into a mystery show where nobody really knew what was real and what was
: fake. Garry is still a human being, no? Already your assumption of a clean
: "thought-process" one could get from a human player was very superficial. A
: human being has emotions, feelings. An invisible artificial opponent alone
: is
: sufficient to provoke feelings of anxiety. If then -- justified or not --
: the impression grows up that something is wrong with the whole setting,
: especially for someone who already is believing in supernatural events,
: then the asking for the output is not at all a naive nonsense but a last
: try to save the situation. History will speak the verdict about the details
: of the event.
I too would *love* to see DB play in human events. So we can determine how
strong it is. We *know* it can beat Kasparov, something that does not happen
very often. But we don't know how strong it truly is. IE can it totally wax
the under 2600 GM's? No idea.
But we also know it can play chess. As evidenced in NY.
: I have already made my decision. A team of highly intelligent technocrats
: with a machine had a meeting with an artist, a great sportsman, a genius. A
: human being still. With many faults and many good sides. What I have wished
: for a fair show would have been sort of chairman for the machine's team who
: at least had a glimpse of insight into the world of a chess genius. Someone
: from Russia. From Aserbeidjan. The first ridiculous factor is when a man
: called "The Operator" is placed in front of the genius, as if *he* would
: play the game against him. Sure the operator has the monitor at his side.
: But then in ther meantime he also stares at the board as if he was trying
: to think ... This must be changed. It's troubling the basic concentration
: of a player. Because normaly he has his real opponent in front. Here he has
: to abstract from the operator and the display to the real machine some
: kilometers away. That has to be changed. The machine has to be in the same
: room where the board and the audience are. The operator has not the right
: to be seated at the board but he should have his place in front of the
: monitor. He then stands up makes a move on the board and goes away again.
: Even better would be a monitor solution next to the board and a neutral
: referee would move for the machine. There must not be an operator from the
: team itself. This way the situation becomes clearer a fight man vs machine
: with the teamster totally in the background.
That's all interesting, and *none* of it represents something I haven't heard
before. I'll say it again: playing against a computer is *far* different than
playing against a human. No psych-games possible. No fatigue. no momentary
lapse in concentration. No simple tactical oversights or blunders. No fairly
deep tactical oversights. It is like pushing against a *very* heavy truck.
You might budge it, but you have to push as hard as you possibly can, *every*
step of the way, never getting a moment to take a deep breath. No slamming a
piece down and momentarily stunning your opponent. It is a different game.
But then the computer is based on silicon and not carbon, electrical signals
vs electro-chemical signals, etc. they are *different*. what else is new?
: I described all that not because all that had to be processed like that.
: But I wanted to demonstrate a sample of the hundreds of factors that could
: be important for a human player and his perception. Focussing on the human
: thought process is a good research project but not a sufficient description
: of the complete story. This is difficult to explain to someone who never
: was himself in such an extreme psychological tension of a chess genius.
:
No argument. But Kasparov knew what he was playing before he made his first
move. He thought that the prize fund justified the risk.
: Yes, because a chess genius isn't automatically a psychological scientist
: of his own self.
:
:
Nope...
: Ah, you're so easily to be instructed as agent provocateur? :)
No... just "been there, seen that, done that..."
: I grant you that right, Bob.
: But then I reserve me the right to tell you that you absolutely missed my
: point.
: And I must state this with all seriousity. We're not in Las Vegas. We're
: not in gambling. And if we play against a machine we want a high
: probability that it was the machine who made the moves. Was that clear
: enough? We as scientists should support such a triviality. And to make
: another thing clear. Because you always repeat that Kasparov had all the
: cards to define whatever he wanted ...
Please stop that line of reasoning (that it was the machine that made the
moves.) DB played the moves it played. Output for the highly debated
Be4 vs Qb6 was one example. Other programs also found this although it
took us *tons* of time, not 15 minutes like DB. So that line of reasoning
is simply "off limits." No cheating. No discussion abut possible cheating.
It didn't happen. And only a very few with no computer chess savvy have
suggested such.
:
:
: But you made a weak statement. Because I don't care, and Kasparov doesn't
: care about what (now please pay attention) the *m a c h i n e* is
: doing by itself. That's the deal in my eyes. But the machine has to be set
: into the playing hall and all the operators, allegators and master(ator)s
: have to be placed in the audience like Garry's mother. Far away from the
: machine. It's a special misbehaviour that the teamsters tried to pretend
: that their cryptic output of the machine fell completely into the
: competence of the team itself. What would you say if (I repeat an older
: idea) an athlet would claim the right to run 200 miles away and then
: transmitted by a TV satellite ...? The moves of a chessplaying machine must
: come visible and understandable for a really neutral referee.
Total, utter, nonsense. I'm not going to debate the cheating claim. It
was totally disproved on Be4 vs Qb6. It is *not* something to debate any
further...
: And finally all the possible tweaking and personality twisting has to be
: done before the match. I mean the human input. The machine itself might
: change itself whatever whenever it wanted. Please not the stuff about Garry
: talking to his mother, looking TV and eating lobsters again. Because you
: once told me that therefore also the machine had to have chats with Joel
: Benjamin ... Or with Hsu even better. :)
I disagree. Unless you seal the computer in one room, and Kasparov in
another. No phone. No visitors. No *nothing*. I'd take such a match
opportunity, *any* day...
Rolf Tueschen
>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
>: Please. I thought I had told you my point with irony. But that seemed to be
>: a failure. Therefore here in capitals. You don't need to add more examples
>: on the same level. I understood it, mainly agreed with you, but then tried
>: to say that this is not the sort of stuff Kasparov needed to exploit
>: weaknesses of a machine. All these trivial examples are true but also
>: totally wrong. At least in the sense you tried to use them. To make this
>: quite clear. For you as a programmer of a specific computerchess program
>: these details are useful and important. But in human chess the GMs don't
>: need such petitesses. Because if they had to rely on such tricks or if they
>: had opponents who could be busted regularly with such trivialities, it
>: would be no longer GM-chess.
>I totally disagree. Why do you think he wanted to see *recent* games of
>the program? For recreation, or because he thought that he could study them
>to learn something about how it evaluates/searches?
We have a new misunderstanding here. We were talking about the output. My
view was that we needed it because we wanted to be sure about the identity
of the output and DB thoughts. It was not to win insight into the thought
process as such.
Then the question of some games of your next opponent. We totally agree
that in special Kasparov could use such data in his favor. No doubt about
it. But then I went further and tried to explain that Kasparov didn't
*need* the like data to make his plans. It's sufficient to see of what
quality more or less the games of any opponent would be. If NOT of highest
GM quality, he can go to sleep. If however of top GM quality he surely -- I
agree with you 100% again -- will go into details like all the top GMs down
to expert A class level have done it since centuries ago. But my argument
went about the amount of necessity, Kasparov needed such data. And I told
you that guys of his kind would play good chess no matter of the
opposition. Let me explain. Garry would just try the Vienna Game with the
aggressive 5.d3. Simply because it's not prospective enough for White. But
Rolf e.g. did play it all the time because if you don't know too exact with
the Blacks what to do you will get buried head-down into the dust. Or take
the Blackmar/Diemer Gambit. You won't see Kasparov play it even in simuls.
And NOT because he feared any loss. But it's simply the reputation as Wch
that prevents him from playing such lines. Because he knows that it's not
the best chess line. I don't know what you understand here, but writing
that I did NOT mean that nobody should play the BDG. As an amateur you
could play the line with no disadvantages. The same with my 5.d3. Also if
you want to test the opening understanding of your opponent you might play
1.e3 e5 2. e4, just in case the guy never liked to open his games with e4
himself. Now he has to play it with Black but the "advantage" of the first
move. Against FRITZ I played hundreds of games with the black pieces simply
because I could better learn about the strength of it.
But this is not top GM chess. If you want to join a nice discussion, come
look into rgcm. There I defend the opinion that J. Polgar is NOT a creative
striong top GM. An artist so to speak. But all the male top GM are *also*
artists who gave us wonderful examples of their creativity. Kasparov in the
first ranks of course. The beauty comes out of the impression what the guys
are able to do using all the well known openings we amateurs could discover
in the ECO. No secret sidelines. No, the main line down into the 25 moves.
And then new ideas. We amateurs simply can't compete with this. Because the
amount of data becomes confusing after a while.
That's the "cheat/fake" with the actual programs. They can't walk without
sticks, but the can dance along the ECO lines better than the Wch.
You as a father of such a monster must know, and you are one great
exception who also talked about that, that it's more or less hot air. Of
course a player like Garry Kasparov could discover hundreds of fallacies in
certain opening lines, he has created himself, BTW. So, if you want to hide
such secrets away from him long enough (therefore Ed played only two
tournament games where one draw already saved the honour gold medal), you
surely could beat even the human Wch. No doubt. Garry will also lose many
exhibition games either because he was too lazy to take a certain trick
serious enough or because he felt that a decent play should be honoured
with his support ...
That is a complete misunderstanding when you meant that (raising your
voice) 3,5 was more than 2,5. Yes, of course. But it doesn't mean so much
statistically. Now guess, who was that guy who doesn't give a rest to
explain such a triviality here on rgcc on and on?
The tragedy of DB was for me that it didn't prove anything. Anything what a
decent micro could show us too. Sure, it had the moves in shorter time, but
the main points (the principle weaknesses of silicon valley) were still
there. All savys know that. And you too. :)
>: Let me repeat. A real GM doesn't need to wait for the moment when the
>: "opponent" finally falls into that sort of traps. A GM plays his game no
>: matter if it's a weak or a strong opponent. If the pressure finally is
>: rising it's normal that more and more of such tactical tricks become
>: important. But it would mean nothing to know before the game that opponent
>: A was a victim of a double attack in zeitnot. A GM doesn't think "well,
>: let's find a good situation for a double attack for this guy." Why? Simply
>: because double attacks are the daily bread for each chessplayer, not only a
>: GM. At first comes the game. The pressure must be created. So, what I'm
>: tryinfg to tell you is the non-importance of such knowledge.
>This is also wrong. I can lead you to a couple of GM's that will give you
>some *eye-opening* input on this subject. But they *all* consider their
>opponent, and the openings their opponents do well with and the openings they
>do poorly against, etc. When I was in the midst of getting ready for the
>proposed Yermolinsky match (which fell through) I asked Roman for suggestions.
>Boy did I get suggestions... he pulled up his trusty database, checked to see
>what he liked to play, and gave me some ideas of what I might try to lead him
>away from his normal openings...
>So I simply disagree... GM's consider the opponent all the time. And they
>adjust their game plan to the opponent, whether they need a win or draw,
>possible fatigue issues that might lead to mistakes, etc...
You bet. me too. We don't differ here. Thanks for the little insight into
your kitchen. It's sad that I could never enjoy it live on ICC ...
>If you believe it isn't important, then you have to answer the question "why
>did Kasparov want to see games played by DB. Why, after he lost the match,
>did he start making demands about the next match, even when one was not on the
>drawing board? IE I can quote his remarks, as relayed by Yasser in the ICCA
>Journal, where he demanded "N" games played against known opponents before the
>next match.
Easy one. If the quality of games were there, then he knew what to play ...
For me the "N" is another form of evidence like the "output". Of course K.
needs such information. You don't understand GM chess. It's naturally about
the opponent's qualities. Each simul exhibition could justify a new
creation of the "real" Wch, just because some little boy had beaten the
champ. But in all sports it's about a longer series of qualifications,
right?
"They" acted unfriendly against Garry and what he's doing, he simply
reminds the world of certain undeniable truths. Which he had formerly kept
away from public eyes out of politeness/generosity. Bob, you mix it all up.
First came the unfriedliness then Garry's reaction.
>If he couldn't/wouldn't alter his style or strategy based on these games, what
>good could they *possibly* do?
Easy one. At first prove the real capacity of the "thing".
>: I see another weak point however in your theory. We know that strong GMs
>: don't lose to a 7th rank attack because they didn't know the danger of it.
>: No, the game finally escalates into a weak 7th for one side. But in today's
>: chess this is never a singular pattern but a complexity of several
>: patterns. Both players "know" all about the single items. The art is to
>: find a complex strategy that goes a little step farther the opponent had
>: thought of.
>I wasn't suggesting that the 7th rank was even worth discussing. Just that
>Kasparov would *never* sit down with an opponent he was playing in a match
>next week and discuss his strategy, or give lessons in positional understanding
>to that opponent...
It never happened, but not because he then could no longer win a match,
Bob!
He's a professional. He's not paid by the State of Alabama ...
Give him a million dollars and he will tell you more than GMRoman.
>: I'm talking about human chess as Rolf is fantasysing it. How do you think
>: Kasparov could explain to (whom??) how to beat him theoretically? Let me
>: help you. I would take his queen at first and then mate him. Period.
>: Go ahead with crafty this way but don't dream that you could ever come to
>: an end with such successive steps. With such descriptions you show me that
>: you still don't think about real GM chess. I fear you talk too often with
>: GMRoman. Of course he could talk for months about such points. He surely
>: knows many more than me for example. But I still don't understand what he's
>: really doing. Keeping you busy for the next years? Or is he trying to
>: explain his thought process? Watch out, Bob. He's a chess addict.
>I know that. There are several such "addicts" around. But you seem to think
>in terms of "revolution" which is ok. I think in terms of "evolution" which
>nature has proven to work quite well, if somewhat slowly. I'll keep patching
>the holes I find, and the holes that IM/GM players point out to me, while you
>make the revolutionary jump. But somehow I suspect that evolution will win.
>It took Deep Blue a long time and a lot of feedback from GM players to reach its
>current level. There's plenty left that can be done without throwing out the
>current approach that is playing pretty well for lots of us...
1. You never thought of your life span?
2. You never worried the fire break in the silicon tissue?
3. I'm not saying that your approach was nuts. It's ok. I like it. Most of
your findings will go into commercial programs of the next season. Not bad.
I was referring to your deeply hidden core of a possible future nobel prize
winner ... But I might have exegerated.
>: Why don't you start a cooperation of some sort with Chris? He could tell
>: you the same about isolated pawns and open files. But he has the links to
>: the future world of "knowledge". To give you my little warning: Don't
>: believe that you could add all the steps and hope for an automatic birth of
>: intelligent chess. It won't function this way.
>Of course not. No program of mine has ever beaten a GM player, not in blitz,
>not in game/30, not in 40/2hrs. Was all a dream. Other programs based on the
>same sort of paradigm, including Rebel, Junior, etc. are not beating IM/GM
>players with enough regularity to raise warning flags either. It's totally
>hopeless to think we could ever do this, right?
Let's pray. Lord, I shallt refrain my eager fast flying temptations.
Short-cuts. Addictions of heave dosed fallacies. Amen.
>: Yes, I already bought that. But I also said that one could hide the secret
>: parts and still leaving the parts open that could prove that all what
>: happened was kosher. Because that was the point for the asking. NOT to
>: detect some spooky tricks to exploit. Bob, we're talking about real chess
>: and not just you and Ed playing 24 h/move games. :)
>There never was any doubt about "kosher" behavior. The printout published
>in the NY Times was clear enough in its meaning... The "cheating" was just
>pure baloney, conjured up by a sore loser with a petty attitude making excuses
>for his own stupidity...
Let's adjourn that one, please. We'll come back later ...
>: I knew since long that we had something in common ...
>:> He knew it was stupid to ask for them. Everyone there knew it
>: >was stupid for him to ask for them. Almost everyone *here* could figure
>: >out that it was stupid for him to ask for them. Because he
>: wouldn't/couldn't
>: >return the favor.
>:
>: >And chess *is* a "zero-sum" game...
>: But now, Bob, it looks this way:
>: Kasparov, one of thze best players of the world was busted by a team of
>: mediocre, not even this, by complete chessic ignorants with the exception
>: of some advising GMs. Busted with the help of a machine that is NEVER of
>: real GM strength. That is as if Michael Johnson would be beaten by a
>: Vietnam veteran in his wheel chair...
>I don't know what you mean. It *must* be of GM strength to beat Kasparov,
>as he has *never* lost a match to a non-GM in his career. In fact, he hasn't
>lost enough matches to count... no matter what the opposition...
Usually I'm prepared to take a look at the quality of a game before I make
my comments.
>: And then you come and try to argue that it's all Garry's own fault because
>: he had to know better about the American business. How could he after the
>: first meetings were held in all sophisticated friendship? It was as if you
>: was friendly enough to let your bathroom-window open for a guest whom you
>: didn't want to salute at your front entrance. You can't understand what
>: unwritten rules the IBM /DB teamsters have violated? Could you please leave
>: aside your buddy-motivation for a moment and to better concentrate on your
>: task as scientist? I don't doubt for a second that you're man enough to
>: fight for "them" but then you pay a too high price in my eyes. You come in
>: danger with your reputation as a scientist. The same I told you with Chris
>: and his ideas. I think you have modified your politics a little bit and
>: that's great. But as far as DB is concerned you should simply try to let
>: live both sides. Both sides have their pro's and con's. It would honour you
>: if you would also defend the side you personally don't appreciate at all.
>: Get the idea?
>Nope. I have *zero* respect for Kasparov after the way he acted. He should
>have walked to the stage, said "I screwed up, I listened to bad advice, I
>blundered in the final game, I deserved to lose, and I'd like to have a rematch
>since we are now 1-1 in match results." He didn't. He acted like a complete
>jackass.
Only fair to see it this way. I'm more on his side as a psychologist.
Excuse me.
>: This wouldn't weaken your position but would only shed a brighter light on
>: your reputattion as an expert.
>:
>I don't do this for a "reputation as an expert." You want my opinion, you
>get *my* opinion. Nothing more. Nothing less.
I thank you for that opinion. But all I did was appealing to your heart and
super conscience. Why not try the evolution in our own self?
>: Again this is too low stuff. Too trivial. Too average. Tomorrow you tell us
>: that he won't speak about his wish to win the game ...
>Again, if he wouldn't use the DB info to adjust his "plan" why would he
>want games and output for anything *else*? It would be useless if he didn't
>study and adjust...
Why do referees exist in this world of flowers and peace? Ed? :)
>: I can't bet because I think and wrote the same. But you sound as if you
>: would say that Kasparov did ask for the stuff to get an unfair advantage.
>: That's ridiculous. That's how *I* try to bust FRITZ but Garry has other
>: talents to win against a machine.
>Not enough, apparently. Remember, 3.5-2.5... 3.5 is greater than 2.5.
>: You generally don't refuse to repeat always the same good arguments. So let
>: me repeat mine. It's ridiculous to make a match about the question mankind
>: (!) or machine and then behind the curtains transform, better pervert it
>: into a mystery show where nobody really knew what was real and what was
>: fake. Garry is still a human being, no? Already your assumption of a clean
>: "thought-process" one could get from a human player was very superficial. A
>: human being has emotions, feelings. An invisible artificial opponent alone
>: is
>: sufficient to provoke feelings of anxiety. If then -- justified or not --
>: the impression grows up that something is wrong with the whole setting,
>: especially for someone who already is believing in supernatural events,
>: then the asking for the output is not at all a naive nonsense but a last
>: try to save the situation. History will speak the verdict about the details
>: of the event.
>I too would *love* to see DB play in human events. So we can determine how
>strong it is. We *know* it can beat Kasparov, something that does not happen
>very often. But we don't know how strong it truly is. IE can it totally wax
>the under 2600 GM's? No idea.
Ok then. We agree.
>But we also know it can play chess. As evidenced in NY.
100% ok.
Thanks. I was trying a first approach as "newbie" in computerchess. Thanks
for your acceptancy.
>: I described all that not because all that had to be processed like that.
>: But I wanted to demonstrate a sample of the hundreds of factors that could
>: be important for a human player and his perception. Focussing on the human
>: thought process is a good research project but not a sufficient description
>: of the complete story. This is difficult to explain to someone who never
>: was himself in such an extreme psychological tension of a chess genius.
>:
>No argument. But Kasparov knew what he was playing before he made his first
>move. He thought that the prize fund justified the risk.
He knew the "machine" factor but he didn't know too good the rest we talked
about.
>: Yes, because a chess genius isn't automatically a psychological scientist
>: of his own self.
>:
>:
>Nope...
>: Ah, you're so easily to be instructed as agent provocateur? :)
>No... just "been there, seen that, done that..."
I know. Let's make a joke from time to time to keep us in shape.
>: I grant you that right, Bob.
>: But then I reserve me the right to tell you that you absolutely missed my
>: point.
>: And I must state this with all seriousity. We're not in Las Vegas. We're
>: not in gambling. And if we play against a machine we want a high
>: probability that it was the machine who made the moves. Was that clear
>: enough? We as scientists should support such a triviality. And to make
>: another thing clear. Because you always repeat that Kasparov had all the
>: cards to define whatever he wanted ...
>Please stop that line of reasoning (that it was the machine that made the
>moves.) DB played the moves it played. Output for the highly debated
>Be4 vs Qb6 was one example. Other programs also found this although it
>took us *tons* of time, not 15 minutes like DB. So that line of reasoning
>is simply "off limits." No cheating. No discussion abut possible cheating.
>It didn't happen. And only a very few with no computer chess savvy have
>suggested such.
Adjournement. Accepted?
>:
>:
>: But you made a weak statement. Because I don't care, and Kasparov doesn't
>: care about what (now please pay attention) the *m a c h i n e* is
>: doing by itself. That's the deal in my eyes. But the machine has to be set
>: into the playing hall and all the operators, allegators and master(ator)s
>: have to be placed in the audience like Garry's mother. Far away from the
>: machine. It's a special misbehaviour that the teamsters tried to pretend
>: that their cryptic output of the machine fell completely into the
>: competence of the team itself. What would you say if (I repeat an older
>: idea) an athlet would claim the right to run 200 miles away and then
>: transmitted by a TV satellite ...? The moves of a chessplaying machine must
>: come visible and understandable for a really neutral referee.
>Total, utter, nonsense. I'm not going to debate the cheating claim. It
>was totally disproved on Be4 vs Qb6. It is *not* something to debate any
>further...
What about the Kh1 or Kf1 topic?
>: And finally all the possible tweaking and personality twisting has to be
>: done before the match. I mean the human input. The machine itself might
>: change itself whatever whenever it wanted. Please not the stuff about Garry
>: talking to his mother, looking TV and eating lobsters again. Because you
>: once told me that therefore also the machine had to have chats with Joel
>: Benjamin ... Or with Hsu even better. :)
>I disagree. Unless you seal the computer in one room, and Kasparov in
>another. No phone. No visitors. No *nothing*. I'd take such a match
>opportunity, *any* day...
Frankenstein!
Robert Hyatt
: Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
:>Rolf Tueschen <TUESCHEN.MEDIZ...@t-online.de> wrote:
:>: Please. I thought I had told you my point with irony. But that seemed to be
:>: a failure. Therefore here in capitals. You don't need to add more examples
:>: on the same level. I understood it, mainly agreed with you, but then tried
:>: to say that this is not the sort of stuff Kasparov needed to exploit
:>: weaknesses of a machine. All these trivial examples are true but also
:>: totally wrong. At least in the sense you tried to use them. To make this
:>: quite clear. For you as a programmer of a specific computerchess program
:>: these details are useful and important. But in human chess the GMs don't
:>: need such petitesses. Because if they had to rely on such tricks or if they
:>: had opponents who could be busted regularly with such trivialities, it
:>: would be no longer GM-chess.
:>I totally disagree. Why do you think he wanted to see *recent* games of
:>the program? For recreation, or because he thought that he could study them
:>to learn something about how it evaluates/searches?
: We have a new misunderstanding here. We were talking about the output. My
: view was that we needed it because we wanted to be sure about the identity
: of the output and DB thoughts. It was not to win insight into the thought
: process as such.
There were two issues he brought up during and after the match, particularly
after the match:
(1) he couldn't believe that a program would play Be4 rather than winning
a pawn (he thought) with Qb6. Several of us proved that Qb6 leads to a
draw, although it takes a hellacious search (for a micro) to find this.
They apparently found it OTB. And switched to Be4 which does not lead to
the same draw. He wanted a printout of this move, which he eventually
received, and same was published in the NY times, maybe even here in this
newsgroup. But in the Times for sure. It seemed obvious to those of us
looking at it (except for maybe one person, perhaps) that it was a legitimate
move by the computer, and our later analysis using other computers proved that
there was a valid reason for avoiding Qb6. That's a case closed issue...
(2) he said that for the next match he wanted game scores between DB and
known strong opponents. This was the subject that I thought interesting,
because if, as you claim, he wouldn't study DB's play to adjust his style,
then *what* would he do with the game scores? I think he'd do just like he'd
do if he was playing Kramnik, Short, Anand, Karpov, Shirov, etc... he'd look
for favored openings and try to find a way to obtain an advantage. But against
a computer I think that he'd go further and look carefully for apparent
positional misunderstanding, and try to exploit that as well. He has *said*
this on several occasions... that if he could study program output and/or
program games, that he would find weaknesses and discover how to beat the
machine. He did this OTB in match 1, but was unable to do so in match 2
because it behaved somewhat differently between games, to keep him off-
balance...
: Then the question of some games of your next opponent. We totally agree
: that in special Kasparov could use such data in his favor. No doubt about
: it. But then I went further and tried to explain that Kasparov didn't
: *need* the like data to make his plans. It's sufficient to see of what
: quality more or less the games of any opponent would be. If NOT of highest
: GM quality, he can go to sleep. If however of top GM quality he surely -- I
: agree with you 100% again -- will go into details like all the top GMs down
: to expert A class level have done it since centuries ago. But my argument
: went about the amount of necessity, Kasparov needed such data. And I told
: you that guys of his kind would play good chess no matter of the
: opposition. Let me explain. Garry would just try the Vienna Game with the
: aggressive 5.d3. Simply because it's not prospective enough for White. But
: Rolf e.g. did play it all the time because if you don't know too exact with
: the Blacks what to do you will get buried head-down into the dust. Or take
: the Blackmar/Diemer Gambit. You won't see Kasparov play it even in simuls.
: And NOT because he feared any loss. But it's simply the reputation as Wch
: that prevents him from playing such lines. Because he knows that it's not
: the best chess line. I don't know what you understand here, but writing
: that I did NOT mean that nobody should play the BDG. As an amateur you
: could play the line with no disadvantages. The same with my 5.d3. Also if
: you want to test the opening understanding of your opponent you might play
: 1.e3 e5 2. e4, just in case the guy never liked to open his games with e4
: himself. Now he has to play it with Black but the "advantage" of the first
: move. Against FRITZ I played hundreds of games with the black pieces simply
: because I could better learn about the strength of it.
I don't disagree with you here. It depends on what you are trying to
achieve with Deep Blue. If *the goal* is to beat Kasparov, then secrecy is
the *obvious* way to go. Don't give him a clue about your weaknesses, or
about your opening choices, etc...
If the goal is to prove you have the program that is the best player in the
world, then it has to be proved in a different way, such as a year's competition
in GM tournaments. IE like many of us do on the chess servers. Say "come and
get us if you can" and then try to survive as they play hundreds of games,
probing for weaknesses...
I believe that they were quite clear, that their goal was to beat Kasparov in
a match. Nothing else. That was the Fredkin prize requirement. It has been
a computer chess goal since I wrote my first program (for me) and for others
it dates back even further.
Beaing the best on the planet is a valid goal. It is *another* goal beyond the
goal of beating Kasparov. I'd like to see them pursue that to see if they are
good enough to reach it. If they don't, there are plenty of us that will keep
working and try to do so.
But for the goal they set, their approach was the correct one... because
they wanted to beat Kasparov in a match (actually, beat the world champion
in a match, but since that was/is Kasparov...).
: But this is not top GM chess. If you want to join a nice discussion, come
: look into rgcm. There I defend the opinion that J. Polgar is NOT a creative
: striong top GM. An artist so to speak. But all the male top GM are *also*
: artists who gave us wonderful examples of their creativity. Kasparov in the
: first ranks of course. The beauty comes out of the impression what the guys
: are able to do using all the well known openings we amateurs could discover
: in the ECO. No secret sidelines. No, the main line down into the 25 moves.
: And then new ideas. We amateurs simply can't compete with this. Because the
: amount of data becomes confusing after a while.
: That's the "cheat/fake" with the actual programs. They can't walk without
: sticks, but the can dance along the ECO lines better than the Wch.
: You as a father of such a monster must know, and you are one great
: exception who also talked about that, that it's more or less hot air. Of
: course a player like Garry Kasparov could discover hundreds of fallacies in
: certain opening lines, he has created himself, BTW. So, if you want to hide
: such secrets away from him long enough (therefore Ed played only two
: tournament games where one draw already saved the honour gold medal), you
: surely could beat even the human Wch. No doubt. Garry will also lose many
: exhibition games either because he was too lazy to take a certain trick
: serious enough or because he felt that a decent play should be honoured
: with his support ...
Note that I consider "micro programs to be less than GM's today" but I'm no
longer convinced that will be true in 5 more years. They are beginning to
become serious GM opponents. Which likely means that (a) GM players are not
quite as infallible as we'd like to think; (b) chess might not be quite as
difficult as many of us have thought for a very long time; (c) search may
be able to make up for some sorely missing knowledge that the humans possess.
I don't know that any of that is true. But anecdotal evidence suggests that
it might be, because of results that I see every day. I can point you to an
IM running a chess program on a server, something I thought would *never*
happen. Why? In his words, he loves to watch it play, and defend inferior
middlegame and endgame positions as fiercely as Korchnoi would do. IE he is
seeing something in a computer's play that many of us would not attribute to
a computer at all. I won't mention the program unless you are interested. It
might be surprising..
: That is a complete misunderstanding when you meant that (raising your
: voice) 3,5 was more than 2,5. Yes, of course. But it doesn't mean so much
: statistically. Now guess, who was that guy who doesn't give a rest to
: explain such a triviality here on rgcc on and on?
What I meant was that Kasparov lost. Not that he was the worse player. Not
that DB was clearly the better player. But that in *that* match, Kasparov
lost. Something thought impossible before it happened... Something that had
*never* happened in the past 40+ years of computer chess... and something that
may not happen again if DB stays inactive... at least for many years.
: The tragedy of DB was for me that it didn't prove anything. Anything what a
: decent micro could show us too. Sure, it had the moves in shorter time, but
: the main points (the principle weaknesses of silicon valley) were still
: there. All savys know that. And you too. :)
Yep... never said otherwise. Just their search is so much better than "our"
search (our = micro programs) that they don't overlook a lot. So they have
far fewer weaknesses than we do. And, based on watching the last match, these
weaknesses that they have are *very* difficult to exploit, because it takes
tremendous concentration to avoid making a tactical mistake that you *know*
the machine won't overlook. That mental pressure may well offset the rather
notable positional mistakes it made here and there...
:>: Let me repeat. A real GM doesn't need to wait for the moment when the
:>: "opponent" finally falls into that sort of traps. A GM plays his game no
:>: matter if it's a weak or a strong opponent. If the pressure finally is
:>: rising it's normal that more and more of such tactical tricks become
:>: important. But it would mean nothing to know before the game that opponent
:>: A was a victim of a double attack in zeitnot. A GM doesn't think "well,
:>: let's find a good situation for a double attack for this guy." Why? Simply
:>: because double attacks are the daily bread for each chessplayer, not only a
:>: GM. At first comes the game. The pressure must be created. So, what I'm
:>: tryinfg to tell you is the non-importance of such knowledge.
:>This is also wrong. I can lead you to a couple of GM's that will give you
:>some *eye-opening* input on this subject. But they *all* consider their
:>opponent, and the openings their opponents do well with and the openings they
:>do poorly against, etc. When I was in the midst of getting ready for the
:>proposed Yermolinsky match (which fell through) I asked Roman for suggestions.
:>Boy did I get suggestions... he pulled up his trusty database, checked to see
:>what he liked to play, and gave me some ideas of what I might try to lead him
:>away from his normal openings...
:>So I simply disagree... GM's consider the opponent all the time. And they
:>adjust their game plan to the opponent, whether they need a win or draw,
:>possible fatigue issues that might lead to mistakes, etc...
: You bet. me too. We don't differ here. Thanks for the little insight into
: your kitchen. It's sad that I could never enjoy it live on ICC ...
:)
ICC is easy to visit. Free to log into in fact... as a guest. You can
still chat, discuss, watch games, etc...
:>If you believe it isn't important, then you have to answer the question "why
:>did Kasparov want to see games played by DB. Why, after he lost the match,
:>did he start making demands about the next match, even when one was not on the
:>drawing board? IE I can quote his remarks, as relayed by Yasser in the ICCA
:>Journal, where he demanded "N" games played against known opponents before the
:>next match.
: Easy one. If the quality of games were there, then he knew what to play ...
: For me the "N" is another form of evidence like the "output". Of course K.
: needs such information. You don't understand GM chess. It's naturally about
: the opponent's qualities. Each simul exhibition could justify a new
: creation of the "real" Wch, just because some little boy had beaten the
: champ. But in all sports it's about a longer series of qualifications,
: right?
: "They" acted unfriendly against Garry and what he's doing, he simply
: reminds the world of certain undeniable truths. Which he had formerly kept
: away from public eyes out of politeness/generosity. Bob, you mix it all up.
: First came the unfriedliness then Garry's reaction.
Here's my take. When I compete with someone, things "change". My best friend
and I used to have a lot of fun doing several things, from playing tennis, to
playing chess, to competing in karate matches. But when we competed, whether
it be in tennis, or karate, we *competed*. I did everything I could to beat him,
he did the same to me, and we'd shake hands after it was over and vow to do
better the next time around.
Kasparov has to realize that this was a *competition*. With a lot at
stake. And he was not so stupid as to think that IBM would hand him everything
he wanted to help him beat the machine. Rather, they were going to do everything
possible to beat him. If he expected anything else, he is more stupid than I
though.
:>If he couldn't/wouldn't alter his style or strategy based on these games, what
:>good could they *possibly* do?
: Easy one. At first prove the real capacity of the "thing".
He knew the "real capacity". He had played its predecessor already, and lost
a game to it. So there was no doubt that it was strong and serious...
:>: I see another weak point however in your theory. We know that strong GMs
:>: don't lose to a 7th rank attack because they didn't know the danger of it.
:>: No, the game finally escalates into a weak 7th for one side. But in today's
:>: chess this is never a singular pattern but a complexity of several
:>: patterns. Both players "know" all about the single items. The art is to
:>: find a complex strategy that goes a little step farther the opponent had
:>: thought of.
:>I wasn't suggesting that the 7th rank was even worth discussing. Just that
:>Kasparov would *never* sit down with an opponent he was playing in a match
:>next week and discuss his strategy, or give lessons in positional understanding
:>to that opponent...
: It never happened, but not because he then could no longer win a match,
: Bob!
: He's a professional. He's not paid by the State of Alabama ...
: Give him a million dollars and he will tell you more than GMRoman.
If he'd pay IBM a million bucks, I'd bet he could buy those printouts
*right now*.
:>: I'm talking about human chess as Rolf is fantasysing it. How do you think
:>: Kasparov could explain to (whom??) how to beat him theoretically? Let me
:>: help you. I would take his queen at first and then mate him. Period.
:>: Go ahead with crafty this way but don't dream that you could ever come to
:>: an end with such successive steps. With such descriptions you show me that
:>: you still don't think about real GM chess. I fear you talk too often with
:>: GMRoman. Of course he could talk for months about such points. He surely
:>: knows many more than me for example. But I still don't understand what he's
:>: really doing. Keeping you busy for the next years? Or is he trying to
:>: explain his thought process? Watch out, Bob. He's a chess addict.
:>I know that. There are several such "addicts" around. But you seem to think
:>in terms of "revolution" which is ok. I think in terms of "evolution" which
:>nature has proven to work quite well, if somewhat slowly. I'll keep patching
:>the holes I find, and the holes that IM/GM players point out to me, while you
:>make the revolutionary jump. But somehow I suspect that evolution will win.
:>It took Deep Blue a long time and a lot of feedback from GM players to reach its
:>current level. There's plenty left that can be done without throwing out the
:>current approach that is playing pretty well for lots of us...
: 1. You never thought of your life span?
Nope. Life's too short, as it is, without dwelling on something that I
can't avoid. Whether Crafty can beat Kasparov (or the reigning world champ
at the time) before I die or not is someting I can't control, other than to
make as much progress as I can, and hope for the best. And if I don't do it,
someone else will, whether it be a successor of Crafty, or a successor of
Rebel, or whatever...
: 2. You never worried the fire break in the silicon tissue?
: 3. I'm not saying that your approach was nuts. It's ok. I like it. Most of
: your findings will go into commercial programs of the next season. Not bad.
: I was referring to your deeply hidden core of a possible future nobel prize
: winner ... But I might have exegerated.
Very little chance of that... and it's not an aspiration of mine anyway. I
am merely "having fun" now...
:>: Why don't you start a cooperation of some sort with Chris? He could tell
:>: you the same about isolated pawns and open files. But he has the links to
:>: the future world of "knowledge". To give you my little warning: Don't
:>: believe that you could add all the steps and hope for an automatic birth of
:>: intelligent chess. It won't function this way.
:>Of course not. No program of mine has ever beaten a GM player, not in blitz,
:>not in game/30, not in 40/2hrs. Was all a dream. Other programs based on the
:>same sort of paradigm, including Rebel, Junior, etc. are not beating IM/GM
:>players with enough regularity to raise warning flags either. It's totally
:>hopeless to think we could ever do this, right?
: Let's pray. Lord, I shallt refrain my eager fast flying temptations.
: Short-cuts. Addictions of heave dosed fallacies. Amen.
eh?
:>: Yes, I already bought that. But I also said that one could hide the secret
To me that would be hopeless. I'm not able to look at a game and say "this
must be a GM or that must be an IM". I can recognize "strong" and I can
probably recognize any example of "computer". But if I see someone play a
GM, and beat him 3.5 to 2.5, I am pretty certain, based on sound statistical
theory, that this guy is probably a GM himself. We can compute some
probabilities if you'd like, to see how accurate that might be...
:>: And then you come and try to argue that it's all Garry's own fault because
: Ok then. We agree.
: 100% ok.
:>Nope...
: Adjournement. Accepted?
Fine by me...
:>:
:>:
:>: But you made a weak statement. Because I don't care, and Kasparov doesn't
:>: care about what (now please pay attention) the *m a c h i n e* is
:>: doing by itself. That's the deal in my eyes. But the machine has to be set
:>: into the playing hall and all the operators, allegators and master(ator)s
:>: have to be placed in the audience like Garry's mother. Far away from the
:>: machine. It's a special misbehaviour that the teamsters tried to pretend
:>: that their cryptic output of the machine fell completely into the
:>: competence of the team itself. What would you say if (I repeat an older
:>: idea) an athlet would claim the right to run 200 miles away and then
:>: transmitted by a TV satellite ...? The moves of a chessplaying machine must
:>: come visible and understandable for a really neutral referee.
:>Total, utter, nonsense. I'm not going to debate the cheating claim. It
:>was totally disproved on Be4 vs Qb6. It is *not* something to debate any
:>further...
: What about the Kh1 or Kf1 topic?
Not an issue either. Ed analyzed this with Rebel. Kasparov's comment was
so far out of line it is not even worth discussion. IE he said "a repetition
of 8-10 plies." This draw by repetition is almost exactly *60* plies deep
from this point. Proven by careful analysis with a computer. *no* machine
can search that deep at present. Apparently no human either as Kasparov never
saw the repetition...
If you don't see the repetition, Kf1 makes sense in moving toward the
center. If you can see the repetition, Kf1 is a mistake. It was just a
lack of depth, in an impossibly deep position. But remember, DB played
the wrong move. Kasparov did *not* see the repetition. So it either
wasn't "obvious, as he had claimed" or else he blundered somehow...
:>: And finally all the possible tweaking and personality twisting has to be
:>: done before the match. I mean the human input. The machine itself might
:>: change itself whatever whenever it wanted. Please not the stuff about Garry
:>: talking to his mother, looking TV and eating lobsters again. Because you
:>: once told me that therefore also the machine had to have chats with Joel
:>: Benjamin ... Or with Hsu even better. :)
:>I disagree. Unless you seal the computer in one room, and Kasparov in
:>another. No phone. No visitors. No *nothing*. I'd take such a match
:>opportunity, *any* day...
: Frankenstein!
and the basic idea fails anyway... what about setting things up? what if
the machine crashes due to power failure? Hardware problem? software
problem? How could you be *sure* that the person fixing it didn't also
tweak the program?
Vincent Diepeveen
>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>
>: As a response to Stefan's post to CCC (sorry, but I'm banned from posting
>: there) I can confirm, despite nonsensical allegations to the contrary from
>: those who should know better:
>
>As the one who originally complained about auto232, I'll respond. Note
>that I *specifically* addressed "auto232" and not "winauto232". auto232
>is poorly designed, from the ground up. The protocol is unnecessarily
>complicated, requiring a tab here, a space there, etc. It does *not*
>allow underpromotion of any kind, which is rediculous. It has timing
>bugs that make moving too quickly hang it. Or moving too slowly causes
>it to time out. Just like Goldilocks and the three bears, it has to be
>"just right" to work.
>
>: 1. the auto232 code supplied by Stefan works just fine.
>
>Never said it didn't, although it is based on a protocol that is flawed
>badly. Stefan did fix the underpromotion problem, but only if you play
>a win version vs a win version, not a dos version. But the protocol is
>still byzantine and unnecessarily complicated. Winboard/Xboard interface
>protocol is very easy to understand, with keyword commands, and specific
>(and easy to find) commands to the engine, and from the engine. Including
>information about the opponent's time and so forth...
>
>
My program supports both winboard and auto232.
Auto232 took me 2 hours to make.
Winboard support still has some problems (as you must claim a draw).
For auto232 you only need:
gets(line);
fprintf(stdprn,"...etc\n");
That are the 2 things you need
That underpromotion is not supported in auto232 doesn't matter.
That's only 1 in a thousand games at most, and even then already in a lost
position. So that doesn't matter.
Auto232 protocol is a rude protocol that is *very easy* to get working
within a few hours, where winboard can't be done within a few hours.
For example: my program didn't claim a draw at icc, and the rules for
winboard are that it must claim a draw itself. This means that
you forfeit every say 10th game.
Then another problem is that you claim a draw, but this draw offer is
sent *before* you have read that game has already ended!
If you lag, then the opponent can then rematch and will get a draw offer
at the first move already. They nearly all take that, as Judgeturpin
run by Bradley Woodward plays everybody (in contradiction to DiepX, crafty
and others who play only people that have at most 400 rating points
less or something).
That kind of problem with xboard/winboard is not easy to fix. It's a lot
of work, where the rude auto232 works directly and easy.
For windows you need to have a non-console mode program (crafty is currently
a console program just like diep in order to work with winboard) in order
to support auto232, but that'll be not that difficult either. Source from
Stefan is very well readable.
There is no need to make your own auto player unless you want to cheat somehow.
There are good reasons to go from DOS to Win32 auto232 player, as
the new windows microsoft visual c++ 6.0 is at least 10% faster than msvc 5.0,
which was already way faster than any other compiler.
Greetings,
Vincent
--
+----------------------------------------------------+
| Vincent Diepeveen email: vdie...@cs.ruu.nl |
| http://www.students.cs.ruu.nl/~vdiepeve/ |
+----------------------------------------------------+
Vincent Diepeveen
>On Sat, 12 Sep 1998 19:53:42 +0100, in news:<905626498.8516.0...@news.demon.co.uk>,
>Chris Whittington <chr...@cpsoft.demon.co.uk> wrote in rec.games.chess.computer:
>>>offering/accepting a draw?
>>Big deal for autoplayer games between comps. Play it out.
>
>>>changing the "protocol system" to work with ethernet? Have to recompile
>>>*all* the engines...
>>
>>Get a standard system, like the rest of us.
>
>Ethernet, not standard enough ? You must be kidding.
>
>>>What about time? can you find out how much time your opponent actually
>>>has?
>>
>>Unless you want to mess around with bullet, this don't matter.
>
>I think it does. At least many humans try to take advantage of their
>opponent's zeitnot by playing faster (which may or may not turn out to
>be a good idea). Computer could do that too.
Tell me what program gets into time trouble like ahuman does?
I mean i played last Sunday a game vs IM v/d Weide.
At a certain time i had 5 minutes time left, and v/d Weide had 50 minutes
left.
I've never seen computer games with that much difference in time.
They think nearly all about the same time.
So if A is in time trouble then B too.
Further you KNOW that you lose say 2 seconds a move.
So for 60 moves inaccuracy is like 2 minutes. After 60 moves
next game gets started.
So for 40 moves in 2 hours 2 minutes difference is nothing.
That's less than 2%.
Further you *know* that it's 500 milliseconds a character,
so if you'd program for that you get even more accurate.
auto232 player is a rude auto player which is designed to get working
fast, and to work easily, unlike winboard.
Winboard is meant for unattended service at the internet, which
includes draw offers, 50 move rule etc.
Greetings,
Vincent
>
>--
>Marc Roger
Robert Hyatt
: My program supports both winboard and auto232.
: Auto232 took me 2 hours to make.
: Winboard support still has some problems (as you must claim a draw).
: For auto232 you only need:
: gets(line);
: fprintf(stdprn,"...etc\n");
: That are the 2 things you need
You need more. It took me 30 minutes to do the first auto232
version. It took months to figure out how to delay at the right
times to avoid hanging the silly thing. And when in tablebase
endings Crafty (and other programs using them) hang the thing at
times because we move too quickly, still, or the disk IO is causing
some system silliness or something...
: That underpromotion is not supported in auto232 doesn't matter.
: That's only 1 in a thousand games at most, and even then already in a lost
: position. So that doesn't matter.
Dead wrong. In 1992 at the ACM event a program (lachex) was winning.
Outright winning. But it was in a very sharp endgame, and the winning move
was missed at one point because LaChex didn't understand underpromotion
either, and it queended (not with check) thinking the opponent was going to
queen too, but the opponent underpromoted into a knight with check, and won
quickly. I watched a game Crafty vs some IM a couple of years ago that
was a KNN vs KP ending, where the only reason the game could be drawn was
that at the right moment, the pawn promoted into a knight with check, We
had a bug in winboard/xboard interface that made underpromotion not work,
it promoted into a queen and got mated by the two knights.
So this does happen. And it should be in any interface protocol that is
designed to handle program-to-program games. IE winboard/xboard does it.
The new winauto232 does it...
: Auto232 protocol is a rude protocol that is *very easy* to get working
: within a few hours, where winboard can't be done within a few hours.
I had the first crafty working about as quickly as auto232. Of course,
adding resuming adjourned games took handling the "edit" command, but
I don't see any difference in the complexity of the two protocols. Both
ship moves back and forth. the winboard protocol does update your clock
after each move, but most programs already have code to do that...
: For example: my program didn't claim a draw at icc, and the rules for
: winboard are that it must claim a draw itself. This means that
: you forfeit every say 10th game.
that's part of the game of chess. When you play a human do you tell him
that "the game is drawn by the 50-move rule" or "the game is drawn by
three-fold repetition"?? If so, you can just output "draw" when using
winboard and that will do it... if you don't claim the draw, nothing
says the human must, it is optional to go beyond three repetitions or
the 50 move rule if both want to...
: Then another problem is that you claim a draw, but this draw offer is
: sent *before* you have read that game has already ended!
: If you lag, then the opponent can then rematch and will get a draw offer
: at the first move already. They nearly all take that, as Judgeturpin
: run by Bradley Woodward plays everybody (in contradiction to DiepX, crafty
: and others who play only people that have at most 400 rating points
: less or something).
: That kind of problem with xboard/winboard is not easy to fix. It's a lot
: of work, where the rude auto232 works directly and easy.
Yes... by totally ignoring it. Which is not exactly a solution, IMHO...
: For windows you need to have a non-console mode program (crafty is currently
: a console program just like diep in order to work with winboard) in order
: to support auto232, but that'll be not that difficult either. Source from
: Stefan is very well readable.
: There is no need to make your own auto player unless you want to cheat somehow.
: There are good reasons to go from DOS to Win32 auto232 player, as
: the new windows microsoft visual c++ 6.0 is at least 10% faster than msvc 5.0,
And there are also good reasons to write a protocol that "actually conforms
to the rules of chess" so that you can offer/accept draws, check the official
game clock, underpromote, not hang due to TSR timing issues, etc...