AEGON IS UNFAIR!
Jouni Uski
are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
Clearly the chances for more points/bigger ELO are for the first type of
participants. In many years there has been praises for program X for result say 5/6, but but second
version of X with 2/6 is totally "forgotten".
Comments?
Jouni Uski
James Garner
: I think AEGON tournament is unfair because some programs like Rebel, Quest
Correct.
In fact, the potential for conflict is even worse when these
machines "meet" and the machine that is doing poorly in the tournament up
to then "loses."
Marcel van Kervinck
: In fact, the potential for conflict is even worse when these
: to then "loses."
How exactly do you imagine two machines 'meet' in a tournament that is
computer vs. human only?
Marcel
-- _ _
_| |_|_|
|_ |_ Marcel van Kervinck
|_| bue...@urc.tue.nl
M D. Grimminck
>I think AEGON tournament is unfair because some programs like Rebel, Quest
>are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
>Clearly the chances for more points/bigger ELO are for the first type of
>participants. In many years there has been praises for program X for result say 5/6, but but second
>version of X with 2/6 is totally "forgotten".
>Comments?
>Jouni Uski
The Aegon tournament is completely fair, it's just the programmers/sellers
that *potentially* are not. I am afraid it is common practice
to emphasize good results and fail to remember to tell the bad ones. (And not
just for chess programs!) Remember a program that reported a 2805 TPR
some time ago? (TPR=tournament prestation rating)
In any tournament there will always be players with an exceptionally
high TPR. In fact, if you get a perfect score, your TPR will be infinite,
and nobody will argue that he/she/it is an infinitely strong player,
because of one TPR it once achieved.
Ofcourse having a machine playing 12 or 18 games instead of 6,
means the programmer (and the public) gets more accurate information on
the playing strength of the program.
However, I doubt that the results of the different version indicate
their relative playing strength, as the number of games played is
simply too small. You need to play at least 100 games or so to
get any significant result.
--
Michel Grimminck, Computational Physics, University of Amsterdam.
draughts/checker page: http://carol.fwi.uva.nl/~grimmink/draughts.html
Chris Whittington
>
> I think AEGON tournament is unfair because some programs like Rebel, Quest
> are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
> Clearly the chances for more points/bigger ELO are for the first type of
> participants. In many years there has been praises for program X for result say 5/6, but but second
> version of X with 2/6 is totally "forgotten".
>
> Comments?
>
> Jouni Uski
>
>
Life is unfair.
This has been going on for ages.
Currently:
2 x Mephisto and a Genius (3 Lang entries)
2 x Virtual Chess
Fritz and Quest (2 Morsch entries)
2 x Rebel
2 x De Konig (R30 and The King)
For me it is no problem, why not have more programs.
The problem comes when the marketing supporters extol the virtues
of a program because it scored 5/6 or something, when it also scores
2/6.
This is not an AEGON problem, but a sales and marketing problem.
You could take the AEGON results, average out all the 'multiple entries'
and post the revised results to the news group ?
Then heavily attack any 'optimised' marketing claims ?
Bets regards,
Chris Whittington
Robert Hyatt
Jouni Uski <Jouni...@nce.nokia.com> wrote:
>I think AEGON tournament is unfair because some programs like Rebel, Quest
>are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
>Clearly the chances for more points/bigger ELO are for the first type of
>participants. In many years there has been praises for program X for result say 5/6, but but second
>version of X with 2/6 is totally "forgotten".
>
>Comments?
>
>Jouni Uski
>
>
Hi Jouni... This has been a problem for years. One example was Fidelity
entering multiple copies of their machine in (say) the US open, then advertising
the "best" finisher. Ditto for various microcomputer events. There was even
cases of tampering with the results, by controlling the outcome of a game should
two programs from the same vendor meet in a round.
I don't understand the purpose of the Aegon event anyway, it's less controlled
than (say) the Harvard Cup (as you pointed out above). In any case, you simply
have to take any "results" from such events with a small (or sometimes large)
grain of salt.
Bob
--
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
James Garner
: James Garner (da...@tiac.net) wrote:
: : In fact, the potential for conflict is even worse when these
: : machines "meet" and the machine that is doing poorly in the tournament up
: : to then "loses."
: How exactly do you imagine two machines 'meet' in a tournament that is
: computer vs. human only?
Not at all.
I was reflecting on some earlier 1980's computer only tournaments,
where one manufacturer would enter several models. In one situation,
perhaps Steve can recall, there was a small scandal about how one of the
models from one manufacturer (was this Fidelity?) lost to another Fidelity
machine....perhaps someone recalls the exact circumstances.
Chris Whittington
>
snip-snip
>
> I don't understand the purpose of the Aegon event anyway, it's less controlled
> than (say) the Harvard Cup (as you pointed out above). In any case, you simply
> have to take any "results" from such events with a small (or sometimes large)
> grain of salt.
>
> Bob
What is the purpose, eh ? A more mechanistic comment would be hard to
generate.
What is the purpose of a butterfly, what is the purpose of a Robert
Hyatt, what is the purpose of this posting ?
THEY JUST ARE.
Aegon.
1. Its organised by de Gorter, because he finds it interesting and fun.
2. Its attended by many Dutch people, The Dutch are the biggest players
of all sorts of mind games in the World.
3. Its reviewed on Dutch, German, and French TV.
4. Its reported in most European newspapers.
5. Its great to attend, to talk to other programmers, to talk to the players.
6. Its fun.
7. Want more reasons ?
8. It doesn't take place in America.
9. It isn't controlled by the ICCA mafia.
10. It isn't controlled by any mafia.
11. Bob Hyatt doesn't play there.
12. Bob Hyatt could play there if he wanted.
13. Bob Hyatt could just send Crafty, and the nice Dutch people would
provide an operator.
14. It provides a measure of 50 programs collectively against 50 humans.
15. It provides evidence that computer programs are massively overrated
on grading lists. Like knock off 200 ELO's.
16. It provides some kind of progress measure of programs against
computers, or, more recently of humans getting better against computers.
17. Er.
18. That's it for now.
Chris Whittington
Robert Hyatt
Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
>hy...@cis.uab.edu (Robert Hyatt) wrote:
>>
>snip-snip
>>
>> I don't understand the purpose of the Aegon event anyway, it's less controlled
>> than (say) the Harvard Cup (as you pointed out above). In any case, you simply
>> have to take any "results" from such events with a small (or sometimes large)
>> grain of salt.
>>
>> Bob
>
>What is the purpose, eh ? A more mechanistic comment would be hard to
>generate.
>
>What is the purpose of a butterfly, what is the purpose of a Robert
>Hyatt, what is the purpose of this posting ?
>
>THEY JUST ARE.
>
>Aegon.
>
>1. Its organised by de Gorter, because he finds it interesting and fun.
>2. Its attended by many Dutch people, The Dutch are the biggest players
>of all sorts of mind games in the World.
Debatable, but not worth arguing about... :)
>3. Its reviewed on Dutch, German, and French TV.
>4. Its reported in most European newspapers.
>5. Its great to attend, to talk to other programmers, to talk to the players.
What programmers are there?
>6. Its fun.
>7. Want more reasons ?
>
>8. It doesn't take place in America.
>9. It isn't controlled by the ICCA mafia.
>10. It isn't controlled by any mafia.
Don't know what you mean about the ICCA mafia. Been associated with them
since it was created in 1977. Don't agree with everything they do, but
there is *no one* else out there looking out for the interests of computer
chess.
>11. Bob Hyatt doesn't play there.
>12. Bob Hyatt could play there if he wanted.
Debatable. I was contacted, but nothing happened. I have no qualms about
playing anywhere, because I enjoy this. Ferret also was not invited, although
I won't venture a guess as to what that means. I've never been asked to compete
there, either with Cray Blitz or Crafty. Didn't bother me at all. I've never
been asked to the White House either.
>13. Bob Hyatt could just send Crafty, and the nice Dutch people would
>provide an operator.
See above.
>14. It provides a measure of 50 programs collectively against 50 humans.
>15. It provides evidence that computer programs are massively overrated
>on grading lists. Like knock off 200 ELO's.
I agree with that... :)
>16. It provides some kind of progress measure of programs against
>computers, or, more recently of humans getting better against computers.
>17. Er.
>18. That's it for now.
>
>Chris Whittington
My query is "where's the science?" Wouldn't the computer-chess group
look better with a dozen Genius clones there? How would you compare that
to a year ago with only two genius clones? Etc. In short, to draw
conclusions, you can change only one thing at a time, in this case, the
playing skill of the programs. It simply strikes me as an event that
should be fun to enter, and interesting to watch, but scientifically
useless if you try to draw conclusions about how many games are won/lost
by humans and by the computers as two "groups."
John Stanback
gives the best estimate of how computer programs actually rate against human
competition.
But I agree with your opinion that it is misleading to compare results when
some authors have entries under multiple names which improves their chances for
a good result. I think it would be best if there were only one program allowed
for each author.
John Stanback
Ed Schröder
Rebel plays with:
a) the normal commercial version Rebel 7.0
b) an experimental version named Rebel Aegon
Also Frans Morsch does the same:
a) Fritz4 (the commercial version)
b) Quest (his experimental version)
Please note that AEGON has no problems with that, they allow programmers
to play with their -STRONGEST- (most up to date version).
Others (Lang, Hirsch etc.) will have the -SAME- possibility but they
don't make use of it.
BTW (1) : Genius is present with THREE (commercial) programs.
BTW (2) : Chessica is also Frans Morsch, it was NOT planned but Socrates
didn't show up.
Ed Schröder
>I think AEGON tournament is unfair because some programs like Rebel, Quest
>are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
>Clearly the chances for more points/bigger ELO are for the first type of
>participants. In many years there has been praises for program X for result say 5/6, but but second
>version of X with 2/6 is totally "forgotten".
>
>Comments?
>
>Jouni Uski
- Ed Schroder -
Sorry for not signing the previous message.
Vincent Diepeveen
my compliments. A great setup!
In <4ktbao$2...@axl02it.ntc.nokia.com> Jouni Uski <Jouni...@nce.nokia.com> writes:
>I think AEGON tournament is unfair because some programs like Rebel, Quest
>are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
>Clearly the chances for more points/bigger ELO are for the first type of
>participants. In many years there has been praises for program X for result say 5/6, but but second
>version of X with 2/6 is totally "forgotten".
>
>Comments?
I guess it is to fill up the number of participants.
Also:
Can someone tell me the difference between
Genius
and
Mephisto Genius?
But let's take a look at your own computer first.
Do you have several versions of Fritz on your disk,
and only 1 of Zarkov?
Fact is that there is 1 type of position at the Aegon tournament
all programs loose, no matter their opponent, no matter the programs
name: a position without open files and with a closed centre is without
chances for programs. You can only pray that the opponent will give
away a piece.
In this respect it is not surprisingly that certain strong
players have lost a lot of points.
On the other hand, a position wih an open files the programs
sometimes seem to play perfectly.
For example my own program played second round against
Vaganian: open e-file, isolated pawn
at d5, halfopen c,d file (french). Diep played great, and
would have had the better chances in the endgame, if it would have
centralized its King. I still don't know why it didn't do this,
as it gets a lot of points for centralizing King in these endings.
Fact is that after 4 moves my program was out of book, and managed
to get a nice ending (Bishop+Knight against Rook up) against one
of the world top players.
Now take for example a program like Genius that should be much
better than my own program, even joins 2 times in the tournament
(not to mention the hardware from the same programmer that joins).
Because of its hazardeously play it seems to be without chances:
Usually after few moves the human players get programs out of book,
which takes care that a lot of so called 'beasts' getting over
100,000 nodes a second have a lot of positional problems even against
1900 players. As the average age is quite high in the aegon tournament
(mainly piece-giving away journalists), one should not conlude what
one sees as a result. The first 20 to 30 moves show clearly the
drawbacks of programs. For example my own program is missing certain
information concerned with defending its own kings position. So
it simply lost against Michael Hoving, where replacing Michael Hoving
by a computer would have caused my program to win: Michael Hoving
was busy putting my program to mate, so he didn't care about his
own kingsposition too much and he didn't want to exchange the Queen,
where a program playing with white would have done 1 of the
things you could afford with white: it would
have exchanged Queen, or it would have cared about its own kings position
too much, or it wouldn't have played for a mate.
Programs have a lot to learn. They seem to miss goals, and are only
great in waiting for the opponent to make action.
Depth doesn't count against humans. Tactically humans are outsearched
anyway. Knowledge works. Programs following ideas work even better
(although this is in fact a result of knowledge).
You must be lucky that your opponent doesn't
play the type of game your program doesn't like. Better openingsbook
trying to get into the type of positions you like doesn't work
against human, they are smart.
Joining with 2 instead of 1 program
will certainly give you a greater chance that your program
scores well, but still you must be lucky that your opponent makes a
tactical mistake, because even in positions programs should play well,
it seems that not-giving-pieces-away opponents should win. Only the
chance that they miss something in these positions seems much higher,
especially if you look after the average age of the opponent... :)
Anyway, the tournament is a great success, and as a programmer you can
learn a lot about how to improve your program. Humans keep on finding
weaknesses in your program... :)
Another interesting thing i see in this tournament is that
deeply calculating programs play better than heavy selective programs
using forward pruning (without search-replacement like nullmove).
It is a great topic to talk about!
Another interesting thing is that equal amateur-programs,
using slow evaluation (with static evaluation in the leaves) seem
to perform better than amateur programs using fast evaluation; piece-square
tables only. May be because in computer-computer tournaments the depth
you get is more important, because of horizon effects?
Another great topic!
>Jouni Uski
Vincent Diepeveen
vdie...@cs.ruu.nl
--
+--------------------------------------+
|| email : vdie...@cs.ruu.nl ||
|| Vincent Diepeveen ||
+======================================+
Vincent Diepeveen
>>5. Its great to attend, to talk to other programmers, to talk to the players.
>What programmers are there?
All European programmers, and a few american programmers.
So everyone that can pay a ticket, or can come by car.
However if you look at the programs that join, the only missing
2 are Deep Blue and Crafty.
>>9. It isn't controlled by the ICCA mafia.
>>10. It isn't controlled by any mafia.
Wrong, the opening was controlled by one.
There was an openingsspeech of prof. V/d Herik.
As my opponent of that evening said: "a speech of previous year".
I couldn't follow the long speech. Taking too much systemtime... :)
Still there is an indication that his optimism begins to change.
In this 3 horse race V/d Herik is currently betting on 3 horses.
His estimation written down about 140 to 160 points scored by the computers
50 games every day, 6 rounds. Makes 300 points.
140 points means that computers loose.
150 points means that computers play equally.
160 points means that computers win.
So he bets on all 3 possible outcomes: win,draw,loss!
>My query is "where's the science?"
It is taking science back to where it belongs: serving the audience.
Why making chessprograms? Because the public that has bought a computer
wants to buy a program. The Aegon tournament also gives huge ability
to put chess/computerchess in the spotlights.
> Wouldn't the computer-chess group
>look better with a dozen Genius clones there? How would you compare that
>to a year ago with only two genius clones? Etc. In short, to draw
>conclusions, you can change only one thing at a time, in this case, the
>playing skill of the programs. It simply strikes me as an event that
>should be fun to enter, and interesting to watch, but scientifically
>useless if you try to draw conclusions about how many games are won/lost
>by humans and by the computers as two "groups."
Why investigating how to proof programs if there is
evidence that you cannot check a program to be correct?
Quest (Fritz 5) is playing there, and is performing very well.
At home, there are no grandmasters to test his program,
there are not even professional chessplayers to test his program.
Only other programs.
Ever played 90 minutes a game + 20 seconds added every move in
ICS at a P166 against Grandmasters?
In this tournament this possibility exists, and it exists in
the Netherlands, and NOT in America!
Robert Hyatt
Ed Schröder <10065...@compuserve.com> wrote:
-->Jouni Uski <Jouni...@nce.nokia.com> wrote:
-->>I think AEGON tournament is unfair because some programs like Rebel, Quest
-->>are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
-->>Clearly the chances for more points/bigger ELO are for the first type of
-->>participants. In many years there has been praises for program X for result say 5/6, but but second
-->>version of X with 2/6 is totally "forgotten".
-->>
-->>Comments?
-->>
-->>Jouni Uski
-->
-->
-->Your conclusions are not correct.
-->
-->Rebel plays with:
-->a) the normal commercial version Rebel 7.0
-->b) an experimental version named Rebel Aegon
-->
-->Also Frans Morsch does the same:
-->a) Fritz4 (the commercial version)
-->b) Quest (his experimental version)
-->
-->Please note that AEGON has no problems with that, they allow programmers
-->to play with their -STRONGEST- (most up to date version).
-->Others (Lang, Hirsch etc.) will have the -SAME- possibility but they
-->don't make use of it.
-->
-->BTW (1) : Genius is present with THREE (commercial) programs.
-->
-->BTW (2) : Chessica is also Frans Morsch, it was NOT planned but Socrates
-->didn't show up.
-->
-->- Ed Schroder -
-->
-->Sorry for not signing the previous message.
-->
-->
-->
Hi Ed... Hope you didn't take my later comments as a personal thing
directed toward you. My two "interests" are:
(1) some will enter multiple copies, then advertise the best result,
even though an identical copy finished with a much worse result. Sort
of like cheating the bell curve: if you enter enough clones, one will
be way out (some bad, but some also very good). That's simply confusing.
It's not nearly so bad as I've seen in other events in the past (and no,
they were not Aegon at all, but the venerable US Open (fidelity and
others come to mind) as well as the ICCA-WMCCC. I think the abuses
there might have been affected by greed, because each machine pays an
entry fee if it's commercial, encouraging the ICCA to accept 'em and
the authors (because of the prestige) to enter 'em.
(2) There are some absolutely notable absences there. The biggest, of
course, is Deep Blue. However, Ferret is another program that has been
proven to be among the very best group, and he wasn't even contacted
about participating. Cray Blitz has never been invited nor has it ever
participated, again not a major issue, but a possible "flaw." Perhaps
it's a simple issue of "inviting yourself" there, I don't know. However,
with the first two omissions above, the result is probably not as good as
it could be, which detracts from anything "scientific". I'd agree that
it sounds like fun; that's why I've entered a human event here and there
over the years myself. However, it's a least "purported" to be more than
a simple tournament, it's "the" man-machine event of the year.. yet there's
some serious holes in who's there...
In any case, from my perspective, you've been one of the more (most?) open
of the commercial guys about what you're doing. Keep it up. :)
Bob
Robert Hyatt
Vincent Diepeveen <vdie...@cs.ruu.nl> wrote:
-->
-->Another interesting thing i see in this tournament is that
-->deeply calculating programs play better than heavy selective programs
-->using forward pruning (without search-replacement like nullmove).
-->
-->
-->Another interesting thing is that equal amateur-programs,
-->using slow evaluation (with static evaluation in the leaves) seem
-->to perform better than amateur programs using fast evaluation; piece-square
-->tables only. May be because in computer-computer tournaments the depth
-->you get is more important, because of horizon effects?
-->
-->Another great topic!
-->
I think I've said that a dozen times. There are two goals here that are
not necessarily taking us along the same path. (1) beat computers (2)
beat humans. In writing a program, these seem to be somewhat at odds at
times. Makes for a lively discussion, although it's done this for at least
26 years now since "Tech" hit the ACM event... and programs like Tech have
been around forever...
Robert Hyatt
Vincent Diepeveen <vdie...@cs.ruu.nl> wrote:
-->In <4kurm3$j...@pelham.cis.uab.edu> hy...@cis.uab.edu (Robert Hyatt) writes:
-->
-->>>5. Its great to attend, to talk to other programmers, to talk to the players.
-->
-->>What programmers are there?
-->
-->All European programmers, and a few american programmers.
-->So everyone that can pay a ticket, or can come by car.
-->However if you look at the programs that join, the only missing
-->2 are Deep Blue and Crafty.
Don't forget Ferret, which is probably as good as any program playing
there right now...
-->
-->>>9. It isn't controlled by the ICCA mafia.
-->>>10. It isn't controlled by any mafia.
-->
-->Wrong, the opening was controlled by one.
-->
-->There was an openingsspeech of prof. V/d Herik.
-->As my opponent of that evening said: "a speech of previous year".
-->I couldn't follow the long speech. Taking too much systemtime... :)
-->
-->Still there is an indication that his optimism begins to change.
-->In this 3 horse race V/d Herik is currently betting on 3 horses.
-->
-->His estimation written down about 140 to 160 points scored by the computers
-->
-->50 games every day, 6 rounds. Makes 300 points.
-->140 points means that computers loose.
-->150 points means that computers play equally.
-->160 points means that computers win.
-->
-->So he bets on all 3 possible outcomes: win,draw,loss!
-->
-->>My query is "where's the science?"
-->
-->It is taking science back to where it belongs: serving the audience.
-->Why making chessprograms? Because the public that has bought a computer
-->wants to buy a program. The Aegon tournament also gives huge ability
-->to put chess/computerchess in the spotlights.
-->
-->> Wouldn't the computer-chess group
-->>look better with a dozen Genius clones there? How would you compare that
-->>to a year ago with only two genius clones? Etc. In short, to draw
-->>conclusions, you can change only one thing at a time, in this case, the
-->>playing skill of the programs. It simply strikes me as an event that
-->>should be fun to enter, and interesting to watch, but scientifically
-->>useless if you try to draw conclusions about how many games are won/lost
-->>by humans and by the computers as two "groups."
-->
-->Why investigating how to proof programs if there is
-->evidence that you cannot check a program to be correct?
-->
-->Quest (Fritz 5) is playing there, and is performing very well.
-->At home, there are no grandmasters to test his program,
-->there are not even professional chessplayers to test his program.
-->Only other programs.
-->
-->Ever played 90 minutes a game + 20 seconds added every move in
-->ICS at a P166 against Grandmasters?
Yes, although not on a P166. I've played hundreds of "standard" games
on ICC however. Against GMs, IMs, and FM's...
-->
-->In this tournament this possibility exists, and it exists in
-->the Netherlands, and NOT in America!
-->
-->Vincent Diepeveen
-->vdie...@cs.ruu.nl
-->
-->
-->
-->--
--> +--------------------------------------+
--> || email : vdie...@cs.ruu.nl ||
--> || Vincent Diepeveen ||
--> +======================================+
Brian Neenan
Marcel van Kervinck posted:
"How exactly do you imagine two machines
'meet' in a tournament that is computer vs.
human only?"
Surreptitiously, by a secluded power socket in the
analysis room after their chaperons have pushed
off to the bar.
Regards. ~~ ~~
~~
~~ ~~
SYSOP
> 5/6, but but second
JU> Comments?
JU> Jouni Uski
I do not agree because they are different programs. The two programs of
Schoreder are REBEL 7.0 (the commercial version, so the customers can show
a "real practical test") and REBEL AEGON, a new version under development
which is not commercially available at this moment. Quest is Frans Morsh
"personal" version of Fritz, not the commercial one. Besides, Schroeder and
Morsch are "dutchmens" and surely for this they were able to put more than
one machine in scene.
And I have not hear any complaints from Marty Hirsch (MChess programmer)
and the same for Mark Uniacke (Hiarcs). AEGON is a good tournament for
promoting sales, but playing with two versions is risky ! What about if
REBEL 7.0 or FRITZ scored very bad ? Actually you can buy only the commercial
versions, not the "experimental" !
I think that mostly of chessplayers who have real intentions for buying
a chess program will decide under the followings statements:
a) strenght and prize !
b) general features
c) review of several games played by the chess programs.
d) final score in AEGON tournament
My view,
______________________________________________________________________________
Roberto Gabriel Alvarez e-mail: rob...@alvarez.satlink.net
CAISSA BBS -Ajedrez para Todos el...@giga.com.ar
Villa Ballester - Bs.As. -ARGENTINA BBS: (54) (1) 847-1521 14400 8,N,1
------------------------------------------------------------------------------
---
* SPEED 2.00 #2722 * AJEDREZ INTEGRAL -La Libreria de Morgado Tel:342-7304
SYSOP
JG> Jouni Uski (Jouni...@nce.nokia.com) wrote:
JG> : I think AEGON tournament is unfair because some programs like Rebel, Quest
> : are in 2 or 3 machines, whereas others like Mchess and Hiarcs only on one!
> : Clearly the chances for more points/bigger ELO are for the first type of
> : participants. In many years there has been praises for program X for result s
> ay 5/6, but but second
> : version of X with 2/6 is totally "forgotten".
JG> Correct.
JG> In fact, the potential for conflict is even worse when these
> machines "meet" and the machine that is doing poorly in the tournament up
> to then "loses."
AEGON is a "man vs computer" tournament, so it is impossible !!!!
______________________________________________________________________________
Roberto Gabriel Alvarez e-mail: rob...@alvarez.satlink.net
CAISSA BBS -Ajedrez para Todos el...@giga.com.ar
Villa Ballester - Bs.As. -ARGENTINA BBS: (54) (1) 847-1521 14400 8,N,1
------------------------------------------------------------------------------
---
* SPEED 2.00 #2722 * Circulo de Ajedrez de Villa Ballester, el mejor club!
Ed Seedhouse
> Remember a program that reported a 2805 TPR
>some time ago? (TPR=tournament prestation rating)
Prestation? That's a word I don't know. I thought it meant
"Tournament Performance Rating".
>In any tournament there will always be players with an exceptionally
>high TPR. In fact, if you get a perfect score, your TPR will be infinite,
Oh no it won't. It will be 400 points above the average of your
oponents if you use the C.F.C. rating formula. For FIDE the TPR may
be more or less but it *definitely* *will* *not* be infinite.
>However, I doubt that the results of the different version indicate
>their relative playing strength, as the number of games played is
>simply too small. You need to play at least 100 games or so to
>get any significant result.
The commonly accepted number is 25, not 100.
>Michel Grimminck, Computational Physics, University of Amsterdam.
>draughts/checker page: http://carol.fwi.uva.nl/~grimmink/draughts.html
Ed Seedhouse
President, Victoria Chess Club.
CFC Rating: 2058
Ed Schröder
>In article <4l020d$1...@krant.cs.ruu.nl>,
>Vincent Diepeveen <vdie...@cs.ruu.nl> wrote:
>-->In <4kurm3$j...@pelham.cis.uab.edu> hy...@cis.uab.edu (Robert Hyatt) writes:
>-->
>-->>>5. Its great to attend, to talk to other programmers, to talk to the players.
>-->
>-->>What programmers are there?
>-->
>-->All European programmers, and a few american programmers.
>-->So everyone that can pay a ticket, or can come by car.
>-->However if you look at the programs that join, the only missing
>-->2 are Deep Blue and Crafty.
>
>Don't forget Ferret, which is probably as good as any program playing
>there right now...
>Robert Hyatt Computer and Information Sciences
Yesterday I contacted AEGON about the Ferret / Crafty subject.
AEGON says that both Crafty and Ferret were invited to participate.
Hope both of you will be able to participate next year.
- Ed Schroder -
Ed Schröder
>(205) 934-2213 115A Campbell Hall, UAB Station
>(205) 934-5473 FAX Birmingham, AL 35294-1170
>
Hi Bob...
No, I did not feel offended at all.
About multiple copies...
You have a point here on the other hand I agree with Chris, that is if
program X scores (much) better than -identical- program Y and (only)
program X is used for advertising purposes, you simply attack the
producer.
In my case I mention both results (Rebel7 and Rebel AEGON).
I should be a fool if I only would mention the highest ranked one here in
the newsgroup.
About Crafty and Ferret...
This I do not understand.
From the AEGON organizer "Cock de Gorter" I heard that both Ferret and
Crafty were invited to participate. Don't know what went wrong, I only
hope that both of you will participate next year because both of you
belong there.
- Ed Schroder -
Shep
>grim...@fwi.uva.nl (M D. Grimminck) wrote:
>> Remember a program that reported a 2805 TPR
>>some time ago? (TPR=tournament prestation rating)
>Prestation? That's a word I don't know. I thought it meant
>"Tournament Performance Rating".
I guess that's what he meant. "Prestatie" is Dutch for "Performance".
Shep
Matt
>
> I think AEGON tournament is unfair because some programs like Rebel, Quest
>
>
> Jouni Uskithis is full o.k. !!!!!
Peter Schreiner
Robert Hyatt
Ed Schröder <10065...@compuserve.com> wrote:
-->hy...@cis.uab.edu (Robert Hyatt) wrote:
-->>In article <4l020d$1...@krant.cs.ruu.nl>,
-->>Vincent Diepeveen <vdie...@cs.ruu.nl> wrote:
-->>-->In <4kurm3$j...@pelham.cis.uab.edu> hy...@cis.uab.edu (Robert Hyatt) writes:
-->>-->
-->>-->>>5. Its great to attend, to talk to other programmers, to talk to the players.
-->>-->
-->>-->>What programmers are there?
-->>-->
-->>-->All European programmers, and a few american programmers.
-->>-->However if you look at the programs that join, the only missing
-->>-->2 are Deep Blue and Crafty.
-->>
-->>Don't forget Ferret, which is probably as good as any program playing
-->>there right now...
-->
-->>Robert Hyatt Computer and Information Sciences
-->
-->Yesterday I contacted AEGON about the Ferret / Crafty subject.
-->AEGON says that both Crafty and Ferret were invited to participate.
-->Hope both of you will be able to participate next year.
-->
-->- Ed Schroder -
-->
-->
As I responded to Ed privately, something got seriously overlooked here.
I was contacted and asked to send a list of FM/IM/GM players Crafty has
played on ICC and how it did against them. I did, and that was the *last*
thing I heard from anyone associated with Aegon. No "we'll see you there"
or anything.
In talking with Bruce (Ferret) he heard nothing from day one, so that doesn't
mesh with the above either. Steven Edwards is a 3rd chess programmer that has
participated in past ACM events, yet he heard nothing either...
I simply wanted to make it clear that I had planned on attending had I been
asked. I had talked to a "high-performance workstation vendor" about using
something fast (and parallel) for the event. However, when nothing happened,
I simply assumed that the "selectivity" criteria were such that Crafty didn't
make the cut, and I simply wrote it off...
Maybe I've been spoiled by the ACM and World Championship events, where there's
a formal entry procedure and a formal acceptance notification?
Lonnie Cook
>Don't forget Ferret, which is probably as good as any program playing
>there right now...
>-->
>-->Quest (Fritz 5) is playing there, and is performing very well.
>-->At home, there are no grandmasters to test his program,
>-->there are not even professional chessplayers to test his program.
>-->Only other programs.
>-->
>-->Ever played 90 minutes a game + 20 seconds added every move in
>-->ICS at a P166 against Grandmasters?
>Yes, although not on a P166. I've played hundreds of "standard" games
>on ICC however. Against GMs, IMs, and FM's...
>-->
>-->In this tournament this possibility exists, and it exists in
>-->the Netherlands, and NOT in America!
>-->
>-->Vincent Diepeveen
>-->vdie...@cs.ruu.nl
>-->
>-->
>-->
>-->--
>--> +--------------------------------------+
>--> || email : vdie...@cs.ruu.nl ||
>--> || Vincent Diepeveen ||
>--> +======================================+
>--
>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
I totally agee with Bob on this one. I talked to Ferret on ICC a few days ago and asked m why he wasn't particpating in AEGON. He said he wasn't officially invited(although he said he probably could have
went into the tourney.) Last but not least which is some of his point too is the expense of going to the tourney. It is a shame Ferret did not enter. It certainly would have
been in the upper sections of the place listings. He's running the program at 200! Crafty will be running his at 200 also shortly! These programs have been tested in the
real world of ICC,GM's,IM's & FM's etc. I have watched these programs play and know that they would have a good placing
Respectfully Yours,
Lonnie J. Cook
"Destroyer"on ICC
<lonni...@riconnect.com>
RIconneCT (401) 596 7341
Westerly,RI own ISP
Robert Hyatt
-->>Don't forget Ferret, which is probably as good as any program playing
-->>there right now...
-->
-->
-->>-->
-->>-->Quest (Fritz 5) is playing there, and is performing very well.
-->>-->there are not even professional chessplayers to test his program.
-->>-->Only other programs.
-->>-->
-->>-->Ever played 90 minutes a game + 20 seconds added every move in
-->
-->>Yes, although not on a P166. I've played hundreds of "standard" games
-->>on ICC however. Against GMs, IMs, and FM's...
-->
-->
-->>-->
-->>-->In this tournament this possibility exists, and it exists in
-->>-->
-->>-->Vincent Diepeveen
-->>-->
-->>-->
-->>-->
-->>-->--
-->>--> +--------------------------------------+
-->>--> || Vincent Diepeveen ||
-->>--> +======================================+
-->
-->
-->>--
-->>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
-->
-->I totally agee with Bob on this one. I talked to Ferret on ICC a few days ago and asked m why he wasn't particpating in AEGON. He said he wasn't officially invited(although he said he probably could have
-->went into the tourney.) Last but not least which is some of his point too is the expense of going to the tourney. It is a shame Ferret did not enter. It certainly would have
-->been in the upper sections of the place listings. He's running the program at 200! Crafty will be running his at 200 also shortly! These programs have been tested in the
-->real world of ICC,GM's,IM's & FM's etc. I have watched these programs play and know that they would have a good placing
-->Respectfully Yours,
-->Lonnie J. Cook
-->"Destroyer"on ICC
--><lonni...@riconnect.com>
-->RIconneCT (401) 596 7341
-->Westerly,RI own ISP
Ugh. The dreaded "Destroyer"... :) While we are on the subject of Crafty,
thought I'd point out something really interesting I'm now seeing. Lonnie
plays a fair number of Fritz4 vs Crafty games on ICC, and I've been noticing
something fairly interesting, that also offers a contrast between computer vs
computer, and computer vs human.
As most here know, I've been specifically tuning Crafty to try and avoid the
locked positions, to play more aggressively, etc. In short, to not get into
a "shuffle" mode if possible. I watched two games yesterday when Lonnie was
playing Fritz4 vs crafty and they were most interesting. game one was a slow,
positional sort of game, that Crafty managed to win by simply not creating any
pawn weaknesses, etc. However, game two was quite different. In this game,
they both created threats, parried them, for a good while. Then Crafty got the
idea of playing h5 (it was black, both castled short) which was likely a good
move most of the time, but in this game, one rook was out of play (relative to
the kingside attack it was launching). When I saw it, I thought "aw sh**" :)
White's pieces had a better chance of getting to the kingside. I was right.
Crafty opened the h-file, but got killed on it about 15 moves later. The
interesting point? This sort of play is a lot of fun, but Fritz is known to
be very aggressive, with less attention to positional play, more on tactics,
and yet here goes Crafty, playing right into the teeth of the fritz program.
Obviously, it needs more evaluation, because what I'm doing is doing well
against carbon opponents, but it's not advisable against a computer opponent,
unless the computer opponent is *not* very aggressive.
As I've mentioned in the past, playing man or machine are two completely
different things on one level. BTW, I don't have summary data, but Fritz
4 is certainly winning more than it's losing, although I'm still working to
rectify this. :)
My "fear" however, is that to do better against programs like Fritz, is
likely to do worse against humans, because the more aggressive play is at
fault, and "turning it down" leads to less interesting games. I might
try some internal tuning, based on the opponent, like I already do for
the opening book.
Vincent Diepeveen
It appeared that
Mephisto Genius was a hardware chesscomputer.
Same for 1 of the 'The Kings'
>John Stanback
Chris Whittington
>
>
> What programmers are there (AEGON) ?
About 20, mostly from Europe (that's to the north of africa, BTW), but
also Berliner, who is there every year.
>
> Debatable. I was contacted, but nothing happened. I have no qualms about
> playing anywhere, because I enjoy this. Ferret also was not invited, although
> I won't venture a guess as to what that means. I've never been asked to compete
> there, either with Cray Blitz or Crafty. Didn't bother me at all. I've never
> been asked to the White House either.
I have it on good authority that you were invited but didn't respond.
AEGON does have a formal invite/acceptance proceedure, but you must
understand that their problem is not with computer entries, but with
getting 50+ reliable humans of appropriate strength.
This does tend to mean that they will not chase programmers who don't
show interest.
If you want to play in the future (and you should), then I'm sure
they will accomodate you.
>
>
> My query is "where's the science?" Wouldn't the computer-chess group
> look better with a dozen Genius clones there? How would you compare that
> to a year ago with only two genius clones? Etc. In short, to draw
> conclusions, you can change only one thing at a time, in this case, the
> playing skill of the programs. It simply strikes me as an event that
> should be fun to enter, and interesting to watch, but scientifically
> useless if you try to draw conclusions about how many games are won/lost
> by humans and by the computers as two "groups."
> --
> Robert Hyatt Computer and Information Sciences
Rubbish !
I've come to expect university researchers to fall back on this when
all else fails. After all, you are the only people who can do the
science (!) cunning, eh. The font of all knowledge, experience and valid
truth can only come from the university - the high-priests of all
knowledge can only be the professors.
Hmm, methinks not.
Where's the science in any sport, league table, boxing match ?
Best regards,
Chris Whittington
Robert Hyatt
Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
-->hy...@cis.uab.edu (Robert Hyatt) wrote:
-->>
-->
-->
-->> What programmers are there (AEGON) ?
-->
-->About 20, mostly from Europe (that's to the north of africa, BTW), but
-->also Berliner, who is there every year.
-->
-->>
-->> Debatable. I was contacted, but nothing happened. I have no qualms about
-->> playing anywhere, because I enjoy this. Ferret also was not invited, although
-->> I won't venture a guess as to what that means. I've never been asked to compete
-->> there, either with Cray Blitz or Crafty. Didn't bother me at all. I've never
-->> been asked to the White House either.
-->
-->I have it on good authority that you were invited but didn't respond.
-->AEGON does have a formal invite/acceptance proceedure, but you must
-->understand that their problem is not with computer entries, but with
-->getting 50+ reliable humans of appropriate strength.
-->This does tend to mean that they will not chase programmers who don't
-->show interest.
-->If you want to play in the future (and you should), then I'm sure
-->they will accomodate you.
You've got it from the "horse's mouth" that they didn't. I was contacted
once on the phone. I asked for information, was told things are not yet
final as to dates, rounds, etc. Told them I'd be happy to enter, although
I doubt I'd be able to attend personally. Expected a phone call back,
but never got anything. *not* a big deal to me. Bruce also did *not*
get an invitation, but again, I doubt he's any more concerned that I am.
-->>
-->>
-->> My query is "where's the science?" Wouldn't the computer-chess group
-->> look better with a dozen Genius clones there? How would you compare that
-->> to a year ago with only two genius clones? Etc. In short, to draw
-->> conclusions, you can change only one thing at a time, in this case, the
-->> playing skill of the programs. It simply strikes me as an event that
-->> should be fun to enter, and interesting to watch, but scientifically
-->> useless if you try to draw conclusions about how many games are won/lost
-->> by humans and by the computers as two "groups."
-->> --
-->> Robert Hyatt Computer and Information Sciences
-->
-->Rubbish !
-->I've come to expect university researchers to fall back on this when
-->all else fails. After all, you are the only people who can do the
-->science (!) cunning, eh. The font of all knowledge, experience and valid
-->truth can only come from the university - the high-priests of all
-->knowledge can only be the professors.
Counter-Rubbish. I didn't even imply that I would do the science. I simply
asked "where's the controls that make this event comparable year after year?"
There's not much. Multiple copies, missing best programs, ... That was my
only complaint... For all it's faults, the ACM event was very careful. The
time control never varied (excepting one time I believe) since 1970. Makes
it easy to compare programs of past against programs of present, when common
opponents are available. The bottom line is, I wasn't invited, I have not
made an issue of not being invited, and don't think that I'll lose sleep if
I'm never invited. It'd be fun, but it not a life-or-death issue for me,
I've played in enough computer events that one more, more or less, is not
going to make much difference to who I am or what I do. However, for the
most part, at least the "reporting" does come from academics. When's the
last time you saw something in writing about the insides of "xxx" where xxx
is a commercial program? Of course, the fact that xxx is built on all the
results produced by the poor academic types is beside the point?
Sorry you're so bitter toward academics... guess you had a bad experience
somewhere along the way. Fortunately, I haven't...
-->
-->Hmm, methinks not.
-->
-->Where's the science in any sport, league table, boxing match ?
anywhere the rules are constant, the participation is controlled, the
very best are always competing, etc. Of course, a Mike Tyson vs a
Joe-the-rag-man sanchez or whatever is not so interesting.
-->
-->Best regards,
-->
-->Chris Whittington
-->
--
Robert Hyatt Computer and Information Sciences
Tom Kerrigan
about AEGON. I gave him my address, phone number, etc. and he told me he would be
in touch. I noticed that he was having similar conversations with several other
programmers. At this point, Bruce was trying hard to not starve to death during
one of his longer games. The fellow observed the game for a few minutes, and went
off in search of another programmer. I can only assume this is why Bruce wasn't
invited while I was, and I can only laugh at speculation on this non-invitation.
Regarding science: I have explored the AEGON web site in some depth, and evidently
I missed the part about it being scientific. For some reason, I just couln't find
a purpose, hypothesis, procedure, or conclusion. Perhaps Hyatt would be so kind as
to tell us all what these are?
Cheers,
Tom
_______________________________________________________________________________
Tom Kerrigan kerr...@frii.com O-
All true wisdom is found on T-shirts.
Robert Hyatt
Easy:
The results published by Aegon include the line "computers:xxx humans:yyy"
It's also compared to last year's results, etc. The implication to me, as
a scientist, is that somehow the two tournaments are comparable and somehow
say something about how the computers have gotten better/worse since last
year.
From my perspective, this doesn't work. The strongest machine in the world
wasn't there, one of the top three PC programs (Ferret) wasn't there. There
were multiple copies of other programs there. There were many virtually
unknown programs there. In short, it was a tournament with people and
computers, but there was very little control which makes comparisons to past
(or future) events difficult if not impossible.
Personally, the event was interesting. However, as one of two man/machine
events held per year (the Harvard Cup is the other) it'd be nice if (a) there
were fewer entrants (check out the pairings, you'll note that programs that
finished with equal scores had drastically different tournament performance
results because of the wide spread of human players... perhaps too wide for
the few number of rounds played. That leaves a strong element of chance that
makes this even more hard to compare to anything.
About the only thing worse that can happen is to expand the field to 200, so
that multiple copies of the best, and single copies of every other program on
the planet can participate. Nothing wrong with that, of course, but it makes
the outcome much less "scientific". I had viewed this as a "scaled up Harvard
Cup" sort of event. In that regard, it didn't come close. If the intent is to
just play humans and computers, it was a success. If the intent was to measure
anything compared to last year or before, it's a scientific "flop".
Maybe I miss the point of the event completely, but the instant someone compares
this year to last year, I suspect they must have too, because the "implication"
is that all that's changed is the programs have gotten better. I haven't looked
at the complete list of humans, but it's certainly possible that the average
rating was down or up. Simply makes comparisons impossible, particularly when
two programs finish with a score of X, but have a performance rating of Y+/-200
points (or more).
Chris Whittington
>
Methinks you protest too much.
You weren't there.
You've never been there.
Its in Europe not America.
The ICCA doesn't control it.
You don't control it.
Its a centre for chess outside your's, acedemic, and ICCA control.
It doesn't regard you as the high priest of computer chess.
You just don't like the idea.
Admit it.
Best regards,
Chris Whittington
Chris Whittington
>
snip-snip
> -->I have it on good authority that you were invited but didn't respond.
> -->AEGON does have a formal invite/acceptance proceedure, but you must
> -->understand that their problem is not with computer entries, but with
> -->getting 50+ reliable humans of appropriate strength.
> -->This does tend to mean that they will not chase programmers who don't
> -->show interest.
> -->If you want to play in the future (and you should), then I'm sure
> -->they will accomodate you.
>
> You've got it from the "horse's mouth" that they didn't. I was contacted
> once on the phone. I asked for information, was told things are not yet
> final as to dates, rounds, etc. Told them I'd be happy to enter, although
> I doubt I'd be able to attend personally. Expected a phone call back,
> but never got anything. *not* a big deal to me. Bruce also did *not*
> get an invitation, but again, I doubt he's any more concerned that I am.
>
There's another horse's mouth of course in Holland, and that mouth
says they did. They say it really quite emphatically.
People will just have to choose who they believe, or not, I suppose.
snip-snip
Best regards,
Chris Whittington
Robert Hyatt
Chris Whittington <chr...@cpsoft.demon.co.uk> wrote:
right.
>You've never been there.
right.
>Its in Europe not America.
right. but what does this have to do with anything? Cray Blitz has
participated in World Championship events in Canada, Germany, the US,
and I don't remember where else. It's done demonstrations in England,
Paris, the Netherlands, etc. never has geography been an issue for me,
other than for costs of course.
>The ICCA doesn't control it.
Who cares? In general, the ICCA hasn't done badly in the tournaments
it holds every year, other than letting the commercial entrants get way
out of line. When they started charging big bucks for commercial
entries, you could see that problems were going to arise, which they have,
in spades. However, what's that got to do with me? I'm not an officer.
>You don't control it.
Again, so what. I'm not interested in organizing a tournament like that.
If I was, and could find the financial support, I would. However, I'm not,
I haven't tried, and I won't. I've never wanted to "control" any event of
any sort. Doesn't make a lot of sense.
>Its a centre for chess outside your's, acedemic, and ICCA control.
What's with the "control" thing? You seem to equate me with ICCA, which
I don't understand. I'm *not even a member*! Surprised? No doubt...
>It doesn't regard you as the high priest of computer chess.
Never wanted anyone to do so. I have my own opinion of who belongs
in this category. However, one thing I do claim to be is "open". If
someone has a question, I'll do my best to answer it. If they have a
suggestion, I'll do my best to evaluate it. You seem to overlook one
important point, that being that you can choose to ignore anything I
say. A "High Priest" can shove his opinions down your throat, and make
you like it as well.
>
>You just don't like the idea.
In a way, this is correct. However, it's not the "idea" I don't exactly
"like", rather, it's the "implementation".
>Admit it.
>
Remember where this thread started? Two questions, one from "Jouni": the
Aegon seems unfair with multiple copies of the same program. I responded
that this has been a problem since long before Aegon, dating back to the
early WMCCC events where there were often 4 entrants from each company,
and "collusion" was a common occurrence, leading to protest after protest.
I also pointed out that collusion doesn't occur here, but multiple copies
allows interesting advertising possibilities by reporting the best result
even if the "doppelganger" did much worse. Of course, you only have to
go as far back as Fidelity's US Open games to see examples of this...
The second from someone who's name I don't recall, asking why Crafty and
Ferret were not there, when they do well on ICC. I responded that neither
of us were invited. Regardless of what you heard to the contrary, both Bruce
and I stand by this. If you query the programmers on ICC, you'll find
that I continually asked them if they'd heard anything because I had talked
to the AEGON folks one time, but it was well before anything was set date
or time-wise. Also easy to verify that I had gotten an account set up at
HP to run some tests there so we could use one of their parallel workstations
at AEGON had it worked out.
The rest of the discussion is simply an "opinion" of mine, that you happen not
to agree with. Doesn't bother me in the least, because there's lots of things
I don't agree with either, and it's not going to change them.
As I said, I think the AEGON event was interesting, otherwise I would not have
even bothered trying to enter. However, it is also very "loosely" organized
and the results are somewhat difficult to interpret. I'll carefully give my
perspective, you can respond, and then we can call it "done" for the topic if
you agree:
1. The best programs were not all present and accounted for. The strongest
around (Deep Blue) was not there, not even in an older form like Deep Thought II
or whatever. Ferret was not there, yet he managed to beat Genius in the WMCCC.
I'm not even sure Crafty belonged there at all, and that's not a real issue to
me. It'll play in it's share of games. It's played about 40,000 games on ICC
now, with about 1/3 of them against IM and stronger players, so I get results
and feedback, which is really all I'm interested in at present. So, two of
the top 4-5 programs in the world were absent, including the #1 program.
2. The duplicate entrants may or may not be a problem. Since they don't play
each other, it's likely they are not. However, it would probably be just as
interesting to cut the number of computers and humans down to size, rather
than pad with duplicates. Minor point that's not a big deal.
3. the overall rating range on both the computers and the humans was
enormous. The effect of this (my opinion, remember) is that there was therefore
too much "luck" involved. Otherwise, explain the final standings. Take any of
the top groups and for two programs that finished with equal (or within 1/2
point) scores, look at the performance ratings. Who did better? A program with
4.5, TPR of 2300, or a program with 4.0 with a TPR of 2500? My complaint here
is that I don't know how to interpret that. At the Harvard Cup, it doesn't
happen (or at least not nearly so often.)
4. In an event that's once per year, and claims to be "the man-machine event
of the year" it would seem to me that the right way to select computers would
be to ask for applications just like the WMCCC or WCCC or ACM events, let a
screening committee evaluate them and pick the best group for the tournament.
I'm not knocking the AEGON event nor organizers here, just that it would be
nice if we could take the yearly crosstable and say "wow, the computers went
up almost 60 points from last year" or "wow, them machines finished with a
lower overall won-lost record than last year". To me, that's what'd make
AEGON unique. Otherwise, how's it different from any old event that is not
(NC)? In short, a controlled experiment produces scientific results, an
uncontrolled one produces entertainment.
Rene Vermaas
> 1. The best programs were not all present and accounted for. The strongest
> around (Deep Blue) was not there, not even in an older form like Deep Thought II
> or whatever.
It maybe is another (and senseless) discussion, but it's rather debatable if
Deep Blue is really the strongest around. For example, in the magazine MegaByte
Jeroen Noomen examined the first Deep Blue - Kasparov game (1-0). 7 PC programs
examined for 3 minutes about each move Deep Blue played. 25 moves (after the opening)
were examined. As well Rebel7 as Chess Genius3 played in 19 cases exactly the same
move as Deep Blue did. With human eyes examined the 6 different moves weren't worse.
The conclusion from Jeroen Noomen: The better PC programs undoubted should have
beaten Kasparov as well.
Of course I don't want to belittle Deep Blue, it's a real beast... But I wonder
what happens when Ed Schroder gets 200.000.000 positions per second... ;=)
Rene Vermaas.
Robert Hyatt
Rene Vermaas <rver...@globalxs.nl> wrote:
-->Robert Hyatt wrote:
-->
-->> 1. The best programs were not all present and accounted for. The strongest
-->> around (Deep Blue) was not there, not even in an older form like Deep Thought II
-->> or whatever.
-->
-->It maybe is another (and senseless) discussion, but it's rather debatable if
-->Deep Blue is really the strongest around. For example, in the magazine MegaByte
-->Jeroen Noomen examined the first Deep Blue - Kasparov game (1-0). 7 PC programs
-->examined for 3 minutes about each move Deep Blue played. 25 moves (after the opening)
-->were examined. As well Rebel7 as Chess Genius3 played in 19 cases exactly the same
-->move as Deep Blue did. With human eyes examined the 6 different moves weren't worse.
-->The conclusion from Jeroen Noomen: The better PC programs undoubted should have
-->beaten Kasparov as well.
-->
-->Of course I don't want to belittle Deep Blue, it's a real beast... But I wonder
-->what happens when Ed Schroder gets 200.000.000 positions per second... ;=)
-->
-->
-->Rene Vermaas.
Impossible to know, but the point is, right now, nobody does. After the
first round, someone reported that Crafty played every move Deep Blue did
except for (I think) two or three. I responded that it might well be that
those two or three moves were the very ones that were critical. One example
was the Rc6 vs Rc7 debate, where all the comps liked Rc7, yet DB played
Rc6 and it turned out to have a threat that helped to parry the mate in 1
that Kasparov had a few moves later.
In any case, you might take a poll of the programmers to find out if anyone's
willing to step forward and say they can take DB in a match of (say) 6 games.
Cray Blitz searches 1-3M nodes per sec on the latest T90 from Cray, yet that
is still some 100-300 time slower, which translates to a measly 3-4 plies
less search. To get a real feel for what this difference will do, take
Crafty, let it search 1 min/move on one machine, and maybe 60 mins/move on
another (which is still orders of magnitude less than Deep Blue of course).
The difference will astound you.
I regularly play matches between Crafty on the Sparc-20 (30K nodes per
second) and Crafty on my sparc-2 (4-5K nodes per second). The result
is overwhelming. When I examine the games, there's always some tactical
thing the slow one didn't see. Yes, the slow one will on rare occasions
win or draw, but it's in the type of positions where tactics simply don't
count for very much. An notice this is not a putdown of Rebel, Genius,
Mchess Pro, Fritz, or even Crafty. It's just that we are all so far down
on the horsepower curve from DB, that it's unlikely that any of us will
ever beat the thing.
Note also, that I don't believe that none of us can not do as well or better
against humans, because the evaluation plays a big part here. It's very
likely we could find a case where (one day) human X loses to Computer Y
nearly every time, human X beats DB most of the time, and DB beats computer
Y every game. Playing against a computer or playing against a human are two
completely different issues. You *know* that the human won't see every
tactical thing that occurs, while you know that DB not only sees everything
your program see, it sees it 3-5 moves before your program sees it. That's
tough to overcome. Of course, there's nothing to prevent the deep blue guys
from coming up with a killer evaluation over the next few years and slamming
this door in our face too. :)
Bob
Feng-Hsiung Hsu
Robert Hyatt <hy...@cis.uab.edu> wrote:
>those two or three moves were the very ones that were critical. One example
>was the Rc6 vs Rc7 debate, where all the comps liked Rc7, yet DB played
>Rc6 and it turned out to have a threat that helped to parry the mate in 1
>that Kasparov had a few moves later.
Didn't know that there was a debate? Bob apparently swapped the moves.
DB played Rc7, not Rc6. There is a flaw with Rc6. Any guess to what
black's reply to Rc6 should be? Hint: counter play.
Komputer Korner
>
> In article <4l7e5c$1...@europa.frii.com>,
> Tom Kerrigan <kerr...@frii.com> wrote:
> >After one of my games at the most recent WMCCC I had a conversation with a fellow
> >about AEGON. I gave him my address, phone number, etc. and he told me he would be
> >in touch. I noticed that he was having similar conversations with several other
> >programmers. At this point, Bruce was trying hard to not starve to death during
> >one of his longer games. The fellow observed the game for a few minutes, and went
> >off in search of another programmer. I can only assume this is why Bruce wasn't
> >invited while I was, and I can only laugh at speculation on this non-invitation.
> >
> >Regarding science: I have explored the AEGON web site in some depth, and evidently
> >I missed the part about it being scientific. For some reason, I just couln't find
> >a purpose, hypothesis, procedure, or conclusion. Perhaps Hyatt would be so kind as
> >to tell us all what these are?
> >
> >Cheers,
> >Tom
> >
>
>_______________________________________________________________________________
> >Tom Kerrigan kerr...@frii.com O-
> >
> >All true wisdom is found on T-shirts.
>
> Easy:
>
> 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
I agree with Robert, but there is one good thing about Aegon. It remains the
only tournament between computers and humans that is played at slow time
controls. BTW, what were the time controls at Aegon?
Komputer Korner
Robert Hyatt
I'm very good at swapping things at times. In any case, as I was observing,
we got lots of kibitzed moves from various programs and all were expecting
Rc6. I don't remember the position now, but it seemed to be threatening
blacks queen? However, the rook where it ended up was critical later on
because no matter where the black king moved, it ended up in a discovered check
from something... :)
sorry for the error... however that might have been a turning point in the
game should the other move be played...
Bob
Robert Hyatt
Komputer Korner <kor...@iosphere.net> wrote:
>
>I agree with Robert, but there is one good thing about Aegon. It remains the
>only tournament between computers and humans that is played at slow time
>controls. BTW, what were the time controls at Aegon?
> Komputer Korner
As I said, I still think the tournament is a good idea, not only for the
reason you gave but for others (including personal contact between the
programmers). However, it'd be nice were it somewhat more selective so
that year-to-year results can be compared in a meaningful way. In the
States, nearly everything is now (NC) so such events are really needed,
both here and around the world. I agree about the Harvard Cup, and it's
unlikely Crafty will ever play, because there's lots of things I can do with
$1,000 bucks. Were I looking for "advertising" that might well be a good way
to get it, and that's a big plus for Aegon. If there were dozens of these
events in a given year, how they are run would not be much of an issue because
an "average" for the year would be revealing. With only one "non-commerical"
interests tournament (AEGON) it'd simply be nice if it provided more info.
However, it's *not* a big deal of course. "he who pays the bills makes the
rules" and there's nothing wrong with that at all. If they want a 3 round
event, seems they should be able to do that.
Jose Lopez Jr.
Maybe their logic was since the best chess players were not there
then the best chess playing computers should not be present too.
Anyway! Basically, do you think Crafty is better than Quest?
If so, wished it were there!
Rene Vermaas
> I agree with Robert, but there is one good thing about Aegon. It remains the
> only tournament between computers and humans that is played at slow time
> controls. BTW, what were the time controls at Aegon?
Start time 90 minutes + 20 seconds/each move (Fischer clock)
Rene.
Robert Hyatt
Jose Lopez Jr. <jlo...@ix.netcom.com> wrote:
>
> Maybe their logic was since the best chess players were not there
>then the best chess playing computers should not be present too.
> Anyway! Basically, do you think Crafty is better than Quest?
>If so, wished it were there!
No I don't, although I'll call it by it's real name Fritz. :) In any
case, I'm a big fan of Fritz on the tactical front. It's fast, searches
deeply, and plays pretty well. There's lots of things positionally that
Crafty does much better, but as I once saw in a book, "thank goodness,
that before the endgame, the good Lord placed the middle-game." This is
where Fritz really "shines". I play a good number of games against Fritz,
both on the servers and offline during my testing. On equal machines, it
wins at least 2/3 of the games (relatively short time controls) although
in nearly every game I'd evaluate Crafty's position as winning should it
avoid the tactics that Fritz seems to come up with in every game. However
it's difficult to win an ending if you don't get there. :) The very
open style of play I've built into Crafty is often it's undoing against
other programs. If you are playing against someone searching deeper,
you ought to keep things "calm" rather than creating open lines... However,
I've been working on the "human angle" which is not the way to beat other
computers unless you are searching at least as deep as they are. Remember
that Crafty does very poorly on the CCR rating test, yet has stayed around
2600 on ICC playing IM's, GM's and FM's at all sorts of time controls.
It simply plays into positions where it does well against humans because
it spots tactics they overlook. Were Fritz better at this it would really
be something against humans... :)
Once I get Crafty better tuned, I'll certainly play some longer games, as
that should tend to push tactics back a little. At present, however, I
don't pretend to claim superiority over any program, it's much too early
for that. I certainly plan on reaching that point at some future date
of course. :)
Robert Hyatt
Rene Vermaas <rver...@globalxs.nl> wrote:
-->Komputer Korner wrote:
-->
-->> I agree with Robert, but there is one good thing about Aegon. It remains the
-->> only tournament between computers and humans that is played at slow time
-->> controls. BTW, what were the time controls at Aegon?
-->
-->Start time 90 minutes + 20 seconds/each move (Fischer clock)
-->
-->
-->Rene.
I've never understood the Harvard Cup's fascination with short time-
control games, unless it's an attempt to equalize the playing field
since shorter time controls favor machines. Of course, the end result
is to produce a rating that has a * beside it (*=action time control
or whatever) which clouds the issue further...
Komputer Korner
Can anyone tell me what the time controls were for Aegon?
Komputer Korner
Rene Vermaas
I agree... But watch your enormous quote. Please, there's already enough
data-traffic on the internet...
Rene Vermaas.
Bill Newton
>I agree... But watch your enormous quote. Please, there's already enough
>data-traffic on the internet...
>Rene Vermaas.
I'm sure you mean well,but havn't you just repeated an enormous quote
in 'your' post?
Regards
Bill.
MZP
Rene Vermaas
Bill Newton wrote:
> >I agree... But watch your enormous quote. Please, there's already enough
> >data-traffic on the internet...
> I'm sure you mean well,but havn't you just repeated an enormous quote
> in 'your' post?
Sure, but I did it just one time to demonstrate what I meant.
I always spend a lot of time in stripping quotes. I frown my eyebrows
when I read a message with 4 pages quotes and just a '2 sentence' answer...
Rene.
Bruce Moreland
In article <8298572...@cpsoft.demon.co.uk>, chr...@cpsoft.demon.co.uk
says...
>I have it on good authority that you were invited but didn't respond.
>AEGON does have a formal invite/acceptance proceedure, but you must
>understand that their problem is not with computer entries, but with
>getting 50+ reliable humans of appropriate strength.
I'm not sure what the deal is, either my news reader is periodically
marking this as "unread", or I've seem something similar to this several
times.
I am confused as to how you can say what you are saying in the snippet
above. You don't say what the circumstances are, so I'll guess.
I assume that de Gorter or someone else told you that de Gorter invited
Hyatt to AEGON. Hyatt says that he wasn't invited.
At this point you conclude that de Gorter really did invite Hyatt, but
that Hyatt is lying about it.
This is not a reasonable conclusion. It makes much more sense that the
invitation got lost in the mail somewhere, or there is miscommunication
about what an "invitation" is.
de Gorter spoke with me about AEGON last October, but I received nothing
in the mail. If you propose some "American" conspiracy that suggests that
both Hyatt and I are lying about this, I'll know how to regard all of your
future posts, Mr. Whittington.
bruce
Dap Hartmann
brucemo (Bruce Moreland) wrote:
>bruce
I haven't read any of the previous postings on this subject; this is
the first and only one I have read.
But I can supply some additional information.
On Jan 16, 1996 I wrote to Bob Hyatt:
"Maybe you should enter it [Crafty] for the AEGON tournament...
No need to show up there; they'll just have some operator play it on
a Pentium. last year they all got Pentium 90 MHz machines.
The AEGON tournament *is* a very, very nice tournament.
So, I guess you should consider it."
Of course, 'entering' the tournament is not as easy as that. But if
you don't tell the organizers that you're interested, it's difficult
for them to guess you want to play.
Two days later, after Bob hadn't responded to my suggestion in the
email he sent me, I repeated the question:
"So what about Aegon? You wanna play there? I can get you in contact
with the organizers. If you provide a list of 'victims', I'm sure
they'll want Crafty to play, and they will probably find an operator".
This time, Bob responded:
"If the Aegon guys are interested, give 'em my email. I can help the
get it up and running, as it needs the large opening book, and about
250megs for the endgame databases it uses. Not difficult to set up
however, as there have been several thousand down-loads already".
I told him:
"I'll get in contact with the AEGON people.
Thing is: they live in the middle ages, so email is something too new
for them. But I will contact Peter Kouwenhoven, and have him phone
AEGON (Peter is still on the board of the CSVN).
If you want, I can ask them if an Alpha ia possible to arrange.
If 64-bit code is really much faster. Hiarcs played an a Sparc
station last year; don't know how they got it, or if they
brought it themselves".
So I asked Peter Kouwenhoven to ask de Gorter to consider inviting
Hyatt with Crafty.
That's the last *I* heard about it, and I guess the last that Bob
heard about it too.
C'est la vie
dap
George Dixon
Certainly doesn't seem like your fault but definitely was Aegon's loss.
George
Chris Whittington
brucemo (Bruce Moreland) wrote:
>
> In article <8298572...@cpsoft.demon.co.uk>, chr...@cpsoft.demon.co.uk
> says...
>
> >I have it on good authority that you were invited but didn't respond.
> >AEGON does have a formal invite/acceptance proceedure, but you must
> >understand that their problem is not with computer entries, but with
> >getting 50+ reliable humans of appropriate strength.
>
> I'm not sure what the deal is, either my news reader is periodically
> marking this as "unread", or I've seem something similar to this several
> times.
>
> I am confused as to how you can say what you are saying in the snippet
> above. You don't say what the circumstances are, so I'll guess.
Dangerous game - guessing, but then we all do it ...
>
> I assume that de Gorter or someone else told you that de Gorter invited
> Hyatt to AEGON. Hyatt says that he wasn't invited.
Correct.
>
> At this point you conclude that de Gorter really did invite Hyatt, but
> that Hyatt is lying about it.
>
> This is not a reasonable conclusion. It makes much more sense that the
> invitation got lost in the mail somewhere, or there is miscommunication
> about what an "invitation" is.
You're erecting a false tower to shoot at.
There are several permutations of the event ranging from everybody
lying to everybody telling the truth with various levels of
communication failure in between.
Because I've spoken directly to de Gorter and registered his hurt at
(what I feel are) unwarranted and thoughtless attacks - I've
countered Hyatt's assertions. Its called the adversarial principle
- often useful for arriving at the truth.
If two sides to one story can't agree then its open for bystanders
to make up their own minds.
Since I've not registered any conclusion, about lying or otherwise,
you are spinning the falsity.
Right now I don't have a conclusion, since I don't know (and stopped
thinking about it several weeks ago), but it is unlikely to be that
Hyatt is lying.
>
> de Gorter spoke with me about AEGON last October, but I received nothing
> in the mail. If you propose some "American" conspiracy that suggests that
> both Hyatt and I are lying about this, I'll know how to regard all of your
> future posts, Mr. Whittington.
May I paraphrase ?
if (jumping to conclusions)
{
challenge(pistols at dawn);
}
Come on.
1. Didn't think I ever mentioned you.
2. Never used word 'lying' or any variant thereof.
3. Challenging assertion 'i wasn't not invited' with counter-assertion
'i was told you were invited' - does not imply someone is lying, although
such a conclusion is possible, as is the equally valid conclusion of
communication failure.
4. Further: this is chicken and egg. Who denied what by saying what first ?
This thread has built with assertion and counter assertion over
several weeks, it could well be possible to fling 'lying' mud in
several directions simultaneously.
If you read more carefully you'll see the real meaning behind my original
post that you quote. Here's your quote again:
> >I have it on good authority that you were invited but didn't respond.
> >AEGON does have a formal invite/acceptance proceedure, but you must
> >understand that their problem is not with computer entries, but with
> >getting 50+ reliable humans of appropriate strength.
>
It is that de Gorter has no problem finding computer entries, but instead
finds it difficult to get 50+ reliable humans.
The implication was that, if you want to enter a program, it is up to
you to chase de Gorter, he'll ask once and from then on its up to you.
If you go back to the original thread - you'll see it started from the
posting of Hyatt "I don't see the purpose of AEGON anyway" or somesuch.
This throwaway line got several people's backs up, including mine.
This was the unwarranted and thoughtless initiation.
But, with hindsight, perhaps if Hyatt (and your goodself ?) can't see the
purpose of chess-human tourneys in Europe, ten thousand miles away from
the centre of the known universe, its no big deal :)
Chris Whittington
>
> bruce
>
Chris Whittington
Komputer Korner <kor...@iosphere.net> wrote:
>
>
> Can anyone tell me what the time controls were for Aegon?
> Komputer Korner
1 hour 30 minutes + 20 seconds Fischer clock (I think)
Chris Whittington