Deep fritz loses again !
james liggett
0.5
and a win with black at that!
If this continues deep fritz won't be selling and might even be
having a recall for deep loser.
I can hear IBM laughing and saying what a pretender.
The question is how many clowns paid for this loser only to find out
it can't win?
The fritz team is going to be the laughing stock of the chess
programs!
Can the fritz team find a job after this blow out.
Will deep fritz win 1 game? Dam sure don't look like it can.
Kramnik is calling his banker telling him get ready for a million
dollar deposit.
What's that I hear warming up in the background........ sounds like
the fat lady!
tinklemagoo
doesnt seem to be able to make plans or understand dynamics in game
positions. And it seems obsessed with doubling pawns, which Kramnik dealt
with very easily.
Johannes Fürnkranz
Kramnik plays great anti-computer chess. It ain't fun to watch, but it
even worked against Kasparov. After seen these games, I'm quite
convinced that Deep Blue wouldn't stand a chance against Kramnik.
Unfortunately, we'll never know...
Juffi
dph
news:anv9m9$95b$07$1...@news.t-online.com...
At least the Fritz people are up for a good challenge.
Cheers to them.
I am sure what they will take what they learned to guide
future development efforts--and the world of computer chess
will be stronger for their efforts.
Regards,
David
Robert M. Hyatt
> Regards,
> David
The sad thing is, had they spent a lot of time on the various
chess servers, they would have seen this problem _many_ times
in playing the various GM players that are playing online. It
is a serious problem, and the "draw-masters" have developed a
style of play (against computers) that requires a _lot_ of
effort to stymie...
Contrary to popular opinion, chess is not yet "solved" although
the programs are definitely playing very strongly. Just not
2700+ strongly...
--
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
Katy and Tracy Miller
he wins, $600K if he loses??? Sounds like a kind of no-lose proposition for
Kramnik. How about $2 million if he wins, nothing (just expenses) if he
loses. Now THAT'S a challenge.
And why did they name this program Deep Fritz? It's clearly a rip-off
of Deep Blue. IBM should be upset by this. People only casually interested
in this match will confuse the two programs. I've already heard one of my
friends asking why Kramnik is having such an easy time against the computer
that beat Kasparov.
"james liggett" <redzo...@att.net> wrote in message
news:acd2ca8c.0210...@posting.google.com...
John Swartz
> And why did they name this program Deep Fritz? It's clearly a rip-off
> of Deep Blue. IBM should be upset by this.
And "Deep Blue" is a newer version of "Deep Thought" - wonder if Linda
Lovelace is upset by that...
Katy and Tracy Miller
"John Swartz" <jsw...@mitre.org> wrote in message
news:3DA34D61...@mitre.org...
Alberich
@att.net says...
Laugh all you want. Just remember, these software titles will only get
stronger, not weaker. Eventually the software will improve and WILL
defeat the World Champion someday. If memory serves, I think Kramnik
himself says within 20 years the computer software for chess will be
almost impossible to defeat. If these software titles weren't so good,
why did Kramnik use it to analyze the game played by Kasparov/Deep Blue?
Of course I knew Kramnik was going to blow Deep Fritz out of the water.
Even Kasparov said the same thing. But that's not the point. The very
fact that a Kramnik or Kasparov would even bother playing a computer or
software for astonishing amounts of money shows that artificial
intelligence is getting stronger and stronger. Anti computer chess will
work for the time being...untl the software makers redesign the software
to ruthlessly destroy the anti computer chess playing styles employed by
the GMs. When that happens, I doubt you'll be laughing your head off.
The Borg were slow and plodding, but they learn from their mistakes and
only got stronger.
This embarrassment for the Chessbase team will toughen it up and force
them to re-evaluate the code. When that happens, watch out. No more
software doubling up pawn chains, exchanging Queens early and getting
"lost" after book moves.
Hmmmm. I was wondering whether Deep Fritz was even using an Opening Book
repetiore. If Fritz was using an actual opening book by say Kramnik
himself, this contest would be alot different. Because then Fritz could
"learn" from the Kramnik database and make better moves.
camelx
(sorry, couldn't help myself)
"james liggett" <redzo...@att.net> wrote in message
news:acd2ca8c.0210...@posting.google.com...
PSUGorilla
ChessBase dealer within a few months. :)
"james liggett" <redzo...@att.net> wrote in message
news:acd2ca8c.0210...@posting.google.com...
Robert M. Hyatt
> How on earth did Kramnik manage to get a match where he gets $1.0 million if
> he wins, $600K if he loses??? Sounds like a kind of no-lose proposition for
> Kramnik. How about $2 million if he wins, nothing (just expenses) if he
> loses. Now THAT'S a challenge.
> And why did they name this program Deep Fritz? It's clearly a rip-off
> of Deep Blue. IBM should be upset by this. People only casually interested
> in this match will confuse the two programs. I've already heard one of my
> friends asking why Kramnik is having such an easy time against the computer
> that beat Kasparov.
It was started by "deep junior". When they did their parallel version.
They got into some sort of flap with IBM because IBM liked to call the
"baby" version of deep blue "deep blue junior". They decided to return
the favor and call their parallel program "deep junior". The "deep"
title stuck and most everyone has been using that term to indicate the
parallel search version of their program (a few did not, ie there is no
deep crafty, or deep ferret, for example)...
> "james liggett" <redzo...@att.net> wrote in message
> news:acd2ca8c.0210...@posting.google.com...
>> Another win by the WC and it's starting to look like a blowout. 2.5 -
>> 0.5
>> and a win with black at that!
>> If this continues deep fritz won't be selling and might even be
>> having a recall for deep loser.
>>
>> I can hear IBM laughing and saying what a pretender.
>>
>> The question is how many clowns paid for this loser only to find out
>> it can't win?
>> The fritz team is going to be the laughing stock of the chess
>> programs!
>> Can the fritz team find a job after this blow out.
>>
>> Will deep fritz win 1 game? Dam sure don't look like it can.
>> Kramnik is calling his banker telling him get ready for a million
>> dollar deposit.
>> What's that I hear warming up in the background........ sounds like
>> the fat lady!
--
mark.houlsby
Unlikely, what with her being dead and all... the executors of her estate
might be, however...
M
Kym
Kramnik the gets a slight positional/strategic advantage and applies a
torturous slow crush; meantime DF has NO plan and make time wasting moves.
DF has no endgame. Too many pawns and not enough EGTB :-)
It is not that simple for Kramnik: he took something like 30 minutes on one
move, so he must seems to be very careful.
DF will still crush most players, but the 'best' human play wins.
BTW: What is DF performance rating at 2.5-0.5? about 2500?
If the result is 6-2 then the PF is about 2600?
If worst case 7.5-0.5, then about 2300.
Seems like a marketing disaster for ChessBase.
They will be hoping for at least 1 win for DF, and that will *ONLY* happen
if Kramnik makes a middle game tactical mistake, which he will play to
avoid.
Go Vlad Go!
"james liggett" <redzo...@att.net> wrote in message
news:acd2ca8c.0210...@posting.google.com...
Steve
Uh I think you are confused. It's called Deep Thought NOT Deep Throat!
Douglas Adams has nothing to do with Linda Lovelace.... Sorry but your joke
was silly!
Anyway it didn't play chess
;)
"John Swartz" <jsw...@mitre.org> wrote in message
news:3DA34D61...@mitre.org...
>
Steve
understanding. I remember back in college reading a book in the late 80s
written by a grandmaster and he said a computer would never beat a human.
He figured it'd have to analyze all 20 moves deep into the game. Well many
who had a strong mathematical background understood that he was wrong and
challenged.
I think the answer is apparent. It will not be the software makers alone
but much comes fromo the microcomputer advancement as well.
Read the articles reading computer ratings vs depth. If the PC loses, fine
but the advancement over the past 10yrs is far beyond what many ppl ever
thought.
You can extrapolate and get a general idea that if the PC isn't good enough
give it another 10yrs or less and the PC will be well beyond 3000 elo
rating.
I think the next chess champion (human) will have an elo rating
unprecedented. He/She will achieve a score of 3500 by 2010 and it'll mostly
be attributed to PCs that have score in the range of 3200 elo. Humans will
learn and better challenged.
Of course that's my opinion, but wait for 2010 :)
"Alberich" <Albe...@somewhere.com> wrote in message
news:MPG.180d2a0aa...@news.cyburban.com...
Steve
"PSUGorilla" <an...@anon.com> wrote in message
news:OnKo9.639$XF....@news1.central.cox.net...
Steve
to exploit them. That was partly why DF lost.
Also is it really fair that Kramnik has time to practice with DF while DF
team doesn't have time to practice with Kramnik? That seems to me like
Kramnik had an advantage to find a flaw.
Imagine if you were able to play a human and the human opponent wasn't able
to learn from you. You found the human's errors and exploited them without
the human opponent having an equal chance.
In other words, perhaps DF isn't nearly as bad as it appears if Kramnik
feels the necessity to learn from the PC but not the other way around.
"Kym" <send...@to.nul.ru> wrote in message
news:3da3...@duster.adelaide.on.net...
mark.houlsby
_months_ to practise against DF, whereas DF, not having played Kramnik's
anti-computer style, couldn't practise against him at all...
Moron.
<plonk>
TheDoDo <the...@spamfree.org> wrote in message
news:1213915.T...@thedodo.net...
> Steve from scp...@hotmail.com on Tuesday 08 October 2002 09:01 pm
> enscribed:
>
> > I agree a little. However Kramnik did make some errors but DF wasn't
able
> > to exploit them. That was partly why DF lost.
> >
> > Also is it really fair that Kramnik has time to practice with DF while
DF
> > team doesn't have time to practice with Kramnik? That seems to me like
> > Kramnik had an advantage to find a flaw.
>
>
> Are you a moron? DF has an entire DATABASE of every game Kramnik has ever
> played. On the other hand, Kramnik had not a single DF game to study.
KRCH
> Kramnik adopts an entirely different approach when playing programs. He had
> _months_ to practise against DF, whereas DF, not having played Kramnik's
> anti-computer style, couldn't practise against him at all...
>
> Moron.
> <plonk>
> TheDoDo <the...@spamfree.org> wrote in message
> news:1213915.T...@thedodo.net...
> > Steve from scp...@hotmail.com on Tuesday 08 October 2002 09:01 pm
> > enscribed:
> >
> > > I agree a little. However Kramnik did make some errors but DF wasn't
> able
> > > to exploit them. That was partly why DF lost.
> > >
> > > Also is it really fair that Kramnik has time to practice with DF while
> DF
> > > team doesn't have time to practice with Kramnik? That seems to me like
> > > Kramnik had an advantage to find a flaw.
> >
> >
This programm play about 2300 as maximum.Do not need to be Kramnik to beat her.
Johannes Fürnkranz
> I agree a little. However Kramnik did make some errors but DF wasn't able
> to exploit them. That was partly why DF lost.
>
> Also is it really fair that Kramnik has time to practice with DF while DF
> team doesn't have time to practice with Kramnik? That seems to me like
> Kramnik had an advantage to find a flaw.
But there's a database of hundreds of Kramnik's game that DF surely
knows by heart. Problem is, it can't make much sense out of them.
I was surprised that they let DF into that Berlin wall that already
crumbled Kasparov. That's not a line where a computer has any chance of
winning. Likewise, trading off queens for a double pawn like in game 3
is no line for a computer. It doesn't know how to line up a positional
game against a double pawn anyways. So why go for it?
IMHO this match so far shows lousy preparation on the DF side (they
didn't have Joel Benjamin, I guess). I still think with this style of
play Kramnik would win anyways, but they could make it a bit tougher.
They should at least try with white...
Juffi
Johannes Fürnkranz
> Kramnik adopts an entirely different approach when playing programs. He had
> _months_ to practise against DF, whereas DF, not having played Kramnik's
> anti-computer style, couldn't practise against him at all...
Nope. He played the same type of lines that already brought Kasparov to
tears. A little opening preparation would have sufficed to at least not
get into the Berlin wall, where white has absolutely no chance of
winning against Kramnik.
Juffi
crot...@networksplus.net
Is everyone forgetting that Kramnik has had access to the program. He
isnt playing anything that he hasn't already memorized.
Gordon
"Guy Macon" < Guy Macon @ http://www.guymacon.com/home.html > wrote in
message news:6UadneWZnZB...@News.GigaNews.Com...
>
> Steve <scp...@hotmail.com> wrote:
> >
> >I agree a little. However Kramnik did make some errors but DF wasn't
able
> >to exploit them. That was partly why DF lost.
> >
> >Also is it really fair that Kramnik has time to practice with DF while DF
> >team doesn't have time to practice with Kramnik? That seems to me like
> >Kramnik had an advantage to find a flaw.
>
> of books on openings?
No.
This is a human versus computer match. Does the term "human" include books?
No, of course not. Does the term "computer" include a memory subsystem such
as a hard-drive? Yes, it is a common part of a computer. I bet your PC has
a hard-drive, and you have a brain but no "builit in" books as such.
However, there's a big difference between looking at your opponent's games
in advance, compared with actually getting to practice against a copy of the
opponent. Deep Fritz has played match/tournament games prior to this. Why
didn't Kramnik study these games?
Although Kramnik can't determine exactly what openings Deep Fritz will play,
during his practice he could have found out how well he does against it in,
e.g., typical Sicilian positions compared with the Ruy Lopez. He could have
found out which works best as White... e4, d4, etc. Did Deep Fritz know
what it finds most effective against Kramnik prior to the match? No, the DF
team are only now finding out how good or bad the Scotch opening is, etc.
This is unfair.
Gordon
>
>
>
>
> --
> Guy Macon http://www.guymacon.com Guy Macon http://www.guymacon.com
> Guy Macon http://www.guymacon.com Guy Macon http://www.guymacon.com
> Guy Macon http://www.guymacon.com Guy Macon
http://www.guymacon.com
> Guy Macon http://www.guymacon.com Guy Macon
http://www.guymacon.com
>
Kym
just replay a game from memory. What he will have is a *feel* for DF
strengths and weaknesses, ie: good opening + tactics and bad strategy and
EG.
<crot...@networksplus.net> wrote in message
news:3da3f468...@news.networksplus.net...
Gordon
> isnt playing anything that he hasn't already memorized.
I largely agree with this (with some exceptions).
If I've understood the match conditions correctly, Kramnik got to practice
against the exact program prior to the match. However, the Deep Fritz team
can alter its opening book to some extent prior to and during the match.
If my assumptions are wrong, I apologise... please correct me and ignore the
rest... :-)
So, given this, I don't think Kramnik can reliably memorise whole games due
to the unpredicatable variation that can happen during the opening moves. I
think there would be too many possibilities to try to pre-empt.
However, the Deep Fritz team are very unlikely to invent any new openings!
:-) They may change the odd move here and there, but if Kramnik during his
practice has found out, for example, that DF doesn't "understand" typical
plans in the Ruy Lopez, and that he can generally win easily... then the DF
team need to avoid this opening all together, and not just tweak the actual
moves that DF chooses as part of it.
If Kramnik finds out how best to play in typical Ruy Lopez positions,
typical Sicilian, etc. this has to be a big advantage. Is it better to aim
for the ending or not? He doesn't need to guess this... he can base it on
actual results! How well does my opponent play rooks endings? pawn endings?
opposite bishops? etc, etc... No need to guess... just setup typical
positions, try it out and note actual results!
I don't think Kramnik should have been given the exact program... just
what's commercially available to us all. And the DF team should have been
allowed to completely change it's opening book between games, and modify the
program between games if necessary. But no updates during games.
Gordon
mark.houlsby
DF would _avoid_ the Berlin with what, exactly? The Vienna? Ponziani's? Do
you think Kramnik isn't prepared for these and all other white
possibilities?
Think again.
Mark
mark.houlsby
that they couldn't _sell_ it.... DF must compete in the format in which it
sells or the match is pointless, from its programmers point-of-view. That's
why the sponsors were hoping to pay Volodya six figures for losing....
Mark
Johannes Fürnkranz <johannes....@t-online.de> wrote in message
news:ao0pk5$j96$01$1...@news.t-online.com...
PSUGorilla
"Steve" <scp...@hotmail.com> wrote in message
news:Ypidnftat4t...@News.GigaNews.Com...
pc...@delete.this.please.tcp.co.uk
retirement ! ;-)
redzo...@att.net (james liggett) wrote:
>Another win by the WC and it's starting to look like a blowout. 2.5 -
>0.5
>and a win with black at that!
>If this continues deep fritz won't be selling and might even be
>having a recall for deep loser.
--------------------------------------------------------
www.chessandcheckers.com/chess.htm (mine)
www.chessgenius.com (Langsoft official page)
Steve
Really?! I thought I read that Kramnik was playing a version on a 8
processor PC practicing for the tournament and DF were making changes.
Making I got it mixed up with something else. Well it'd not be my first
time ;)
As for DF having a db of every game of Kramnik, I'd think that'd not be
nearly as good if there were games held for practice by both teams. But
again that's only with the assumption that Kramnik was able to study DF.
Someone did comment that Kramnik was good at anti-computer chess. Seems
there is a whole study behind the field.
I think though the fact that there is such a field seems to indicate that
PCs are superior in some sense to the human chess champions. Otherwise such
a line of attack wouldn't be relevant. Just as humans play certain lines of
attack or defense for certain openings. Guess this be the anti-chess line
of attack/defense :)
Guess we learned more from the computer than we realize.
"TheDoDo" <the...@spamfree.org> wrote in message
news:1213915.T...@thedodo.net...
> Steve from scp...@hotmail.com on Tuesday 08 October 2002 09:01 pm
> enscribed:
>
> > I agree a little. However Kramnik did make some errors but DF wasn't
able
> > to exploit them. That was partly why DF lost.
> >
> > Also is it really fair that Kramnik has time to practice with DF while
DF
> > team doesn't have time to practice with Kramnik? That seems to me like
> > Kramnik had an advantage to find a flaw.
>
>
> Are you a moron? DF has an entire DATABASE of every game Kramnik has ever
> played. On the other hand, Kramnik had not a single DF game to study.
>
>
> >
Steve
I couldn't argue with that ;)
"TheDoDo" <the...@spamfree.org> wrote in message
news:1301547.k...@thedodo.net...
> I'll take deep throat over deep thought any day of the week.
>
>
> Steve from scp...@hotmail.com on Tuesday 08 October 2002 08:42 pm
> enscribed:
Steve
had access DF. I don't really think memorizing any games for Kramnik would
make any sense.
The difference is if Kramnik had access to DF, he could find areas to
exploit DFs weakness. Imagine if you are this awesome GM and had a player
that basically played the same and didn't learn anything or sit back and
analyze positions to be avoided later.
So whatever weakness he could find would most likely be present in the real
tournament. The match seems a little one-sided IMHO.
"Steve" <scp...@hotmail.com> wrote in message
news:z6ycndSadrO...@News.GigaNews.Com...
John Swartz
reference (I remember there was another computer called "Deep Thought",
but forgot where I heard of it.
The point was that "Deep Fritz" was seen by a poster as a ripoff of
"Deep Blue"
My point was that "Deep Thought" was at least inspired by the name "Deep
Throat"
Sorry to be silly,
John
Steve
What was the answer again? 47?
"John Swartz" <jsw...@mitre.org> wrote in message
news:3DA43DBA...@mitre.org...
John Swartz
> Someone did comment that Kramnik was good at anti-computer chess. Seems
> there is a whole study behind the field.
Call it what you want, but I see this as simply preparing to play
against an opponent's style. Recall, for example, in the 1972 Fischer
Spassky match when Fischer (who played 1.e4 almost exclusively) opened
several games with 1.c4 Anti-computer chess to me is just another
example of knowing your opponent and trying to throw them off their
game.
John
Lars
> I think it will be much sooner than that. See http://www.qubit.org/
Also my words, Artificial Intelligence will increase much in the future.
http://www.cs.usu.edu/~degaris/artilectwar2.html
-snip-
Let me be more specific. As a professional brain building researcher and former theoretical
physicist, I am in a position to see more clearly than most, the potential of 21st century
technologies to generate "massively intelligent" machines. By "massively intelligent" I mean
the creation of artificial brains which may end up being smarter than human brains by not just
a factor of two or even ten times, but by a factor of trillions of trillions of trillions of
times, i.e. truly godlike.
-snip-
--
Best regards
Lars
Bob Day
"Lars" <l...@get3net.dk> wrote in message
news:MPG.180e8ece6...@news.inet.tele.dk...
Post again when you sober up.
-- Bob Day
Luigi Caselli
news:MPG.180e8ece6...@news.inet.tele.dk...
> -snip-
> Let me be more specific. As a professional brain building researcher and
former theoretical
> physicist, I am in a position to see more clearly than most, the potential
of 21st century
> technologies to generate "massively intelligent" machines. By "massively
intelligent" I mean
> the creation of artificial brains which may end up being smarter than
human brains by not just
> a factor of two or even ten times, but by a factor of trillions of
trillions of trillions of
> times, i.e. truly godlike.
> -snip-
Well, I'm not so sure of this brillant future of artificial brains but in
any case they'll run under Windows 3000 (at least for commercial reasons).
So I don't think they'll be so smart... :-)
Luigi Caselli
Damir Ulovec
Also my words, Artificial Intelligence will increase much in the future.
former theoretical physicist, I am in a position to see more clearly than
most, the potential of 21st century technologies to generate "massively
intelligent" machines. By "massively intelligent" I mean the creation of
artificial brains which may end up being smarter than human brains by not
just a factor of two or even ten times, but by a factor of trillions of
trillions of trillions of times, i.e. truly godlike.
Yes, then (and only from that point) computers will be strongest chess
players than humans. But, there's one more threat... Huuuuuu... Will then
computers (or their heiress-successors) conquer the human world...
Huuuuuuuu... Just joking :o)))
It's no doubt that computer's speed grows 2x every 18+months (19-20). Well,
that means that we will have monster chess computer players after about 4-5
years, which will evaluate 15-20 (or even more) ply's in about 5-10 (or even
less) seconds. By the way, do you remember great computers before 1995?
1993? 486DX5? 486DX4? 486...? 386DX? 386SX (@ 20Mhz, my first :-)?... I know
many peoples that was defeated many times, long time ago, by Fritz 1.2,
Mephisto Genius 2, KnightStalker,... witch "Elo" was about 1900-2100.
Ops, I was type too much words... Sorry for my monotony.
mark.houlsby
>
> > Someone did comment that Kramnik was good at anti-computer chess. Seems
> > there is a whole study behind the field.
>
> Call it what you want, but I see this as simply preparing to play
> against an opponent's style.
<snips some good stuff>
Not quite. My understanding of one particular anti-computer strategy, which
Kramnik has employed with complete success in this match, is to get the
program into an endgame which is not a tablebase, and which, therefore, it
is unable to play. This involves specific strategic decisions which he would
_never_ play against a human opponent, and goes some way towards explaining
his having done so badly in the recent Russia vs. the World match (he was,
we take it, very much in an anti-computer frame of mind).
In other words I am arguing that there is a _qualitative_ difference between
anti-computer strategy and anti-any-given-human-opponent strategy...
Any GM who adopted an anti-computer strategy against a human colleague would
be summarily crushed.
That aside, I agree.
Mark
Johannes Fürnkranz
>
> DF would _avoid_ the Berlin with what, exactly? The Vienna? Ponziani's? Do
I don't know. I would say that's Chessbase's task to decide.
How about a Blackmar Diemer? At least the games would be more fun to
watch. :-)
> you think Kramnik isn't prepared for these and all other white
> possibilities?
Not as much as DF is. Even Kramnik doesn't know ECO by heart.
But in any case, it's completely irrelevant whether Kramnik is prepared
for a line or not. You can't choose lines against Kramnik on the hope
that he ain't prepared for them.
The point is what lines you choose to play. DF is a fierce tactician and
a lousy positional player (lousy for GM standards). If I could play like
that, I wouldn't trade off queens in the opening, and I certainly
wouldn't go into the Berlin after seeing how Kramnik held these
positions against Kasparov.
Juffi
Johannes Fürnkranz
> John Swartz <jsw...@mitre.org> wrote in message
> news:3DA44263...@mitre.org...
>
>>>Someone did comment that Kramnik was good at anti-computer chess. Seems
>>>there is a whole study behind the field.
>>
>>Call it what you want, but I see this as simply preparing to play
>>against an opponent's style.
>
>
> <snips some good stuff>
>
> Not quite. My understanding of one particular anti-computer strategy, which
> Kramnik has employed with complete success in this match, is to get the
> program into an endgame which is not a tablebase, and which, therefore, it
> is unable to play. This involves specific strategic decisions which he would
> _never_ play against a human opponent, and goes some way towards explaining
> his having done so badly in the recent Russia vs. the World match (he was,
> we take it, very much in an anti-computer frame of mind).
Why shouldn't he choose the same strategy against a human that he knows
is bad in endgames?
I haven't seen his games in the Russia vs. the World match, but I agree
that he might have been practicing his strategy against opponents where
the strategy don't fit.
> In other words I am arguing that there is a _qualitative_ difference between
> anti-computer strategy and anti-any-given-human-opponent strategy...
I don't see this.
> Any GM who adopted an anti-computer strategy against a human colleague would
> be summarily crushed.
I don't see much difference between Kramnik's strategy against DF and
Kramnik's strategy against Kasparov, and I think it is very appropriate
in both cases, although for entirely different reasons. In Kasparov's
case, I believe breaking his ego was much more important than holding
these games with black.
Juffi
Johannes Fürnkranz
> See my reply to your other post. Tailoring DF to Kramnik would have meant
> that they couldn't _sell_ it.... DF must compete in the format in which it
> sells or the match is pointless, from its programmers point-of-view. That's
> why the sponsors were hoping to pay Volodya six figures for losing....
Agreed.
Although I am not quite sure why a DF version tailored to Kramnik
wouldn't be right for me, you, or any of us.
Juffi
mark.houlsby
> mark.houlsby wrote:
> >
> > DF would _avoid_ the Berlin with what, exactly? The Vienna? Ponziani's?
Do
>
> I don't know. I would say that's Chessbase's task to decide.
> How about a Blackmar Diemer? At least the games would be more fun to
> watch. :-)
If it played the BDG it'd get killed... I suspect you were joking.
>
> > you think Kramnik isn't prepared for these and all other white
> > possibilities?
>
> Not as much as DF is. Even Kramnik doesn't know ECO by heart.
>
No, but he knows which lines of which openings will lead DF into endgames it
can't play, and I'd be very surprised if he hasn't prepared for _all of
them_
> But in any case, it's completely irrelevant whether Kramnik is prepared
> for a line or not. You can't choose lines against Kramnik on the hope
> that he ain't prepared for them.
>
True.
> The point is what lines you choose to play. DF is a fierce tactician and
> a lousy positional player (lousy for GM standards). If I could play like
> that, I wouldn't trade off queens in the opening, and I certainly
> wouldn't go into the Berlin after seeing how Kramnik held these
> positions against Kasparov.
>
> Juffi
>
My question remains: How does DF _avoid_ the Berlin, given that Kramnik will
reply 1...e5 and 2...Nc6...?
Other white choices are clearly inferior from DF's point-of-view
Mark
mark.houlsby
Umm I'd guess because we don't play like he does. Relatively few people open
1.Nf3. Almost nobody can play the middlegames resulting from his repertoire
the way Volodya can.
Mark
mark.houlsby
> mark.houlsby wrote:
> > John Swartz <jsw...@mitre.org> wrote in message
> > news:3DA44263...@mitre.org...
> >
> >>>Someone did comment that Kramnik was good at anti-computer chess.
Seems
> >>>there is a whole study behind the field.
> >>
> >>Call it what you want, but I see this as simply preparing to play
> >>against an opponent's style.
> >
> >
> > <snips some good stuff>
> >
> > Not quite. My understanding of one particular anti-computer strategy,
which
> > Kramnik has employed with complete success in this match, is to get the
> > program into an endgame which is not a tablebase, and which, therefore,
it
> > is unable to play. This involves specific strategic decisions which he
would
> > _never_ play against a human opponent, and goes some way towards
explaining
> > his having done so badly in the recent Russia vs. the World match (he
was,
> > we take it, very much in an anti-computer frame of mind).
>
> Why shouldn't he choose the same strategy against a human that he knows
> is bad in endgames?
>
This is simple to answer. Simultaneous exhibitions aside, he never gets to
_play_ any humans who handle the endgame anything like as badly as DF
does...
> I haven't seen his games in the Russia vs. the World match, but I agree
> that he might have been practicing his strategy against opponents where
> the strategy don't fit.
>
> > In other words I am arguing that there is a _qualitative_ difference
between
> > anti-computer strategy and anti-any-given-human-opponent strategy...
>
> I don't see this.
>
Maybe because you still haven't grasped how lousy (by Kramnik's standards)
programs are at playing endgames with more than 5 pieces on the board...?
> > Any GM who adopted an anti-computer strategy against a human colleague
would
> > be summarily crushed.
>
> I don't see much difference between Kramnik's strategy against DF and
> Kramnik's strategy against Kasparov, and I think it is very appropriate
> in both cases, although for entirely different reasons. In Kasparov's
> case, I believe breaking his ego was much more important than holding
> these games with black.
>
Against Kasparov, Kramnik won the match because he was better prepared, just
as is happening with DF. He chose to play the Berlin against GKK precisely
because of its _being resistant to computer analysis_, thereby ensuring that
Kasparov and his seconds would have many fruitless, sleepless nights...
Certainly by choosing _that line_ with white against GKK's Gruenfeld,
Kramnik succeeded in outpsyching him.....
Mark
> Juffi
>
Robert M. Hyatt
> mark.houlsby <mark.h...@ntlworld.com> wrote:
>>...how lousy (by Kramnik's standards) programs are at playing
>>endgames with more than 5 pieces on the board...
> Makes you wonder why the chess program vendors haven't made a
> greater effort to produce 6/7/8/etc. piece tablesbases, doesn't it?
No. 7-8 piece tables are _impossible_ at the moment. When we expect
all the 6's to take a terrabyte (or more) the 7's will take roughly
64 times that much and probably more because of the bits needed for
larger mate scores. 8's will take another factor of 64 (roughly).
We have done _all_ the pawnless 6 piece files except for the 5 vs 1
cases. Eugene is getting ready to produce the ones with pawns and
we will be done with the 6's totally...
--
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
Robert M. Hyatt
> Robert M. Hyatt <hy...@crafty.cis.uab.edu> wrote:
>>
>>Guy Macon < Guy Macon @ http://www.guymacon.com/home.html > wrote:
>>
>>> mark.houlsby <mark.h...@ntlworld.com> wrote:
>>
>>>>...how lousy (by Kramnik's standards) programs are at playing
>>>>endgames with more than 5 pieces on the board...
>>
>>> Makes you wonder why the chess program vendors haven't made a
>>> greater effort to produce 6/7/8/etc. piece tablesbases, doesn't it?
>>
>>No. 7-8 piece tables are _impossible_ at the moment. When we expect
>>all the 6's to take a terrabyte (or more) the 7's will take roughly
>>64 times that much and probably more because of the bits needed for
>>larger mate scores. 8's will take another factor of 64 (roughly).
>>
>>We have done _all_ the pawnless 6 piece files except for the 5 vs 1
>>cases. Eugene is getting ready to produce the ones with pawns and
>>we will be done with the 6's totally...
> That implies that you *only* need 4e+26 terrabytes for a 32 piece
> table. I expect daily status reports on you progress... ;)
> A 200-disc DVD-ROM Changer costs around $1000 nowdays. Not an
> unreasonable addition to a computer that's playing for a million
> dollar prize...
Can you _imagine_ the NPS of a program probing those in the search?
1nps? :)
> I realize the advantage of having a comlete TB, but would it be
> worthwhile for a program to have partial 7 or 8 piece tables?
> If so, how much storage would it take for only those positions
> where the material is even? How much more to cover one side
> or the other being a pawn up?
We have some single 6's that are > 4 gigabytes. So some single
sevens will push 250 gigs. Single eights will push 16 terrabytes
which I don't want to think about. :)
pc...@delete.this.please.tcp.co.uk
But I suppose the interesting aspect of question is the
rate of progress. The increase in processor speeds,
and hd capacity, despite fears, have not slowed down.
We could calculate the approximate completion dates
for the 4,5,6 pc databases, the increase in capacity/processing
required by each new set, and hence try to predict when we might
have the 7,8.. pc databases.. :-)
Anyone?
>--
>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
ade
-----------------------------------------------------------
www.chessandcheckers.com/chess.htm (my chessgenius page)
www.chessgenius.com (official chessgenius page)
Simon Waters
>
> But I suppose the interesting aspect of question is the
> rate of progress. The increase in processor speeds,
> and hd capacity, despite fears, have not slowed down.
> We could calculate the approximate completion dates
> for the 4,5,6 pc databases, the increase in capacity/processing
> required by each new set, and hence try to predict when we might
> have the 7,8.. pc databases.. :-)
> Anyone?
I think the power to create and store 6's and possibly 7's is
around, you only need look at the recent supercomputers with
tens or hundreds of Giga bytes of RAM per processors, although
these probably have more important things to do.
Personally my bandwidth (last mile) has only increased 160 fold
in 20 years, at which rate it may well take me several days to
download one 7 piece table in 2025.
Although based on my personal processor speed improvements I
probably won't need them as the machine will just work out the
result for itself.
pc...@delete.this.please.tcp.co.uk
>pc...@delete.This.Please.tcp.co.uk <pc...@delete.This.Please.tcp.co.uk> wrote:
>>>>>We have done _all_ the pawnless 6 piece files except for the 5 vs 1
>>>>>cases. Eugene is getting ready to produce the ones with pawns and
>>>>>we will be done with the 6's totally...
>>But I suppose the interesting aspect of question is the
>>rate of progress. The increase in processor speeds,
>>and hd capacity, despite fears, have not slowed down.
>>We could calculate the approximate completion dates
>>for the 4,5,6 pc databases, the increase in capacity/processing
>>required by each new set, and hence try to predict when we might
>>have the 7,8.. pc databases.. :-)
>>Anyone?
>Here is an oversimplified analysis:
>Assumptions: (these are guesses; please correct me if I got these wrong!(
>Computing power and storage doubles every year.
>2002 Tablebase = 6 pieces
>Storage/time to compute increase needed to add a piece = 64X
>2002 exhaustive search from the opening position = 10 plies
>Computer power increase needed to add an exhaustive search ply: = 64X
>YEAR TB PLIES
>2002 - 6 - 10
>2008 - 7 - 11
>2014 - 8 - 12
>2020 - 9 - 13
>2026 - 10 - 14
>2032 - 11 - 15
>2038 - 12 - 16
>2044 - 13 - 17
>2050 - 14 - 18
>2056 - 15 - 19
>2062 - 16 - 20
>2068 - 17 - 21
>2074 - 18 - 22
>2080 - 19 - 23
>2086 - 20 - 24
>2092 - 21 - 25
>2098 - 22 - 26
>2104 - 23 - 27
>2110 - 24 - 28
>2116 - 25 - 29
>2122 - 26 - 30
>2128 - 27 - 31
>2134 - 28 - 32 <-- Exhaustive search finds mate from opening position
>2140 - 29
>2146 - 30
>2152 - 31
>2158 - 32 <-- Look up mate from opening position in tablebase
Of course we are into the old "more permutations than atoms
in the universe" argument here - has anyone calculated how
many positions need to be stored for each stage?
Mind, by then we will probably be storing data by modulating
wavelet characteristics on sub-atomic particles, 10^32 bits
per atom.. :-)
ade
>The question is, in what year will a computer be able to start with the opening
>position, do an exhaustive search to positions in the tablebase, and thus play
>a perfect game of chess every time?
>--
> Guy Macon http://www.guymacon.com Guy Macon http://www.guymacon.com
David F. Cox
opening
> position, do an exhaustive search to positions in the tablebase, and thus
play
> a perfect game of chess every time?
>
> --
> Guy Macon http://www.guymacon.com Guy Macon http://www.guymacon.com
Perhaps this one. All it has to do is offer a draw?
David F. Cox
Kym
What if 1. d4! wins and 1.e4? loses -or- 1. h4! wins :-)
"David F. Cox" <bigd...@netscape.net> wrote in message
news:Uxfq9.10143$345.4...@newsfep2-win.server.ntli.net...
David F. Cox
> This assumes that perfect play = draw.
> What if 1. d4! wins and 1.e4? loses -or- 1. h4! wins :-)
They don't work for me.
DFC
jeff1101
> Another win by the WC and it's starting to look like a blowout. 2.5 -
> 0.5
> and a win with black at that!
> If this continues deep fritz won't be selling and might even be
> having a recall for deep loser.
>
> I can hear IBM laughing and saying what a pretender.
>
> The question is how many clowns paid for this loser only to find out
> it can't win?
> The fritz team is going to be the laughing stock of the chess
> programs!
> Can the fritz team find a job after this blow out.
>
> Will deep fritz win 1 game? Dam sure don't look like it can.
> Kramnik is calling his banker telling him get ready for a million
> dollar deposit.
> What's that I hear warming up in the background........ sounds like
> the fat lady!
They are now tie 3 - 3! Guess whos laughing now?