Fritz 10 vs Rybka 2.2

Main

Date: 21 Dec 2006 15:47:52
From: Paul
Subject: Fritz 10 vs Rybka 2.2

Hi,

I purchased Fritz 10 and Rybka 2.2.

Rybka keeps beating Fritz 10 and I thought that Fritz 10 was suppose to be
the strongest software program. Is there a setting in Fritz 10 to make it
play stronger or is it just that Rybka 2.2 is stronger?

Any help would be appreicated.

Paul

Date: 05 Feb 2007 13:41:51
From: Thomas T. Veldhouse
Subject: Re: Fritz 10 vs Rybka 2.2

Paul <[email protected] > wrote:
> Hi,
>
> I purchased Fritz 10 and Rybka 2.2.
>
> Rybka keeps beating Fritz 10 and I thought that Fritz 10 was suppose to be
> the strongest software program. Is there a setting in Fritz 10 to make it
> play stronger or is it just that Rybka 2.2 is stronger?
>

I have been playing Rybka 2.2n2 up at Playchess and it handily beats Fritz10
opponents [even deep Fritz]. Junior is a tough match though.

BTW ... Rybka 2.3 is due out on the 12th.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 05 Feb 2007 07:46:53
From: SAT W-7
Subject: Re: Fritz 10 vs Rybka 2.2

what time limits do you use or does it matter ?

Give Fritz 30 min and give Rybka 20 min and see who wins ...Or Fritz
one hour and Rybka 30 or 45 min.........

Also on Fritz i think you can make it play in the style of certain GM's
, Karpove , Kasporove so id try all of their styles + i hear you can
make it play different pieces stronger so id try that too.

Date: 05 Feb 2007 16:01:25
From: Thomas T. Veldhouse
Subject: Re: Fritz 10 vs Rybka 2.2

SAT W-7 <[email protected] > wrote:
>
> Also on Fritz i think you can make it play in the style of certain GM's
> , Karpove , Kasporove so id try all of their styles + i hear you can
> make it play different pieces stronger so id try that too.
>

The latter is a feature of Chessmaster, not Fritz [to my knowledge].

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 05 Feb 2007 09:37:57
From: SAT W-7
Subject: Re: Fritz 10 vs Rybka 2.2

I have a friend who has Fritz 9 and i think he told me that it can play
different GM styles but i could be wrong about that so ill ask him next
time i talk to him ..

Id try everything i could to tweak Fritz I0 to play at its
strongest...

I have Ivan and it is about I700 ELO so it loses when i play strong
computers and it bugs me .. ha ha ha........

I see the weaknesses of ivan and yet i still can not beat it .... I
have only won one game and lost about 400 games at the strongest level..

Date: 05 Feb 2007 18:44:09
From: Thomas T. Veldhouse
Subject: Re: Fritz 10 vs Rybka 2.2

SAT W-7 <[email protected] > wrote:
> I have a friend who has Fritz 9 and i think he told me that it can play
> different GM styles but i could be wrong about that so ill ask him next
> time i talk to him ..
>

I have all the Fritz 7 through 10 and have never seen it in the software.
Chessmaster has had it for years, and who knows if the software analogue is
anything like the real GM.

> Id try everything i could to tweak Fritz I0 to play at its
> strongest...
>

I like Fritz to be handicapped when playing me, as I will never come close to
beating it otherwise. I want to learn from it, but not be completely
slaughtered by it everytime. The only catch I have noticed so far is that
when it plays weeker, it sometimes throws a move so badly that I can't fail to
note it when it happens [feels like a boxer pulling a punch]. It would be
nice if it was a little more subtle.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 08 Feb 2007 18:45:28
From: David Richerby
Subject: Re: Fritz 10 vs Rybka 2.2

Thomas T. Veldhouse <[email protected] > wrote:
> I like Fritz to be handicapped when playing me, as I will never come
> close to beating it otherwise. I want to learn from it, but not be
> completely slaughtered by it everytime. The only catch I have
> noticed so far is that when it plays weeker, it sometimes throws a
> move so badly that I can't fail to note it when it happens [feels
> like a boxer pulling a punch]. It would be nice if it was a little
> more subtle.

I find that it's usually OK in `Handicap and Fun' mode. Friend mode
just causes it to blunder material until it's effectively giving you
odds; sparring mode is a little more interesting (the computer
deliberately leaves tactical opportunities for you to spot) but still
prone to making crass-looking blunders.

Dave.

--
David Richerby Poetic Perforated Boss (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a middle manager but it's full
of holes and in verse!

Date: 05 Feb 2007 15:34:17
From: SAT W-7
Subject: Re: Fritz 10 vs Rybka 2.2

Thanks for that info , i thought i mite be wrong...

When i said i wanted fritz as strong as i can make it is when id play
other computers ..
My friend go's on a site and sets his fritz to play 30 min games and he
can go to work and come home and his computer mite have played 4 or 5
games vs Rybeka , Shredder and others ..
The games are all in his computer so he can play them back.. If he is
home he can watch his machine play as it plays real time ...
That is how i think it works if i understand him .

His fritz does pretty good at different time controls + some of the
computers are better than his so that makes a difference too..

Date: 02 Feb 2007 16:49:13
From: [email protected]
Subject: Re: Fritz 10 vs Rybka 2.2

Well, using the Rybka model (as I understand it) of assigning
probabilities to various moves in the search tree, and searching
deeper in lines that are more likely to be the best, it might be
optimal to have different processors or different machines working on
different problems, but all assigning probabilities to the same search
tree.

For example one machine might serve as mate-finder, and assign maximal
probabilities to moves that lead to a forced mate (or perhaps more
important, assigning zero probability to blunders that allow mate, so
that the other machines don't have to waste time looking at that part
of the tree).

Another machine might be looking only for win of material, and assign
higher probabilities to moves that win material (or again perhaps more
importantly, assign lower probabilities to blunders that lose
material)

Then the master machine could check the lines for positional
correctness, without having to spend much time on lines that the above
machines have ked as probably incorrect.

You don't lose anything by taking such a probabilistic approach as
Rybka takes, since deterministic algorithms are included (for example
by assigning probability 1 to a move that a deterministic algorithm
thinks is best)

Other special purpose machines might be devoted to looking for
zugzwang positions, or searching deeply for lines that hit the Nalimov
tablebases.

If a machine had exhausted the possibilities related to its special
purpose task, it could be reassigned to some other task.

Since there is a certain loss of efficiency when trying to use
seperate processors in parallel on the same task, at some point it may
turn out that using some of the processors for different tasks may be
more efficient than trying to use them all in parallel on the same
task.

Another way of looking at it is using some machines for simple tasks
(such as mate-finding or win of material) and assigning rough
probabilities based on that, while slower algorithms evaluating
positional characteristics assign more and more refined probabilities
to the moves, correcting the rough evaluations assigned earlier.

Date: 05 Feb 2007 07:49:50
From: SAT W-7
Subject: Re: Fritz 10 vs Rybka 2.2

By the way i like Shredder , how does Rybka do vs it ?

Date: 06 Jan 2007 19:18:50
From: [email protected]
Subject: Re: Fritz 10 vs Rybka 2.2

Rybka is very strong, but it plays more positionally than most other
engines. One possibility for improvement might be to run a tactical
version of Rybka, such as Rybka WinFinder, on one machine, and a
positional version of Rybka on another machine.

Tactical positions and strategic positions seem to have different
characteristics, so it might be more efficient to have 2 different
programs running at the same time, one program based on the premise
that there are no tactical shots in the positions it looks at, and that
the problem is to choose between several good positional moves, and the
other program based on the premise that there is a unique best tactical
shot in the positions it looks at, and focused on trying to find it.

Then if the tactical program does find a position where there is a
unique best tactical move, write the evaluation of that position into
the other program's hash table or something like that. So the
positional program wouldn't have to analyze that position, but could
just use the value given to it by the tactical program.

Using the Nalimov tablebases a lot, I find that tactical positions are
more likely to have a unique best move than ordinary positions. Of
course sometimes there are several equally good moves, such as when the
winning idea is to move a rook along the 7th rank from A7 to H7,G7, or
F7 which are all equally good. But there is often basically a single
best idea.

So a tactical program might be basically looking for a unique best move
or set of moves that are kedly better than the alternatives, whereas
a positional program might be focused on finding subtle differences
between several moves that look almost equally good at first. So it
might be better to seperate the two tasks and give them to 2 seperate
programs to try and solve, rather than having a single algorithm trying
to solve two such different problems at the same time.

Hope that makes sense :-)

Date: 15 Jan 2007 00:24:01
From: Johnny T
Subject: Re: Fritz 10 vs Rybka 2.2

[email protected] wrote:
> Rybka is very strong, but it plays more positionally than most other
> engines. One possibility for improvement might be to run a tactical
> version of Rybka, such as Rybka WinFinder, on one machine, and a
> positional version of Rybka on another machine.
>
> Tactical positions and strategic positions seem to have different
> characteristics, so it might be more efficient to have 2 different
> programs running at the same time, one program based on the premise
> that there are no tactical shots in the positions it looks at, and that
> the problem is to choose between several good positional moves, and the
> other program based on the premise that there is a unique best tactical
> shot in the positions it looks at, and focused on trying to find it.
>
> Then if the tactical program does find a position where there is a
> unique best tactical move, write the evaluation of that position into
> the other program's hash table or something like that. So the
> positional program wouldn't have to analyze that position, but could
> just use the value given to it by the tactical program.
>
> Using the Nalimov tablebases a lot, I find that tactical positions are
> more likely to have a unique best move than ordinary positions. Of
> course sometimes there are several equally good moves, such as when the
> winning idea is to move a rook along the 7th rank from A7 to H7,G7, or
> F7 which are all equally good. But there is often basically a single
> best idea.
>
> So a tactical program might be basically looking for a unique best move
> or set of moves that are kedly better than the alternatives, whereas
> a positional program might be focused on finding subtle differences
> between several moves that look almost equally good at first. So it
> might be better to seperate the two tasks and give them to 2 seperate
> programs to try and solve, rather than having a single algorithm trying
> to solve two such different problems at the same time.
>
> Hope that makes sense :-)
>
Made sense, but is almost nonsensical. But don't take that too badly.
It is just a misunderstanding of reality.

First the dual position question. The whole strategic vs tactical
question. This is wrong for a few reasons, any one of the reasons
enough to not warrant anything past the discussion. But lets try one of
the easiest. First, there is no chess engine on earth, rybka included
that is going around NOT looking at tactics. This is the number one,
best thing they are good about. Move pieces around, capture stuff, see
if we are still in forcing sequences, if so, who has the most material,
and are there any gross imbalances in the position? Is it enough to
win? Many programs stop paying attention after a win has been demonstrated.

Essentially there is no engine that really "plays" tactics better than
another engine. There was an age when there were deeper engines than
another, but we are beyond that age (except for Hydra, and probably deep
blue. But they may not have been deep enough for the likes of rybka,
and brethren. We will see).

The other way to look at the question is that there is either a
fundamentally "best" move, but we have no way to test that, or we have
moves that are better against a given opponent, and we do that already.
All the time in the CPU vs CPU realm. As a matter of fact that was
one of the complaints were that the games were getting to specifically
opponent oriented, and that accept for SSDF standings, the chess wasn't
necessarily getting "better" especially against the other test subjects,
GM's, Super-GM's and world champions.

At the end of the day, all the "meat" in a chess program is in the quiet
areas. Where ideas will portend tactics someday. And those that best
perform here, are the best at winning.

The interesting thing is when the machines change how we think about
certaing positons or tableux. This has happened as recently as the
Kramnik v. Fritz 10 match, but I don't know how much more we (the
general public), will get to see of this kind of new thinking, and how
much will be just so much verbage in the pile of machine v. machine
games that will be happening. I do believe that desktop class machines
are in a new race to be much stronger. And it isn't clever programming
like you have shown, it is going to be in just adding ever more bits of
sts in the program. But we will tend to only see this in the
standings. Us patzers won't be able to see or understand anything
profound happening. Those that would, aren't going to be looking. They
will be off doing tourneys and stuff. But it may seep in during
training, we will see.

Date: 14 Jan 2007 14:09:34
From: SAT W-7
Subject: Re: Fritz 10 vs Rybka 2.2

Which one won ?

Have you had those two play Shredder or Hercules ? That would be
two good games.

Date: 21 Dec 2006 14:35:15
From: Rob
Subject: Re: Fritz 10 vs Rybka 2.2

David Richerby wrote:
> Rob Mitchell <[email protected]> wrote:
> > Does Rybka have a function to turn off tablebases?
>
> Early versions of Rybka didn't even have a function to turn on
> tablebases! :-)
>
> But you should be able to turn off tablebases for any engine just by
> moving the files and not telling it where they are. Most engines
> should cope with that.
>
>
> Dave.
>
> --
> David Richerby Addictive Technicolor Gnome (TM):
> www.chiark.greenend.org.uk/~davidr/ it's like a smiling garden ornament
> but it's in realistic colour and you
> can never put it down!

Thanks David!

Date: 21 Dec 2006 12:29:50
From: Rob Mitchell
Subject: Re: Fritz 10 vs Rybka 2.2

Ruud wrote:
> "Paul" <[email protected]> schreef in bericht
> news:[email protected]...
> > Thanks Ruud. Even though I have Rybka 2.2 x32bit rather than x64, it still
> > takes Fritz 10 apart. Is there any tweeking that can be done in Fritz 10
> > to make it a bit stronger.
>
> You could experiment with the engine-parameters, but the default-settings
> are usually the best.
>
> > Do you know why Chessbase does not fight back to make their products the
> > strongest.
>
> I'm pretty sure they are.
> It's just that nowadays there are better chess-programmers out there.
> In the beginning of the 90s there were Fritz, Rebel, and some others, that
> dominated.
> Now Shredder, Fruit, Zappa, and now Rybka, have overtaken the Fritz of the
> 90s.
>
>
> >I would have thought this would be a good selling point.
>
> If they are again the top, I'm sure that it would become the main point of
> advertizing again.
> Still, the interface of Chessbase is still inviting to programmers and
> players, so they're ok, I think.

Does Rybka have a function to turn off tablebases? Can't you get that
program from Chessville.com?

Date: 21 Dec 2006 21:34:31
From: David Richerby
Subject: Re: Fritz 10 vs Rybka 2.2

Rob Mitchell <[email protected] > wrote:
> Does Rybka have a function to turn off tablebases?

Early versions of Rybka didn't even have a function to turn on
tablebases! :-)

But you should be able to turn off tablebases for any engine just by
moving the files and not telling it where they are. Most engines
should cope with that.

Dave.

--
David Richerby Addictive Technicolor Gnome (TM):
www.chiark.greenend.org.uk/~davidr/ it's like a smiling garden ornament
but it's in realistic colour and you
can never put it down!

Date: 21 Dec 2006 16:56:51
From: Ruud
Subject: Re: Fritz 10 vs Rybka 2.2

"Paul" <[email protected] > schreef in bericht
news:[email protected]...
> Hi,
>
> I purchased Fritz 10 and Rybka 2.2.
>
> Rybka keeps beating Fritz 10 and I thought that Fritz 10 was suppose to be
> the strongest software program. Is there a setting in Fritz 10 to make it
> play stronger or is it just that Rybka 2.2 is stronger?
>
> Any help would be appreicated.
>
> Paul

Rybka is simply the strongest engine around now, according to pure playing
strength.
Fritz 10 however has some advanced training- and analysis-functions, which
make it the best for learning, getting a better chessplayer overall.
Chessbase, the company that distributes the Fritz-software, tends to
emphasize this, ever since other programmers have managed to beat their
engines.
Maybe that's what causes the confusion.

Rating list.
http://www.husvankempen.de/nunn/40_120_ratinglist/ratinglist/rangliste.html
Rybka 2.2 is even stronger than 2.1

Greetings from Ruud.

Date: 21 Dec 2006 16:17:40
From: Paul
Subject: Re: Fritz 10 vs Rybka 2.2

Thanks Ruud. Even though I have Rybka 2.2 x32bit rather than x64, it still
takes Fritz 10 apart. Is there any tweeking that can be done in Fritz 10 to
make it a bit stronger. Do you know why Chessbase does not fight back to
make their products the strongest. I would have thought this would be a good
selling point.

Regards

Paul

"Ruud" <[email protected] > wrote in message
news:[email protected]...
>
> "Paul" <[email protected]> schreef in bericht
> news:[email protected]...
>> Hi,
>>
>> I purchased Fritz 10 and Rybka 2.2.
>>
>> Rybka keeps beating Fritz 10 and I thought that Fritz 10 was suppose to
>> be the strongest software program. Is there a setting in Fritz 10 to make
>> it play stronger or is it just that Rybka 2.2 is stronger?
>>
>> Any help would be appreicated.
>>
>> Paul
>
> Rybka is simply the strongest engine around now, according to pure playing
> strength.
> Fritz 10 however has some advanced training- and analysis-functions, which
> make it the best for learning, getting a better chessplayer overall.
> Chessbase, the company that distributes the Fritz-software, tends to
> emphasize this, ever since other programmers have managed to beat their
> engines.
> Maybe that's what causes the confusion.
>
> Rating list.
> http://www.husvankempen.de/nunn/40_120_ratinglist/ratinglist/rangliste.html
> Rybka 2.2 is even stronger than 2.1
>
> Greetings from Ruud.
>

Date: 21 Dec 2006 19:05:25
From: Ruud
Subject: Re: Fritz 10 vs Rybka 2.2

"Paul" <[email protected] > schreef in bericht
news:[email protected]...
> Thanks Ruud. Even though I have Rybka 2.2 x32bit rather than x64, it still
> takes Fritz 10 apart. Is there any tweeking that can be done in Fritz 10
> to make it a bit stronger.

You could experiment with the engine-parameters, but the default-settings
are usually the best.

> Do you know why Chessbase does not fight back to make their products the
> strongest.

I'm pretty sure they are.
It's just that nowadays there are better chess-programmers out there.
In the beginning of the 90s there were Fritz, Rebel, and some others, that
dominated.
Now Shredder, Fruit, Zappa, and now Rybka, have overtaken the Fritz of the
90s.

>I would have thought this would be a good selling point.

If they are again the top, I'm sure that it would become the main point of
advertizing again.
Still, the interface of Chessbase is still inviting to programmers and
players, so they're ok, I think.