Chess One wrote:

... Mr. Hyatt ...

You aren't worthy to lick Professor Hyatt's boots, troll.

I didn't see the earlier messages, so please excuse me...

"help bot" wrote in message
ups.com...

I was obviously replying to comments by IM Innes
which presumed that BH ought to have taken an
"academic" approach (i.e. teach his program to
truly learn chess, not just to win the easy way).

Easier said than done, of course...

*except* that Dr Hyatt was able
to use his experience with CB

Right. Cray Blitz, not Crafty. Mr. Hyatt always
insisted that his chess program was *inseparable*
from the Cray, but then turned around and wrote
a chess program for PCs just the same, when it

It is true that CB was very itimately tied to the Cray's vector hardware.

I've read two versions of the CrayBliz source and although it had a portable
fortran version, the performance was significantly slower.

Even modern SSE3 etc. kind of instructions wouldn't help.

Plus, Cray Blitz took advantage of the vastly higher memory bandwidth
available in a supercomputer. Trying to do the same things on a PC would
result in terrible memory latencies.

became clear that others were having success
*without* super-speed. His open-source approach

No, actually he stopped using the Cray's because it was getting harder &
harder to get enough time to actively work on the program. He often didn't
even have enough time for tournaments, let alone development, testing,
tuning etc.

(He's talked about cases where by the end of the tournament, his program had
been shifted from the faster systems to the slower systems. When he was
playing the final games of a tournament against the strongest players, his
program would sometimes be running on slower hardware with a 10x performance
loss or such.)

When he started Crafty, the Pentium 2 etc. were available and no, they
couldn't even come close to what his Cray's could provide.


In the early years, the speed of the Cray was
credited by some for the success of Cray Blitz,
and this really ticked Mr. Hyatt off because he
wanted to get the credit. When critics requested
a conversion to PC code for comparison purposes,
the sh*t hit the fan. He insisted that his chess
algorithms could no way be translated, period.

He's wrong and right.

The early years, the Cray *was* responsible for Cray Blitz' performance.
True, later on he did develop a lot of extra algorithms etc. But in the
early years, it was mostly the Cray.

As for converting Cray Blitz to the PC.... (snort) I've seen the source for
CrayBlitz and porting it to a PC would have been an utter waste of time.
We're talking a factor of 20 or more in performance loss due to the program
structure & algorithms not matching the hardware.

Once he started vectorizing the program and taking advantage of the large
memory bandwidth and parallel processing, there was no going back to a plain
scalar system.

Cray Blitz was a 'mailbox' chess program, but it also used bitboards and
took heavy advantage of the vector hardware and memory bandwidth. Doing
that on a PC would have a been a waste of time for anything but curosity or
basic program testing.


Mr.
Hyatt, the creator of Crafty, simply exploited the raw
speed of a mainframe he had (virtually unique) access to.

Actually, it was Larry Nelson who exploited the raw speed,
and Bob Hyatt and Bert Gower who wrote the chess and the algorithms.

In the early years, yes. Larry Nelson did do a lot of CAL conversion. But
that also meant that Bob had to write the program so that it could be easily
vectorized. No amount of CAL will improve a plain scalar program on a
vector system such as the Cray.

In the later years, Bob took over the CAL development. (CAL is Cray
Assembly Language.)


Wrote the chess? What does that mean?

I take it you are reminding us that although BH
often gets the credit, there were actually three men
involved in the creation of Cray Blitz, not counting
the multitude responsible for the development of
the Cray itself. Careful here; one false step and you
might incriminate Bob Hyatt, who has always insisted
that the chess algorithms could not be extracted from
CB, no way, no how.

Extracted, sure. Efficiently used on a micro, no way.





The whole caboodle was *far* from simple, as anyone who tries to
write highly-parallel chess code soon discovers. There was some
very clever code, both high- and low-level, in Cray Blitz.

Too bad the PC guys soon began to catch up
in terms of results. The thing is, how can you
(rationally) take credit for awesome programming
skill, summarily dismiss the brute power of the
Cray, and yet watch the competition close the
gap in your rear-view mirror when they are
driving ordinary cars and you're in a Porsche?
Something is amiss. One should at least admit
that the fast car has something to do with it. If
it's a Ford Pinto you see closing in, well, find a
better driver.

I do agree that Bob should give a lot of credit to the Cray... It allowed
much deeper searches and massive hash tables.

When he was having heavy competition from even 8 bit micro's, it does show
there's something not quite right.

However, the diminishing returns in chess searching also plays a role here.
Your computer may be 16 times faster but you'll probably only be searching a
couple ply deeper.



Similar comments apply to Deep Thought and other programs
"accused" of "simply" being fast. Other things being equal, faster
is better. But other things are not equal, and there are seriously
bright people behind the successful programs.

Nobody is saying otherwise.

The real issue was the fact that despite vastly
superior hardware, the Crays and Deep Blues
were unable to maintain a substantive lead for any
length of time over programmers who chose to
work their skills on commonplace hardware.

It's also a lot about the skill of the programmer and the amount of time
they can spend developing their program.

When you have hardware that is available to you whenever you want, as long
as you want, you are much more likely to do more tweaking of the program.

For example, with today's PC's, many chess programmers (including Bob) have
whole 'farms' of computers playing games against each other as a way to test
modifications. That's something you just can't do with a Cray or any super
computer or specialized hardware.



Logic would seem to indicate that were the
conditions right, the best PC programmers might
switch places with these "geniuses" and we
would see a different story; perhaps a widening of

Sorry, but that's not quite true. Different people think differently and
are better suited to different style hardware, etc.

A 'mailbox' programmer would be lost with the 64 bit systems and vector
hardware of a Cray. They wouldn't know how to take advantage of the extra
power and abilities.


the gap. (Obviously, the PC programmers would
need to magically transmute their PC-programming
skills into Cray-programming skills, and vice versa
for the other guys.)


That's really pushing it there.... Porting between those hardwares involves
a lot of changes in algorithms and board representations and data
structures.








----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----

On May 4, 5:38 pm, raylopez99 wrote:
On May 4, 11:39 am, JohnnyT wrote:
[deleted: pean to Rybka, about how great it is and radically
different]

Bull Shiite. Rybka is kicking fanny because the number of games
played by it are still small. It's too early to tell. Regression to
the mean will follow.

FWIW, when I first learned of the existence of Rybka,
it was described as having a (performance) rating of
around 3000. Now it seems to have already started
a downward spiral, one list showing a new version of
Hiarcs to be within 50 or so points of Rybka.

Of all the articles I have seen thus far, the one that
impressed me most is where a game between GMs
Kasparov and Anand, along with a multitude of
others which soon followed suit, had the world
champ moving his Queen around (first to d4, then
to h4, and finally to g3) in the opening, which seems
like one of the classic beginner-style errors. Yet
everybody copied the line. Then Rybka came along
and reassessed things, showing that giving up the
exchange leads perforce to a return of some material,
along with a strong position for Black. IMO, it's
almost as if a line were drawn in the sand, and
everybody and his brother was on the crude
materialistic side, except for Rybka. Of course,
this may well have been the same game where
Rybka only won the ending due to its computer
opponent refusing to advance it pawns, but this is
beside the point. The real point is that nobody else
seems to have spotted the *idea* of ...Bh6, despite
the move being right there, in plain sight.

-- help bot

On May 5, 7:14 am, Ralf Callenberg wrote:
05.05.2007 11:13, raylopez99:

With SSDF the pool is computer vs. strong
human chess player.

Where did you get this idea from? Of course SSDF is purely based on
engine-engine matches.

Ray Lopez is obviously a few bricks short of a load.
His "facts" quite often appear out of thin air, and are
about equally substantive.

However, he may be right that Rybka's rating will
gradually revert toward the other top programs' --
time will tell.

It still amazes me that somehow new chess
programs can seem to suddenly appear out of thin
air, popping into first place by a substantial margin.
The very idea that, say, Fritz 10.0 is a masterpiece
which has taken years to hone and tune is blasted
away by the sudden appearance of a 1.x program
which simply leaves it in the dust. I would like to
know exactly why some of these other programs
can so easily be bested by a newcomer.

In the early days, the answer was obvious: either
they grabbed someone who knew little about chess
but who was a good programmer, or just the reverse.
But after all these years of intense competition, you
would think things would have changed dramatically.

-- help bot

On May 5, 10:36 am, wrote:
help bot wrote:
I was obviously replying to comments by IM Innes

There's where you made your mistake. Innes is an idiot and
a tTroll, and the best policy is to ignore his blatherings.

'''
(0 0)
+-------oOO---(_)---OOo-------+
| Please don't feed the Troll |
+-----------------------------+
\ _ /
|||
ooO Ooo


Something is no lined up right. The eyes are well
to the left of the troll's nose, and the bottom right
portion of the sign is missing. Besides, IM Innes
has *red* hair, and is a bit taller.

-- help bot

On May 5, 4:36 pm, wrote:
Chess One wrote:
... Mr. Hyatt ...

You aren't worthy to lick Professor Hyatt's boots, troll.

Bob Hyatt reportedly doesn't wear boots. He wears
sneakers, which he finds are more comfortable when
driving the bus (to run down certain types of criminals).

-- help bot

On May 5, 5:32 pm, "Hello" wrote:

I was obviously replying to comments by IM Innes
which presumed that BH ought to have taken an
"academic" approach (i.e. teach his program to
truly learn chess, not just to win the easy way).

Easier said than done, of course...

For starters, what's so difficult about removing the
by-rote opening book which, BTW, took considerable
time to choose and create? Time which might have
been devoted to teaching the program to evaluate
opening moves for itself. Plugging in moves is the
polar opposite of what is desired in terms of A.I.


Right. Cray Blitz, not Crafty. Mr. Hyatt always
insisted that his chess program was *inseparable*
from the Cray, but then turned around and wrote
a chess program for PCs just the same, when it

It is true that CB was very itimately tied to the Cray's vector hardware.

A few posts back, Mr. Walker was insisting that
it was someone else who did the mechanical work,
that Bob Hyatt and another man wrote what he
called "the chess".


I've read two versions of the CrayBliz source and although it had a portable
fortran version, the performance was significantly slower.


The point. Critics argued that the great success
of Cray Blitz was the result of its vastly superior
speed and power.


became clear that others were having success
*without* super-speed. His open-source approach

No, actually he stopped using the Cray's because it was getting harder &
harder to get enough time to actively work on the program. He often didn't
even have enough time for tournaments, let alone development, testing,
tuning etc.

My, how things change! From what I could see,
he not only had time for this, but also for engaging
in innumerable arguments in the chess newsgroups
until it reached a point where he was bordering on
paranoia, or a severe persecution complex.


(He's talked about cases where by the end of the tournament, his program had
been shifted from the faster systems to the slower systems.

Right. But this was because the tournaments were
held during the day, not at the optimum time for him
to secure free time on the Cray. Playing in certain
tournaments was the only way to achieve *official*
ratings.


When he was
playing the final games of a tournament against the strongest players, his
program would sometimes be running on slower hardware with a 10x performance
loss or such.)


And again, this resulted in a dramatic lessening
of his inherent speed and power advantage, which
was the whole basis of these discussions. (Was
it this speed/power advantage, or was it the amazing
chess-programming skill that gave Cray Blitz its
edge?)


When he started Crafty, the Pentium 2 etc. were available and no, they
couldn't even come close to what his Cray's could provide.

Precisely.


In the early years, the speed of the Cray was
credited by some for the success of Cray Blitz,
and this really ticked Mr. Hyatt off because he
wanted to get the credit. When critics requested
a conversion to PC code for comparison purposes,
the sh*t hit the fan. He insisted that his chess
algorithms could no way be translated, period.

He's wrong and right.

The early years, the Cray *was* responsible for Cray Blitz' performance.
True, later on he did develop a lot of extra algorithms etc. But in the
early years, it was mostly the Cray.


This was essentially the position of BH's "critics".


As for converting Cray Blitz to the PC.... (snort) I've seen the source for
CrayBlitz and porting it to a PC would have been an utter waste of time.
We're talking a factor of 20 or more in performance loss due to the program
structure & algorithms not matching the hardware.

Once again, this singular focus on speed and power
is telling. What about Mr. Hyatt's chess-programming
skill, separate from everything else? Was it enough to
push him to the top of the heap?


Once he started vectorizing the program and taking advantage of the large
memory bandwidth and parallel processing, there was no going back to a plain
scalar system.

A few posting back, this part was attributed to
someone else, to "the third man" of the CB team,
not to BH:

Actually, it was Larry Nelson who exploited the raw speed,
and Bob Hyatt and Bert Gower who wrote the chess and the algorithms.

In the early years, yes. Larry Nelson did do a lot of CAL conversion. But
that also meant that Bob had to write the program so that it could be easily
vectorized. No amount of CAL will improve a plain scalar program on a
vector system such as the Cray.

In the later years, Bob took over the CAL development. (CAL is Cray
Assembly Language.)

I see. He had time to "take over" programming in
Assembly language and for fighting with dozens of
newsgroup denizens, but not enough time to even
manage a tournament, now and then? My view is
that it was his faltering success (because of speed
issues) that scaled back his tournament participation.
In other words, when he had a clear upper-hand, he
played, but when he lost this, he petered out. In
sum, the "critics" were right.

In fairness to Bob Hyatt, I should point out that his
free-ware program was and still is one of the most
successful of its kind. But the whole issue was in
his pretense that it was not the Cray, but the Hyatt
which was responsible for CB's great success.


Wrote the chess? What does that mean?

I take it you are reminding us that although BH
often gets the credit, there were actually three men
involved in the creation of Cray Blitz, not counting
the multitude responsible for the development of
the Cray itself. Careful here; one false step and you
might incriminate Bob Hyatt, who has always insisted
that the chess algorithms could not be extracted from
CB, no way, no how.

Extracted, sure. Efficiently used on a micro, no way.

To sum up: the chess algorithms used for Cray Blitz
were specifically written and optimized to exploit the
killer advantage of its speed and power. They were
not vastly-superior-chess-algorithms which just
happened to have been written for a very fast machine,
whose vast superiority was merely an added bonus.
:D

However, the diminishing returns in chess searching also plays a role here.
Your computer may be 16 times faster but you'll probably only be searching a
couple ply deeper.

With vastly superior hardware, it is possible to
have a combination of brute-force and selective
search methods, and here one might expect a
bit more than a couple of additional plys. In fact,
at tournament speeds (which is how these
programs got rated) the addition of even two
plys back then should have given the Cray a
big edge in head-to-head competition. Granted,
those games lasted for hours, and so there arose
the aforementioned technical issues.


Similar comments apply to Deep Thought and other programs
"accused" of "simply" being fast. Other things being equal, faster
is better. But other things are not equal, and there are seriously
bright people behind the successful programs.


I agree. But sometimes these seriously bright
people do seriously stupid things. Take DeepBlue,
for instance. In one match it went out-of-book
after 1.e3, losing like a fish. When things like
this happen, it is to be expected that critics will
leap on the blunder, and pointing out that the
fellow in charge had a Ph.D. is of little value; you
can't put lipstick on a pig.



For example, with today's PC's, many chess programmers (including Bob) have
whole 'farms' of computers playing games against each other as a way to test
modifications. That's something you just can't do with a Cray or any super
computer or specialized hardware.

But many issues go deeper than merely tweaking
the program. For all the time wasted on fine-tuning
the opening book to optimize results against the
current competition, one might have instead devoted
some time to "teaching" the program to play the
opening properly on its own. For all the time wasted
in battling adversaries on the net, a fellow could
instead have developed an entirely new approach
to analyzing the game. For example, Cray Brute
Force, and Cray Selective Search, and Cray Blitz,
and Cray Infinite Analysis. Each of these might
be optimized specifically for their respective tasks.


Logic would seem to indicate that were the
conditions right, the best PC programmers might
switch places with these "geniuses" and we
would see a different story; perhaps a widening of

Sorry, but that's not quite true. Different people think differently and
are better suited to different style hardware, etc.

Wow. So you think that if we took the one guy
who had access to a Cray supercomputer and
"switch him" with the absolute best chess programmer
on the PC, there might well be no improvement
whatsoever? That's amazing. It's akin to suggesting
that if I were to jump out of my car and let A.J.Foyt
get in and drive, he would not get to McDonald's any
sooner than I would driving his race car, because we
think differently? So in other words, he's not really
any faster a driver than I am; we just drive different
cars. Who knew? :D

-- help bot

"help bot" wrote in message
oups.com...
On May 5, 5:32 pm, "Hello" wrote:

I was obviously replying to comments by IM Innes
which presumed that BH ought to have taken an
"academic" approach (i.e. teach his program to
truly learn chess, not just to win the easy way).

Easier said than done, of course...

For starters, what's so difficult about removing the
by-rote opening book which, BTW, took considerable

First, removing an opening book is easy. Trivial. Just don't have the book
file when you run Crafty.

The problem is that most programs have considerable difficulty playing
reasonable openings. The opening moves have consequences that are too far
into the game for programs to be able to usually determine.

Second, the statement I replied too wasn't about *removing* the opening
book, but the part about teaching "his program to truely learn chess".
That's a lot easier said than done.

The reality is that at this point of computer chess, *nobody* knows how to
truely program a chess program to learn chess openings even vaguely like
humans do.

Most things degrade into various types of rote learning with practically no
generalization.


**BUT** if you know how to do it, then code it into a program, run a few
thousand games and let it learn and publish the results. The computer chess
community will pat you and the back and hearitly congratulate you.


time to choose and create? Time which might have
been devoted to teaching the program to evaluate
opening moves for itself. Plugging in moves is the
polar opposite of what is desired in terms of A.I.

I suppose so.

Two points though... First, many human players learn openings by rote.

Second, computer chess hasn't been considered AI since the days of Mac HACK
VI, 30 years ago.


I've read two versions of the CrayBliz source and although it had a
portable
fortran version, the performance was significantly slower.

The point. Critics argued that the great success
of Cray Blitz was the result of its vastly superior
speed and power.

Agreed for the early versions and partially for the later versions.

The early versions were basically expanded versions of his old Blitz
program. Hyatt was depending on brute force number crunching and massive
hash tables.

Later on, he was still taking advantage of that, but he had gone beyond that
and developed a lot of ideas and algorithms, many of which happened to be
tuned specifically for the Cray architecture.

So by that time he was depending on the architecture because that's what he
had coded for. Changing architectures would have meant a complete rewrite
and re-think for his algorithms. It's pretty much impossible to seperate
the success and the Cray architecture. They were too tied together.

As for Cray Blitz' success depending on the Cray performance... (shrug) All
you really have to do is pretend it was running on a very slow Cray. You
can't seperate the Cray out of CrayBlitz, but you can reduce the performance
of the hardware.

The good old 'Technology Curve' shows when the hardware gets "x" times
faster, you'll search a little deeper and your rating will increase.

The exact shape of that curve is very disputed, but everybody agree's that
it's not linear. The faster you go or the stronger you are, the less change
there is.

A super slow Cray would probably have only effected CrayBlitz by a 100 - 150
points. (My guess. Pure speculation since there's no way it can be proven.)
It might not have been quite the winner it was, but it wouldn't have been a
push-over either.


became clear that others were having success
*without* super-speed. His open-source approach

No, actually he stopped using the Cray's because it was getting harder &
harder to get enough time to actively work on the program. He often
didn't
even have enough time for tournaments, let alone development, testing,
tuning etc.

My, how things change! From what I could see,
he not only had time for this, but also for engaging
in innumerable arguments in the chess newsgroups
until it reached a point where he was bordering on
paranoia, or a severe persecution complex.

Hyatt is still very active in chess forums. Just not here because there are
too many trolls and too much spam.

But talking is different from coding. And testing takes *lots* of time.

Your comments are pretty absurd.




(He's talked about cases where by the end of the tournament, his program
had
been shifted from the faster systems to the slower systems.

Right. But this was because the tournaments were
held during the day, not at the optimum time for him
to secure free time on the Cray. Playing in certain
tournaments was the only way to achieve *official*
ratings.

(frown) Yes... but what does that have to do with what I said?

I said he has commented numerous times that he had trouble getting enough
computer time to run tests during development and even for important
tournaments.

WHY it happened is irrelevant. Whether it was for an official tournament or
development / testing / tuning is irrelevant. The point was he began having
trouble getting enough time on a cray. But on personal computers he had all
the time and access he wanted.


When he was
playing the final games of a tournament against the strongest players,
his
program would sometimes be running on slower hardware with a 10x
performance
loss or such.)


And again, this resulted in a dramatic lessening
of his inherent speed and power advantage, which
was the whole basis of these discussions. (Was

Actually, to support that statment, you'll have to correlate game losses
with him using the slower system.

Unless you can do that, your statements are pure speculation.



it this speed/power advantage, or was it the amazing
chess-programming skill that gave Cray Blitz its
edge?)


When he started Crafty, the Pentium 2 etc. were available and no, they
couldn't even come close to what his Cray's could provide.

Precisely.


In the early years, the speed of the Cray was
credited by some for the success of Cray Blitz,
and this really ticked Mr. Hyatt off because he
wanted to get the credit. When critics requested
a conversion to PC code for comparison purposes,
the sh*t hit the fan. He insisted that his chess
algorithms could no way be translated, period.

He's wrong and right.

The early years, the Cray *was* responsible for Cray Blitz' performance.
True, later on he did develop a lot of extra algorithms etc. But in the
early years, it was mostly the Cray.


This was essentially the position of BH's "critics".

I wouldn't call myself a 'crtic'.

I've never actually met him. We have been friendly in the emails we've done
over the past few years, though.




As for converting Cray Blitz to the PC.... (snort) I've seen the source
for
CrayBlitz and porting it to a PC would have been an utter waste of time.
We're talking a factor of 20 or more in performance loss due to the
program
structure & algorithms not matching the hardware.

Once again, this singular focus on speed and power
is telling. What about Mr. Hyatt's chess-programming
skill, separate from everything else? Was it enough to
push him to the top of the heap?

Hard to say. Really hard to say.

Is he the best around. No. There are better programs.

Is he very good... Yes.

Is he clever. Yes.

Has he been methodical in his chess algorithm testings: Yes.

Has he tested a *lot* of ideas in the last 25 years? Yes.

Super fast hardware and parallel clusters will only get you a couple extra
ply. At the current ratings, that doesn't really get you much extra ratings
points.

If you do have access to the hardware, then you might as well use it.


Once he started vectorizing the program and taking advantage of the large
memory bandwidth and parallel processing, there was no going back to a
plain
scalar system.

A few posting back, this part was attributed to
someone else, to "the third man" of the CB team,
not to BH:

It's not all that simple.

Harry Nelson did indeed do a lot of CAL programming in Cray Blitz. (That
is: convert from Fortran to Cray Assembly Langauge.)

However, that's not "vectorizing" a program. To do that, you need the right
algorithms. When you have that, even the Cray Fortran compiler can do many
simple vectorizations for you.

Harry Nelson could see some obvious vectorizations and put them into CAL. A
loop, for example.

But for the more substantial oportunities, you need to pick the right
algorithms. You need to put the support into the program itself. And so
on. That requires major effort from Bob Hyatt. (For example, CrayBlitz
went from being a pure mailbox program to a hybrid mailbox / bitboard. That
was done to make vectorization easier.)

I'm not saying Harry Nelson didn't make a few suggestions to Dr. Hyatt.
After all, Nelson was very familiar with the Cray and what it could and what
its limiations were, etc.

But I'd have to say the main chess coding needed to make it vectorizable
came from Bob Hyatt.

Also, don't forget that after Nelson left, Cray Blitz continued to survive
and develop.




Actually, it was Larry Nelson who exploited the raw speed,
and Bob Hyatt and Bert Gower who wrote the chess and the algorithms.

In the early years, yes. Larry Nelson did do a lot of CAL conversion.
But

Incidentally, it's "Harry Nelson", not "Larry Nelson". Probably just a
typo, but I figured it should be corrected.


that also meant that Bob had to write the program so that it could be
easily
vectorized. No amount of CAL will improve a plain scalar program on a
vector system such as the Cray.

In the later years, Bob took over the CAL development. (CAL is Cray
Assembly Language.)

I see. He had time to "take over" programming in
Assembly language and for fighting with dozens of
newsgroup denizens, but not enough time to even
manage a tournament, now and then? My view is

*manage* a tournament?? Why would he want to do that?

Wouldn't it be a clear conflict of interest, with his program being in the
tournament?

Even if it wasn't, why would he want to manage a tournament? Not everybody
does.

And it does take time. Lots of time before hand (time better spent actually
making finishing touches before the games begin) and during.


that it was his faltering success (because of speed
issues) that scaled back his tournament participation.
In other words, when he had a clear upper-hand, he
played, but when he lost this, he petered out. In
sum, the "critics" were right.

You are entitled to your opinion.

Of course, that wouldn't explain why he still plays tournaments and does the
chess servers. After all, he doesn't have the top program, so he's not
always winning. He usually looses the top wins. But yet he still plays.
How about that...

Of course, it's also possible that anybody scaling back their activities in
something may be due to other things. Personal problems. Burning out.
Heath issues. All sorts of things. I don't know. I don't much care,
either.

He's here. He's active. He's talkative. He's helpful. His program is
available in source. (shrug) What more could you really want?




In fairness to Bob Hyatt, I should point out that his
free-ware program was and still is one of the most
successful of its kind. But the whole issue was in
his pretense that it was not the Cray, but the Hyatt
which was responsible for CB's great success.

Do you remember how bad his program played when it was first ported to the
Cray?

It lost the 1980 ACM and finished in 5th place!

It took a lot of effort to get it reworked so it could run well. It took a
lot of work to get it to work with the multiprocessor systems, etc.

I will say that Hyatt has often depended on raw computing power for the
extra ratings points. And that without it, his program wouldn't perform
quite as good as it did / does.

However, that's true of all chess programmers. If I had access to super
hardware, I'd be a fool not to use it.

His use of Super Computers (and now large clusters of micro's) is a bit of a
failing of his. He's hooked on the raw power and he depends on it being
there. There's no question of that. He's hooked on the power and he wants
MORE!!!

But his programs perform so well that I don't think it would have
drastically hurt their performance even if the hardware was suddenly only
one tenth as powerful.

I don't think the hardware helps as much as you think. It's just the
classic technology curve. The stronger your program is, or the faster the
hardware is, the less effect there is if you have even faster hardware.
Diminishing returns.




Wrote the chess? What does that mean?

I take it you are reminding us that although BH
often gets the credit, there were actually three men
involved in the creation of Cray Blitz, not counting
the multitude responsible for the development of
the Cray itself. Careful here; one false step and you
might incriminate Bob Hyatt, who has always insisted
that the chess algorithms could not be extracted from
CB, no way, no how.

Extracted, sure. Efficiently used on a micro, no way.

To sum up: the chess algorithms used for Cray Blitz
were specifically written and optimized to exploit the
killer advantage of its speed and power. They were
not vastly-superior-chess-algorithms which just
happened to have been written for a very fast machine,
whose vast superiority was merely an added bonus.
:D

Not quite that simple. You write your program for what you have.

Whether it's an 8 bit micro or a 32 bit PC, or a 64 bit workstation or a
vector computer.

You write your algorithms for what you have. It's not that some algorithm
is inherently better, it's just that it works better on the system you are
using.

Here's a simple example. The hash table. Relatively trivial.

On a regular system, you look at a specific entry and then probably the next
one or two. You use some simple method to decide which to replace.

But on a vector system, you can look at a bunch of entries, and then you can
step through with an increment other than one.

Is it superior? Maybe a little. Perhaps 1%. (Just a guess.) Why was it
chosen? Because on the vector hardware of the Cray, it was cheap and easy.
Easy to program and it improved the performance at least a little. On the
PC, it'd be a killer. Memory bandwidth and latencies would kill it. It
would be an absolute disaster.

He picked an algorithm that worked well on his hardware. If he had used
other hardware, he would have picked different algorithms.

Here's another example, from more modern times.

Bitboards vs. mailbox. Lots of people use bitboard's these days. They
don't think anything about it. But you wouldn't want to use them if you
were writing for an 8 bit micro or even under 16 bit DOS.

You choose the algorithms for the system you are developing on.

I may be running my program on the latest Core 2 Quad, but if I'm running
under 16 bit DOS, then my algorithm choices are limited. Bitboards would be
a bad idea. A mailbox would be the right choice. You pick your algorithms
for the system you are using. It's not that one is automatically better
than the other, it's just that it's better for your situation.



However, the diminishing returns in chess searching also plays a role
here.
Your computer may be 16 times faster but you'll probably only be
searching a
couple ply deeper.

With vastly superior hardware, it is possible to
have a combination of brute-force and selective
search methods, and here one might expect a
bit more than a couple of additional plys. In fact,
at tournament speeds (which is how these
programs got rated) the addition of even two
plys back then should have given the Cray a
big edge in head-to-head competition. Granted,
those games lasted for hours, and so there arose
the aforementioned technical issues.

Modern programs can get an extra ply for every 2-3 speed increase. But at a
cost in complexity and some risk due to the extra pruning.

Programs back then often took 5 or 6 times the speed to increase one ply.
(Although in later versions of Cray Blitz, he did implement more modern
forms of pruning, so that did start to drop down. I don't know what it'd
be, but perhaps as low as 3? Meaning 3x faster hardware gave an extra ply.)


You don't need modern hardware (or even Crays) to have some form of
selective search. Expletive!! Razoring and futility pruning and null
moves were being done back in the days of 8 bit micros!


And when you are searching so deep and are so strong, a couple extra ply may
not be that big of deal. It's the quality of those plies. The knowledge in
them. The kind of extensions used. Not all plies of search are of the same
quality.




Similar comments apply to Deep Thought and other programs
"accused" of "simply" being fast. Other things being equal, faster
is better. But other things are not equal, and there are seriously
bright people behind the successful programs.


I agree. But sometimes these seriously bright
people do seriously stupid things. Take DeepBlue,
for instance. In one match it went out-of-book
after 1.e3, losing like a fish. When things like
this happen, it is to be expected that critics will
leap on the blunder, and pointing out that the
fellow in charge had a Ph.D. is of little value; you
can't put lipstick on a pig.



For example, with today's PC's, many chess programmers (including Bob)
have
whole 'farms' of computers playing games against each other as a way to
test
modifications. That's something you just can't do with a Cray or any
super
computer or specialized hardware.

But many issues go deeper than merely tweaking
the program. For all the time wasted on fine-tuning
the opening book to optimize results against the

Not just fine tuning the opening book, but tweaking the values of the
evaluator terms and adding / removing / modifying evaluator terms.

And yes that does include openings without books, since the oponent might
make a new move or for some reason the book not being available.


current competition, one might have instead devoted
some time to "teaching" the program to play the
opening properly on its own. For all the time wasted
in battling adversaries on the net, a fellow could
instead have developed an entirely new approach
to analyzing the game. For example, Cray Brute
Force, and Cray Selective Search, and Cray Blitz,
and Cray Infinite Analysis. Each of these might
be optimized specifically for their respective tasks.

Easy to say, darn difficult to actually do.

It's easy for a non-chess programmer to make such statements.

Trust me on this, if it was as easy as you seem to claim, don't you think
everybody would be doing it? Heck, don't you think even 5% would be doing
it??

Making a smart selective search program is pretty hard.

Make a program that actually truely learns is even harder...




Logic would seem to indicate that were the
conditions right, the best PC programmers might
switch places with these "geniuses" and we
would see a different story; perhaps a widening of

Sorry, but that's not quite true. Different people think differently and
are better suited to different style hardware, etc.

Wow. So you think that if we took the one guy
who had access to a Cray supercomputer and
"switch him" with the absolute best chess programmer
on the PC, there might well be no improvement
whatsoever? That's amazing. It's akin to suggesting

Not as much as what you think.

A mailbox program wouldn't run that much faster on a cray than it did on the
PC hardware.

You have to think differently for different hardware and methods.

Ever try going from a mailbox style chess program to a bitboard style
program? It's a whole new way of thinking. You have to basically relearn
chess programming because otherwise you keep doing the same old style. You
might be using bitboards, but you will still be thinking in mailbox terms.
It can take years to really make the mental shift.


that if I were to jump out of my car and let A.J.Foyt
get in and drive, he would not get to McDonald's any
sooner than I would driving his race car, because we
think differently? So in other words, he's not really
any faster a driver than I am; we just drive different
cars. Who knew? :D

Actually, he probably isn't that much of a better driver in city traffic.
It's a different type of traffic than going around an oval.

Lots of race car drivers have said that it's easy to drive 200mph on a race
track. The hard part is to do it in traffic when you are 6 inches from the
guy next to you.

That does take skill, but that doesn't mean he can drive New York city
streets any better than anybody else. Different skills, different types of
traffic.





----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----

05.05.2007 21:13, raylopez99:
I recall reading they used human players early on.

Maybe, but at least since the mid 90s they are using pure computer
games. So, you favouring SSDF over some other ranking list, based on
the notion that SSDF uses games against human players is at least outdated.

Greetings,
Ralf

On May 6, 12:20 am, "Hello" wrote:

First, removing an opening book is easy. Trivial. Just don't have the book
file when you run Crafty.

The point as to how it relates to this thread is that
IM Innes has, for years, been asking *the author* to
do this and then test the program against a wide
variety of players, to see how it fares. Apparently,
it is Bob Hyatt who has the power to do this, not
any nearly-an-IM.


The problem is that most programs have considerable difficulty playing
reasonable openings. The opening moves have consequences that are too far
into the game for programs to be able to usually determine.


Exactly. So then, chess programmers have an
Achilles' heel here. It requires careful "protection"
from the elements.


The reality is that at this point of computer chess, *nobody* knows how to
truely program a chess program to learn chess openings even vaguely like
humans do.

Spoken like a true, lazy programmer. Obviously,
if a computer can be taught to play the middle-game,
it can also be taught to play the openings. It is
merely a question of wanting to try. I suspect that
when things begin to get very lopsided in favor of
computers, somebody will do this just to put things
on a more even footing, rather than just continue to
grind humans into pulp.


Two points though... First, many human players learn openings by rote.

I see. Humans do it, so it must be right. We also
overlook mates on the move, so why not write chess
programs which occasionally do this, too? Oh, and
some humans (many, perhaps) cheat. Let's just go
and incorporate that into the program, too. Why
bother trying to improve?


Second, computer chess hasn't been considered AI since the days of Mac HACK
VI, 30 years ago.

Please rebate all government-awarded funds to
this account immediately: .
If you can't come up with the money, just send me
your computers and chess books. Thanks!


The early versions were basically expanded versions of his old Blitz
program. Hyatt was depending on brute force number crunching and massive
hash tables.

This is why certain critics were quick to point out
where the program played like a dimwit, despite its
crushing speed. CB vs. Belle, 1981 would be an
example.


Later on, he was still taking advantage of that, but he had gone beyond that
and developed a lot of ideas and algorithms, many of which happened to be
tuned specifically for the Cray architecture.

Well, every article I pull up makes conflicting
claims. They will maintain that Crafty is the
"direct descendant" of Cray Blitz, and that CB
was translated from Blitz, but at the same time
these articles will claim that Mr. Hyatt wanted
to "start over", to try "something new". In sum,
they can't seem to make up their minds one
way or another. Put differently, it looks as
though BH wants to have it BOTH ways, and
as you can imagine, this sort of prevaricating
is precisely why he had so many adversaries
in the newsgroups. It's like a politician, in that
the "correct answer" is the one he thinks the
audience should hear.

An article I just finished reading had Mr. Hyatt
whining about getting paired as Black against
what he maintained were "all of the programs
expected to do well". Now, while this sort of thing
might help each of them individually, I think that
would tend to get balanced out by the fact that, by
definition, this means that his program gets White
against the others. Typically, these Whites were
never even mentioned, as if *they* didn't count.

That event was the 2005 tourney, and
unfortunately, the article did not cover the 2006
event. (Odds are one-in-two that if Mr. Hyatt did
not win that event, he had yet another excuse
handy.)


So by that time he was depending on the architecture because that's what he
had coded for. Changing architectures would have meant a complete rewrite
and re-think for his algorithms. It's pretty much impossible to seperate
the success and the Cray architecture. They were too tied together.

Here's what I have to say for his algorithms: if
in the game against Belle in 1981, a Cray super-
computer cannot see the obvious Bxh6, then
something is seriously amiss. There's a lot of
electricity being wasted on having a dumb brute
spin aimlessly in circles at blinding speed. :D

Maybe I'm wrong. Maybe hash-tables had not
been invented yet, and a lousy six plys was
state-of-the-art back then. Maybe it didn't see
further because it was thought that brute force
was the best way to exploit the Cray -- I don't
know. But I do know that any human fish
could have seen Bxh6 in a flash, including me.


As for Cray Blitz' success depending on the Cray performance... (shrug) All
you really have to do is pretend it was running on a very slow Cray.


Um, this invariably leads to moaning and groaning
from the Bob Hyatts about how they were unfairly
cheated by Fate because, well, because he somehow
*deserves* a big advantage in speed. (Something
about being descended from royalty, I suppose.)


You can't seperate the Cray out of CrayBlitz, but you can reduce the performance
of the hardware.

I would prefer to improve the software.


The good old 'Technology Curve' shows when the hardware gets "x" times
faster, you'll search a little deeper and your rating will increase.

According to one article, Cray Blitz faltered at tournament
time controls because it was not looking deep enough to
see that Bxh6 (the obvious move) was superior to getting
killed. This was in spite of it being the fastest machine.


The exact shape of that curve is very disputed, but everybody agree's that
it's not linear. The faster you go or the stronger you are, the less change
there is.

A super slow Cray would probably have only effected CrayBlitz by a 100 - 150
points. (My guess. Pure speculation since there's no way it can be proven.)
It might not have been quite the winner it was, but it wouldn't have been a
push-over either.

Well, if all you have is speed and they take much
of it away during a crucial game, you're up a creek.
(But BH never admitted that all he had was speed.)


My, how things change! From what I could see,
he not only had time for this, but also for engaging
in innumerable arguments in the chess newsgroups
until it reached a point where he was bordering on
paranoia, or a severe persecution complex.

Hyatt is still very active in chess forums. Just not here because there are
too many trolls and too much spam.


So now you're saying he does have spare time.

Many of Mr. Hyatt's supporters in the newsgroups
advised him to stop wasting his valuable time and
spend it instead on improving his program. Of
course he ignored them, and continued trolling the
people who, for instance, took strong exception to
his commentaries on the proper handling of certain
criminals. He seemed to be having a lot of fun,
but strongly resented the fact that he was not just
blindly worshiped by everyone for the great success
of Cray Blitz.

An old saying goes: friends come and go; enemies
accumulate. And in Mr. Hyatt's case, this seemed
to be a major problem. Some perceived him as having
an unfair speed advantage, while others were put off
by his abrasiveness, his argumentiveness, or both.


(frown) Yes... but what does that have to do with what I said?

Not everything written here is related to what you
say.

My comment simply means that when entering
these events, Mr. Hyatt should have taken this
into consideration. If he cannot secure adequate
computer time, there is always the alternative of
figuring out precisely when the Cray is not in full
use, and scheduling matches against those who
would be willing to accommodate him. Obviously,
making use of a remote computer has its
disadvantages, just as lugging around a portable
machine has its own. (This is precisely how my
old notebook computer broke, as an example.)

I am envisioning a rated match between Cray
Blitz and Art Bisguier, to be followed by one
between CB and any number of stay-up-all-night
IMs, whose schedules were not nearly so rigid.

Oh well. These kinds of events would not
garner the same level of attention as playing
in a USCF-rated tournament, and presumably,
winning it.

-- help bot