On May 4, 9:35 pm, raylopez99 wrote:
On May 4, 7:41 am, Martin Brown
wrote:

Crafty19 is somewhat more balanced than Fritz8 about choice of
continuation lines, but it looks a bit out of its depth when asked to
rate super GM games that it really doesn't understand (certainly not
at 12 ply).

True enough, but how often do GM games get decided by anything other
than simple tactics? That is the question.

About 96% of the time if the statistics in the paper that you
originally quoted. Less than 4% of GM level games are determined by a
tactical exploitation of an unforced human error or blunder at this
level. Most are determined by a steady buildup of pressure on a
strategic weakness that eventually results in a lesser of two evils
choice.

Kramnik missed a mate in
one in his last man-machine match and you're worried about plys
greater than 12+?

Ply 12 is only 6 moves for each side. It isn't even enough to annotate
club player games reliably. It will only find the most obvious
blunders. Super GM level play from engines is only possible at around
the 18-20 ply mark.

The fact that world champions cannot do better than
draw (Kramnik) or lose (Kasparov) to chess programs gives lie to the
idea that GM games are decided by profound chess positional sacrifices
that chess computers don't understand.

If a human makes any kind of mistake against a chess engine no amount
of bluffing or complicating the position will work against a computer
it will just hasten your demise. But against a human a very strong
player can regain control of a lost or drawn game by playing the odd
dodgy move provided that their opponent can be out psyched OTB.

Also when compared to stronger engines Crafty19.01 (which
is what I have most experience with) shows a tendency to exaggerate
the badness of some of the weaker moves from a given position.

I revise that statement. There is something fundamentaly wrong in
Crafty 19.01 when run for periods in excess of 24h. The answers it
gives at ply 18 for some lines are bad and getting worse - more than
80cp wrong.

currently testing a series of engines on a slightly tricky French
Advance Variation position where powerbooks considers only 7.Qd2 as
playable.

And in OTB play probably 7. Qd2 would be the best move.

Although against club players even 7. b3 is playable (until recently I
had won with it).

If anyone else wants to try running the position for 48 hours and
noting down the move evaluations at all plys reached beyond the 4
minute mark I would like to see what Fruit, Comet and Deep Sjeng make
of this position. Infinite analysis, top 10 lines displayed (and also
the actual top 10 preferred lines found after 48 hours).

I see the problem with Camp #1--they are trying to play the 'perfect'
game of chess--perhaps correspondence chess. That is not the
inquiry. The inquiry is: which human champion is the best OTB
player? Using PCs we can determine this by finding which human GM
made the fewest tactical mistakes. It's true that Crafty will fail to
score certain profound positional sacrifices correctly, but as I have
repeatedly stated in this thread, probably such positional sacrifices
are rare in chess--chess is 99% tactics. Chess is not checkers, but
it's close enough.

Chess is nothing like 99% tactics you are completely and utterly
wrong. It doesn't matter how many times you repeat this idiotic
pronouncement it will not ever be true. If it had been true Chess
software would not have taken so long from Samuels checkers program
reaching world championship level to an equivalent for chess.

One of the main reasons that Deep Blue spooked Kasparov was precisely
because the chess engine found a long term strategic move that up
until then would have been completely beyond any chess software. He
could not believe that a computer could do this and acused IBM of
cheating. These days you would be hard pushed to find a commercial
chess program that *cannot* reproduce the same critical move.

That's why Go and Bridge are considered the last
bastions of human game creativity (though those bastions will
eventually fall as well, once AI is further perfected).

Go requires vastly more strategic planning than chess to overcome the
huge growth in the game tree and is beyond current technology. AI is
all jam tommorow. Perhaps when we have hardware pattern matching
architectures that are up to the job we will see a program that can
take on top human Go players. That will still take a while yet.

Regards,
Martin Brown

On May 4, 10:08 pm, raylopez99 wrote:
On May 4, 5:09 am, "Chess One" wrote:

I am posting this out of order because as far as I can tell Google
will not let me reply to ChessOne's postings. If you can see a copy of
this post attached to ChessOnes response to the post I am replying to
please let me know privately.

On May 5, 2:44 pm, "Chess One" wrote:
"raylopez99" wrote in message

oups.com...

On May 4, 5:09 am, "Chess One" wrote:
Suppose that a version of Crafty favors "defensive" players like
Capa
and Kramnik, while penalizing "attacking" players like Tal and
Fischer. Call this version of Crafty "Defensive

Interesting post, and the para above identifies the flaw in
comparisons
of
player-to-player by engine analysis- that if you are evaluating
players,
then even Tal himself said that anybody at all could find his own
flaws
after the game, or the next day, or even next week. But he chose 'em
because
very few people could find them in real time OTB.

Yes, but this problem is also present in computer chess! Even if the
program is 'backsolving' or anotating a completed game, it has to look
in the chess tree, which means that it faces time constraints similar
to an OTB player vs a 'time is not of the essence' correspondence
player.

Quite so, Ray - in order to go to deeper plies and find a solution the
program must find a way to prune possibilities - and it does this by
scoring
each branching possibility on, say, ply 7 or 8, and if worse than
nominally -2 will prune that line [which may have been eg, a Knight sac]
that can perhaps offer a return of 3 pawns and the initiaitive by ply
15 -
or even something more compelling.

I doubt that any of the modern programs use anything so crude. They
appear to use recursive alpha-beta with refutation lines tried first
for each possible move together with killer moves for each ply and non-
speculative forward pruning as a base level. After that they use
various forms of selective extensions on the most promising looking
variations followed by a quiessence and swapoff rule. You only have to
watch how fast they rip through the no hoper lines at very deep levels
when compared to something like Crafty or Comet that lack some of the
extra go faster tricks.

I did not realize pruning is employed at -2 pawns. Thanks for this
information. I assumed the cutoff was more like -3.

You may be right on the default value - but the issue for the computer is
that if you optimise ply-depth, then you effect pruning score, no? Otherwise
in the amount of time available for the move there is no way to achieve
sufficient ply searching.

If this was the case engines would never find queen sacrifices that
lead to a mating net. Hint: they do.

Yes - one of the MAMS recommendations is to pick a long line, preferably

For those of us on the other side of the pond: Who or what is MAMS ?

I surmised that MAMS must be the book mentioned in this thread on how
to beat Fritz, to be published. I recall a rather simplistic book
called "How to Beat your Computer in Chess" about 10 years ago that
was not that useful for me. Hopefully this book is better. At any
rate, I doubt that in an actual OTB game with fixed time constraints
you will find the opportunities to play some brilliant positional
sacrifice--you'll be mated tactically before you ever get to these
wonderful winning positions.

A have one draw against Rybka, but older programs could be beat by
positional play, time and time again. With book off no amount of tactical
finesse works. I wrote in about 1998 that Crafty with book off was maybe
1800-1900 level.

What version of Crafty was it then?

I am currently benchmarking Crafty 19.01 on a postion from the French
Advance Variation that came up in a game of mine. I noticed that
Fritz8 made a bad fist of it and thought it might shed some light on
the thinking of other engines. Crafty is now at ply 18 after 47 hours
and its evaluations are actually getting *worse* with each increasing
ply. The worst case error in Crafty when on this position compared to
other much strong engines at 18ply is about 80cp.

I suspect there is a bug in it that is hampering its ability to play
compex opening positions. Whatever else we disagree on about computer
chess I think we should both agree that increased search depth should
generally improve the accuracy of move evaluation. And that is not
what I am observing testing Crafty 19.01.

I will test Crafty 20.xx next. I am also running the same test on my
newly purchased Rybka. Its results are a bit off the mark in absolute
value (ie it slightly dislikes a position good for white) but it
relative rating of the continuation lines is very plausible and
improves with each successive ply - and the evaluation results from 4m/
move to a few hours/move do not deviate abruptly unless it finds some
interesting new line at depth.

However, I will give Camp #1 credit in
one respect--they are essentially arguing how a human can beat a PC in
*CORRESPONDENCE CHESS*. The procedure is as follows: (1) fill a
database of positions where a positional sacrifice not within your
opponent computer's move horizon leads to victory, (2) play a series
of forcing moves that lead to the positions in step (1), and, (3)
execute steps (1) and (2) to win.

I see. This is very similar to beating people, and the central idea of the
postional player.

And to beat a chess engine you want to avoid positions where there is
a single forced line leading to victory. If just one move is massively
better than all of the others or a check is used it will trigger
singular extensions, but if 2 or 3 moves are similarly good the engine
will not extend the search the extra ply until it is too late and the
forcing line is already entered.

If this is what you seek, good luck to you, as you might find
"steering" the opponent program ala step (2) is a lot harder than you
think, because I posit these positional sacrifice positions will be
very far and few between in the chess tree--because chess is 99%
tactics.

Chess is how you play it!

If chess were 99% tactics chess engines would have been world chess
champions much sooner. It is the positional long range strategic
planning that computers still need human guidance on - that is why
freestyle chess is so interesting.

The situation is exemplified in the game of Go where the basic rules
are simpler but the long range planning and huge multiplicity of moves
on a full size board put the top level human game out of reach of
current computing methods.

What is intellectual, is that programming-as-we-know-it, is now semiotically
almost a technology rather than anything to do with science. Many

Computing was initially viewed as an impure form of applied
mathematics.

OK, if I understand your point here--you're saying that Fritz in a few
seconds basically is "right onto" the correct analysis that 50 years
of amateur and GM effort have found. Doesn't that support MY point,
of Camp #2? That computers _can_ do an excellent job of analysing
positions, and not just tactical analysing? Sure seems like it.

In this instance, but in other instances computers are really dumb! So I

Actually computers can analyse some positions well and some *very*
badly. And different engines have different strengths and weaknesses.
Fritz8 for instance can be relatively easily blinded to other
possibilities by offering it what looks like a simple equality swap
off sequence with a sting in the tail. It is very surprised when you
don't take the last pawn or whatever is on offer. Shredder is much
harder to mislead in this way.

There is another point here. A computer can do a lot better annotating
a complete GM level game than playing one or even evaluating a given
position without knowing how play continued.

Calculating ab initio it has no guidelines and must figure everything
out for itself which is what caused the original paper that started
this discussion to use a fixed depth 12 ply search in Crafty. This is
well short of what is needed even with much stronger commercial
engines to annotate ELO 2000 club player games reliably. So it stands
no chance at all of giving the right answer on super GM and world
champion games (except for finding medium term tactical blunders).

Calculating backwards from a game with a known outcome it already has
a framework for to hang its analysis onto and can preload the
transposition table cache with relevant future positions. Provided the
GM play is good this massively accerlated cutoffs in the alpha-beta
tree and it can see a lot further into the game tree in a given time.

Now - here it is - Fritz is no better at refuting this maddening line

1 e4 e5
2 Nf3 Nc6
3 Bc4 Nf6
4 Ng5 Bc5
5 Nf7 Bf2

than 50 years of players, BUT - if you want to know the secret of beating
this Traxler line with its improvident 4. ... Bc5 then a few 'moves by
hand'
or MAMS moves, 8-13, allows Fritz to romp home with a solution to win for
white in 20

Are you going to show the line and the position where Fritz baulks at
the right move?

I would hazard a guess it is a pawn grab trap or an apparently equal
swapoff that isn't.

basis which underlines these issues need more than quantification [brute
force in this instance], and you do not seem to understand the role that
qualification plays in programming chess.

Since you apparently have a copy of Rybka have you not noticed that it
is moving in the direction that you claim to support. It looks at
around 40kn/s on a 3GHz machine compared to Fritz 1000kn/s or Shredder
300kn/s. Looking at a factor of 25x fewer nodes with a massively more
complex terminal node evaluation function is a significant change.

Shredder10 is the engine I use which can usually can go deepest, but I
estimate that Rybka2.31 which I am benchmarking at present will
despite its much slower evaluation function overtake Shredder at about
72hours runtime.

Shredder10 reached 36/36 ply 16 at 4 mins and 5/36 ply 24 at 60h
Rybka2.3.1 reached 36/36 ply 14 at 4 mins and 9/36 ply 23 at 48h
Fritz8 reached 36/36 ply 14 at 5 mins and 3/36 ply 20 at 48h
Crafty19.01 reached 9/36 ply 12 at 7 mins and 9/36 ply 18 at 48h (and
results are junk)

Rybka is currently on target to reach 2/36 ply 24 by 60h unless it
slows down. Very impressive so far.

Crafty is very unlikely to finish ply 18 inside 60h but it will get to
run for at least another 10 hours or so because I want to see what
weird answers it will give for the 9 & 10th strongest lines.

You are conversing with an intellectual, but you don't know why that is
necessary - which is like many programmers whose efforts seem like a
complete waste of time, since they have not sufficiently bench-marked their
activity in the classical way to determine even if what they do is
redundant, and all increased in the programmers performace relate to
bus-speed in the CPU

You know that isn't true. Why do you repeat this lie?

This 'intellectual' question is not how fast the program can go, but in
which direction it will go? - and is something of a refutation of brute
forcing a result, since the logic of it is that increased ply depth is
acheived only by reduced evaluation pruning - and the engine will not look
at a -3 score any longer even though in 20 plies that variation wins.

I don't believe any program uses such a crude pruning method. Perhaps
you do know of one. If so name it!

I think that the top ones are using speculative forward pruning and
branch and bound tricks. But they are way more sophisticated than you
are claiming. I am pretty sure that is what Shredder has done when it
returns strange numerical evalutions like #180 or 278.41. It can see
something devastating in the future(either fail high or fail low) but
not well enough to give it a normal score.

Regards,
Martin Brown

PS Apologies if there are multiple copies of this. But it doesn't show
up at all on either of my NNTP servers or on Google despite having now
been posted 3 times. I am hoping for third time lucky!

On May 4, 5:48 pm, (Dr A. N. Walker) wrote:
In article .com,
help bot wrote:

[...] If you want something reproducible, then, short of
help from the commercial companies, you have to do a fair amount
of tweaking.

Agreed. Although reproducibility is obtained at the expense of
accuracy. I can think of plenty of my games that were annotated by a
stronger engines than Crafty at 14 ply and where significant errors
were not found by blundercheck or deep positional analysis until ply
16 was reached.

Nonsense. Every tweak you make to Crafty is
a step toward irreproducibility, unless you expect
everyone to go in and modify their source code to
precisely match yours [...].

I don't *expect* anyone to do that. But anyone who might
want to check my results [or, in the present case, *G&B's* results]
*can* so modify their source code, and can thereby get exactly the
same results, even on different hardware, or can, if they choose,
apply exactly the same methodology to a different collection of
games, or can, if they choose, apply a different reproducible
methodology to some games and compare the results.

The one which would seem to make most sense is to sacrfice uniform
search depth coverage in favour of the deepest possible analysis in a
given total elapsed time per game. That is effectively equivalent to
blundercheck with 1cp window and then parsing the resulting annotated
games. This could be applied to other stronger engines too.

What do you
mean by "reproducible" and how would you get the above from [eg]
Rybka?

I expect it would require some help from the author unless there is a
way to make the evaluation function exactly deterministic (and there
may be for testing purposes).

A better method would be to take the strongest
program available (i.e. Rybka) and have it score
all the games, and only THEN apply some selected
algorithm to the standardized results.

And if I did the scoring again tomorrow, with a different
"seed" to the RNG, or with different contents in the cache/hash,
or perhaps after adding a new disc or RAM to my computer, let
alone if you tried to repeat my experiment, would you expect to
see the same results?

Not exactly the same results but close enough agreement to be
convincing.

A brief search didn't reveal what the actual hardware for the
recent Kramnik-DeepFritz match was, but from the node-counts quoted
it seems to have been roughly equivalent to 6x 2.5GHz PCs.

[...] The point is, my
guess is that this desktop passed up the 12th
ply in just a few seconds,

Note that neither ply counts nor node counts are very good ways
to compare one program with another [OK for comparing the same program
on different hardware, of course]. 12-ply Crafty, 12-ply Fritz and 12-ply
Rybka are quite certainly seeing different things [as well as applying
different evaluations to them].

Indeed. One serious worry I have after benchmarking so far Shredder10,
fritz8, Rybka2.31 and crafty19.01 is that there is some kind of bug in
Crafty 19.01 that compromises improvement with increasing depth. It
only really becomes a big issue on moves that other engines rate at
50cp below the top line in the test I have done. It is possible that I
have hit upon a position that by chance happens to break Crafty but
then it only takes one counterexample.

It's really a matter of standards; are you willing
to accept half-baked analysis? Half-reasoned
justifications for dodgy work?

The half-baking and the half-reason seem to be based on
the semi-pop version put out by Chessbase. You may [well] not
agree with the baking and reasoning of the full paper, but some
reasonably qualified people will have refereed it [both for ICGAJ
and for ICCG5] and were presumably happy. ICGAJ publishes letters;
if you think their standards are slipping, write to them.

I think it probably is worth making the point that analysing the games
from the final position and moving backwards down the game tree will
enable an engine to exploit all the information available to it and
produce a much more thorough and compelling analysis of the move
strengths, but at the expense of uniformity of search depth. In
particular it will have very deep evaluations from the hash tables if
the GMs actually played along the line that the engine thinks is best.

Whether the resulting
investigation is worthwhile is another matter.
Indeed. My take is that it is silly to think a
computer (at this point in time) can accurately
determine who among the world champions was
"the greatest".

Then perhaps it's lucky that G&B weren't trying to determine that?

I think their attempt was very interesting but flawed because Crafty
at fixed ply 12 + quiessence would score too many club games
incorrectly never mind super GM level play. Limited to finding
blunders of 50cp or greater inside its search horizon I suspect it is
perfectly OK. It would help imrpove confidence to see the results for
10 ply and 11ply.

Regards,
Martin Brown

On May 6, 1:09 am, help bot wrote:
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.

Only once the other engines figure out counter measures against
Rybka's somewhat radically different self consistent complex
evaluation function at a (by computer standards) relatively small
number of nodes. This may take some time.

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.

I don't doubt that all the top engine makers are now tuning their
offerings to enable them to beat the current top dog. That is pretty
much how the art advances. It isn't all that long ago that people
published chess positions that chess computers were expected *never*
to understand. These days engines like Shredder solve them pretty much
instantly.

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.

Please can you provide a position from this game where Rybka can be
shown to be the only engine that will get the right answer? It would
be interesting to find a new modern set of test positions that can
discriminate between engines.
Test pieces that only one engine can do correctly would be the ideal.

Regards,
Martin Brown

"Martin Brown" wrote in message
ups.com...

Quite so, Ray - in order to go to deeper plies and find a solution the
program must find a way to prune possibilities - and it does this by
scoring
each branching possibility on, say, ply 7 or 8, and if worse than
nominally -2 will prune that line [which may have been eg, a Knight
sac]
that can perhaps offer a return of 3 pawns and the initiaitive by ply
15 -
or even something more compelling.

I doubt that any of the modern programs use anything so crude.

I am not specualting.

They
appear to use recursive alpha-beta with refutation lines tried first
for each possible move together with killer moves for each ply and non-
speculative forward pruning as a base level. After that they use
various forms of selective extensions on the most promising looking
variations followed by a quiessence and swapoff rule. You only have to
watch how fast they rip through the no hoper lines at very deep levels
when compared to something like Crafty or Comet that lack some of the
extra go faster tricks.

Whether in fact these devices do the trick is examined in the MAMS book - at
least it can be said that Fritz suffers from it - and as above, if an amalgm
of advanced factors don't actually work, then these are much ado about
nothing.

I did not realize pruning is employed at -2 pawns. Thanks for this
information. I assumed the cutoff was more like -3.

You may be right on the default value - but the issue for the computer is
that if you optimise ply-depth, then you effect pruning score, no?
Otherwise
in the amount of time available for the move there is no way to achieve
sufficient ply searching.

If this was the case engines would never find queen sacrifices that
lead to a mating net. Hint: they do.

I can tell you the test game - the original. And it was played by GM
Bronstein against a chess computer! In fact it has become a stand test of
look-ahead efficacy, according to Bob Hyatt.

If the mate occurs within the look ahead range then all is okay - but if its
not mate, but reversal of fortune after 15 moves, that is a noted weakness
if the evaluation only shows up significantly positive in 15 moves. For the
interim 14 moves the program might have been showing maximum loss [or risk
factor] of [?] for Fritz -4.33

The question is in chess terms, not programming terms, when does the program
cut its loses when its chances are flat-lined?

Yes - one of the MAMS recommendations is to pick a long line,
preferably

For those of us on the other side of the pond: Who or what is MAMS ?

The acronym means Man Assisted Machine Chess - but MAMC is ugly, so Chess
becomes German/Dutch as 'Schack'

---
A have one draw against Rybka, but older programs could be beat by
positional play, time and time again. With book off no amount of tactical
finesse works. I wrote in about 1998 that Crafty with book off was maybe
1800-1900 level.

What version of Crafty was it then?

I don't know, the highest version I could then download. I have a more
current version which comes with Rybka.

I am currently benchmarking Crafty 19.01 on a postion from the French
Advance Variation that came up in a game of mine. I noticed that
Fritz8 made a bad fist of it and thought it might shed some light on
the thinking of other engines. Crafty is now at ply 18 after 47 hours
and its evaluations are actually getting *worse* with each increasing
ply. The worst case error in Crafty when on this position compared to
other much strong engines at 18ply is about 80cp.

I suspect there is a bug in it that is hampering its ability to play
compex opening positions.

Yes, again in conversations with Bob Hyatt, he once said me that Crafty
would never play the 3rd move in the Ruy, ie, Bb5. Much preferring 3. Bb4 ~
since the Ruy is pretty open, that is a less extenuated example.

Other programs have trouble with openings which seem to require White not
exchanging in semi-open positions. One example is the Armenian French, when
the black bishop retreats to a5 instead of, acter being challenged with a3,
capturing the c3 Kt. The B can then be further challenged and winds up on
c7, when the White knight can capture it! - but that is where the program
seems to go nuts, as if its compass was spinning and it can't make up its
mind if it really likes or dislikes subsequent positions. In this instance
this is not a fault with Crafty alone, but in semi-open positions ~ the
typical gaining of the 2 bishops at the cost of tempii.

Do you have Rybka so you can run it at the French Advanced?

Whatever else we disagree on about computer
chess I think we should both agree that increased search depth should
generally improve the accuracy of move evaluation. And that is not
what I am observing testing Crafty 19.01.

Yes - I think we are in agreement, though a useful term to consider is by
the author of MAMS, Alberts, who talks of 'risk factors'. Which line to
choose if the next 10 moves render an = eval, while another line has
oscillated + then - then + then...

I gotta go - interesting post - shall try answer anything below later.
Cordially, Phil

I will test Crafty 20.xx next. I am also running the same test on my
newly purchased Rybka. Its results are a bit off the mark in absolute
value (ie it slightly dislikes a position good for white) but it
relative rating of the continuation lines is very plausible and
improves with each successive ply - and the evaluation results from 4m/
move to a few hours/move do not deviate abruptly unless it finds some
interesting new line at depth.

However, I will give Camp #1 credit in
one respect--they are essentially arguing how a human can beat a PC in
*CORRESPONDENCE CHESS*. The procedure is as follows: (1) fill a
database of positions where a positional sacrifice not within your
opponent computer's move horizon leads to victory, (2) play a series
of forcing moves that lead to the positions in step (1), and, (3)
execute steps (1) and (2) to win.

I see. This is very similar to beating people, and the central idea of
the
postional player.

And to beat a chess engine you want to avoid positions where there is
a single forced line leading to victory. If just one move is massively
better than all of the others or a check is used it will trigger
singular extensions, but if 2 or 3 moves are similarly good the engine
will not extend the search the extra ply until it is too late and the
forcing line is already entered.

If this is what you seek, good luck to you, as you might find
"steering" the opponent program ala step (2) is a lot harder than you
think, because I posit these positional sacrifice positions will be
very far and few between in the chess tree--because chess is 99%
tactics.

Chess is how you play it!

If chess were 99% tactics chess engines would have been world chess
champions much sooner. It is the positional long range strategic
planning that computers still need human guidance on - that is why
freestyle chess is so interesting.

The situation is exemplified in the game of Go where the basic rules
are simpler but the long range planning and huge multiplicity of moves
on a full size board put the top level human game out of reach of
current computing methods.

What is intellectual, is that programming-as-we-know-it, is now
semiotically
almost a technology rather than anything to do with science. Many

Computing was initially viewed as an impure form of applied
mathematics.

OK, if I understand your point here--you're saying that Fritz in a few
seconds basically is "right onto" the correct analysis that 50 years
of amateur and GM effort have found. Doesn't that support MY point,
of Camp #2? That computers _can_ do an excellent job of analysing
positions, and not just tactical analysing? Sure seems like it.

In this instance, but in other instances computers are really dumb! So I

Actually computers can analyse some positions well and some *very*
badly. And different engines have different strengths and weaknesses.
Fritz8 for instance can be relatively easily blinded to other
possibilities by offering it what looks like a simple equality swap
off sequence with a sting in the tail. It is very surprised when you
don't take the last pawn or whatever is on offer. Shredder is much
harder to mislead in this way.

There is another point here. A computer can do a lot better annotating
a complete GM level game than playing one or even evaluating a given
position without knowing how play continued.

Calculating ab initio it has no guidelines and must figure everything
out for itself which is what caused the original paper that started
this discussion to use a fixed depth 12 ply search in Crafty. This is
well short of what is needed even with much stronger commercial
engines to annotate ELO 2000 club player games reliably. So it stands
no chance at all of giving the right answer on super GM and world
champion games (except for finding medium term tactical blunders).

Calculating backwards from a game with a known outcome it already has
a framework for to hang its analysis onto and can preload the
transposition table cache with relevant future positions. Provided the
GM play is good this massively accerlated cutoffs in the alpha-beta
tree and it can see a lot further into the game tree in a given time.

Now - here it is - Fritz is no better at refuting this maddening line

1 e4 e5
2 Nf3 Nc6
3 Bc4 Nf6
4 Ng5 Bc5
5 Nf7 Bf2

than 50 years of players, BUT - if you want to know the secret of
beating
this Traxler line with its improvident 4. ... Bc5 then a few 'moves by
hand'
or MAMS moves, 8-13, allows Fritz to romp home with a solution to win
for
white in 20

Are you going to show the line and the position where Fritz baulks at
the right move?

I would hazard a guess it is a pawn grab trap or an apparently equal
swapoff that isn't.

basis which underlines these issues need more than quantification [brute
force in this instance], and you do not seem to understand the role that
qualification plays in programming chess.

Since you apparently have a copy of Rybka have you not noticed that it
is moving in the direction that you claim to support. It looks at
around 40kn/s on a 3GHz machine compared to Fritz 1000kn/s or Shredder
300kn/s. Looking at a factor of 25x fewer nodes with a massively more
complex terminal node evaluation function is a significant change.

Shredder10 is the engine I use which can usually can go deepest, but I
estimate that Rybka2.31 which I am benchmarking at present will
despite its much slower evaluation function overtake Shredder at about
72hours runtime.

Shredder10 reached 36/36 ply 16 at 4 mins and 5/36 ply 24 at 60h
Rybka2.3.1 reached 36/36 ply 14 at 4 mins and 9/36 ply 23 at 48h
Fritz8 reached 36/36 ply 14 at 5 mins and 3/36 ply 20 at 48h
Crafty19.01 reached 9/36 ply 12 at 7 mins and 9/36 ply 18 at 48h (and
results are junk)

Rybka is currently on target to reach 2/36 ply 24 by 60h unless it
slows down. Very impressive so far.

Crafty is very unlikely to finish ply 18 inside 60h but it will get to
run for at least another 10 hours or so because I want to see what
weird answers it will give for the 9 & 10th strongest lines.

You are conversing with an intellectual, but you don't know why that is
necessary - which is like many programmers whose efforts seem like a
complete waste of time, since they have not sufficiently bench-marked
their
activity in the classical way to determine even if what they do is
redundant, and all increased in the programmers performace relate to
bus-speed in the CPU

You know that isn't true. Why do you repeat this lie?

This 'intellectual' question is not how fast the program can go, but in
which direction it will go? - and is something of a refutation of brute
forcing a result, since the logic of it is that increased ply depth is
acheived only by reduced evaluation pruning - and the engine will not
look
at a -3 score any longer even though in 20 plies that variation wins.

I don't believe any program uses such a crude pruning method. Perhaps
you do know of one. If so name it!

I think that the top ones are using speculative forward pruning and
branch and bound tricks. But they are way more sophisticated than you
are claiming. I am pretty sure that is what Shredder has done when it
returns strange numerical evalutions like #180 or 278.41. It can see
something devastating in the future(either fail high or fail low) but
not well enough to give it a normal score.

Regards,
Martin Brown

PS Apologies if there are multiple copies of this. But it doesn't show
up at all on either of my NNTP servers or on Google despite having now
been posted 3 times. I am hoping for third time lucky!

help bot wrote:

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.

Not on my screen it isn't. It look perfect.

Pehaps you forgot that ASCII-Art requires a monospace
font such as Courier?

--------------------------------------------------------

ASCII-art stereograms:

Here are some ASCII stereograms. View with a monospace font.
To see the 3D effect, you need to focus your eyes such that the two
adjacent "X" characters at the top look like three. Once you've done
that, wait a few moments for your eyes to get used to it, then slowly
look down at the rest of the image.








X X
.. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . .
. . . . . . .
.. . . . . . .
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
. . . . . . .
. . . . . . .
. . . . . . . . .
. . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
.. . . . . . . . . . . . . . . . . . . . . . .






X X

h h h h h h h h h h h h h h h h h h
t t t t t t t t t t t t t t t t
t t t t t t t t t t t t t t t t t t
p p p p p p p p p p p p p p t t
: : : : : : : : : : : : : : : : : :
/ / / / / / / / / / / / / / / /
/ / / / / / / / / / / / / / / / /
w w w w w w w w w w w w w w w
w w w w w w w w w w w w w w w w w
w w w w w w w w w w w w w w w
. . . . . . . . . . . . . . . . .
g g g g g g g g g g g g g g g g
u u u u u u u u u u u u u u u u u u
y y y y y y y y y y y y y y y y
m m m m m m m m m m m m m m m m m m
a a a a a a a a a a a a a a a a
c c c c c c c c c c c c c c c c c c
o o o o o o o o o o o o o o o o
n n n n n n n n n n n n n n n n n n
. . . . . . . . . . . . . . .
c c c c c c c c c c c c c c c c c
o o o o o o o o o o o o o o o
m m m m m m m m m m m m m m m m m
/ / / / / / / / / / / / / / /

"Chess One" wrote in message
news:P9m%h.1960$s7.772@trndny05...


Martin, there is one other point here to which I can add a note which is a
specific about playing chess rather than generalities about play or the
components aspects of the program which may or may not be useful

- I set up a Russian Chess journalist [a GM] to annotate his own game [in
our series, this mean he had to annotate a loss] and provide below his own
comments - there is a certain level of play where the engine doesn't see
enough ahead and in fact at this level of play are not taken too seriously
/because/ as you will see in this game tactical considerations count for
very little, and a deep positional move seems to win the game, and in fact
calls into question the entire line - here the program is not used as much
for its computational ability, but for its database of possibilities - this
GM criticises himself for over-reliance on using the db, even though he
plays out various lines against his computer - he is playing inferior lines.

What this example might illustrate for OUR conversation is if any chess
engine can find what Bezgogov's could not. Here is the key text:

"13. Nh4 . I underestimated this plan in my preparation. White's knight
arrives at the very strong g6-square and its almost impossible to drive it
away from this station. The knight distinctly complicates Black's defense.
I must admit one more serious mistake - I played this endgame against one of
the strongest chess programs. My 'inanimate' opponent didn't find White's
13th move and, against any other play Black stands well and I would defend
Black's position easily."

(( entire game and commentary is at
http://www.chessville.com/LessonsLearned/2004Jan.htm ))

some other notes from the game:-

"GM Alexei Bezgodov: My loss to an excellent grandmaster from Moscow,
Evgeny Najer, one can say consists of two parts. First I prepared badly for
the game using an incomplete database and already from the opening got a
very difficult (maybe even lost) endgame."

A critical moment comes here
"10...h6?!

It's sad to assert that I have made a decisive mistake already on the 10th
move, and that was prepared as far back as home analysis! But certainly it
required Najer's correct play to prove the strength of White's position.
Most likely the best move here is a developing one, 10...Nc6, which is
preferred by an expert in this system for Black, the Champion of Europe-2003
Grandmaster Azmaiparashvili."

a few moves later he says

"I underestimated this plan in my preparation. White's knight arrives at
the very strong g6-square and its almost impossible to drive it away from
this station. The knight distinctly complicates Black's defense. I must
admit one more serious mistake - I played this endgame against one of the
strongest chess programs. My 'inanimate' opponent didn't find White's 13th
move and, against any other play Black stands well and I would defend
Black's position easily. I did not suspect the existence of the knight's
maneuver since my database didn't have an old game - Najer-Belikov! A
surprising 'unlucky coincidence of circumstances' where actually only my own
laxity was at fault - nobody prevented me updating the data base and
furthermore, nobody prevented me analyzing this endgame more carefully!"

he concludes after the game

"But as a professional I am obliged to reproach myself for: first, bad
preparation for the variation generally; second, for bad preparation after
solidifying this game to contain it; and third, that after several not bad
moves I nevertheless 'dropped' an already obtained draw!"

---

He is saying, to summarise the above, that his 10 ... h6 becomes refuted by
Whites 13. Nh4. And that, Martin, is where I think a current test might be
useful. Here are the preceeding moves:-

(1) Najer,E - Bezgodov,A [D24]
56th Russian Championship (3.13), 05.09.2003

1.d4 d5 2.c4 dxc4

3.Nf3 Nf6 4.Nc3! c5 5. d5

5...e6 6.e4 exd5 7.e5! 7...Ne4 8.Qxd5 8...Nxc3

9.Qxd8+ Kxd8 10.bxc3

10...h6?! ((Lets call this Alpha Point)

11.Bxc4 11....Be6 12.Bxe6 fxe6 13.Nh4! ((Lets call this Beta Point))

---

A) My assumptions are that (a) a program will play the Alpha point move at
9, or score it as a strong consideration if it (b) does not see the Beta
point move at 13.

B) Can your chess engine (a) find the Beta point move on its own? and (b)
how does it score that move in you hand-enter it, compared with scores for
other White options at 13?

Cordially, Phil Innes



There is another point here. A computer can do a lot better annotating
a complete GM level game than playing one or even evaluating a given
position without knowing how play continued.

Calculating ab initio it has no guidelines and must figure everything
out for itself which is what caused the original paper that started
this discussion to use a fixed depth 12 ply search in Crafty. This is
well short of what is needed even with much stronger commercial
engines to annotate ELO 2000 club player games reliably. So it stands
no chance at all of giving the right answer on super GM and world
champion games (except for finding medium term tactical blunders).

Calculating backwards from a game with a known outcome it already has
a framework for to hang its analysis onto and can preload the
transposition table cache with relevant future positions. Provided the
GM play is good this massively accerlated cutoffs in the alpha-beta
tree and it can see a lot further into the game tree in a given time.

On May 6, 9:44 am, Martin Brown
wrote:

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.

Please can you provide a position from this game where Rybka can be
shown to be the only engine that will get the right answer? It would
be interesting to find a new modern set of test positions that can
discriminate between engines.
Test pieces that only one engine can do correctly would be the ideal.

I downloaded several articles from the Rybka Web site
a while back, and the one under "articles" and then
"Evans Gambit" shows a game between GMs Kasparov
and Anand, Riga 1995, which begins:

1.e4 e5

2.Nf3 Nc6

3.Bc4 Bc5

4.b4 Bb4

5.c3 Be7

6.d4 Na5

7.Be2 ed

8.Qxd4 Nf6

9.e5 Nc6

10.Qh4 Nd5

11.Qg3

Now here, the author claims that Rybka's move
against Shredder, ...O-O, had never been tried
before in a tournament and he quotes GM Ftaknic
stating that this move leads to a "much worse"
position for Black. Rybka continued:

11. ... O-O

12.Bh6 g6

13.Bxf8 Bxf8

14.O-O Bh6

This is the move I have mentioned several times
here, the one which seems to have been missed
or else mis-evaluated. Clearly, White's Queen is
embarrassed, and what's more, the glaring
weaknesses left in the wake of the QB exchange
are highlighted by this move. Shredder decided to
chuck a center pawn rather than go for the Rd1 line
given by someone referred to as "Costa" from
ChessBase magazine, which the author also says
is skewered by Rybka, not "unclear"; Apparently,
Shredder agreed. Rybka went on to win in fifty
moves despite being down the exchange.

---

In game 2, the same line was re-tested, but
Shredder varied with 14.Bc4. Again, Rybka
sacrificed the exchange for a very active position
where, again, the White Queen is indisposed
and there is plenty of compensation for the
exchange. Rybka won a bit slower this time.

---

Perhaps there are clearer examples of where
Rybka is "the only one" to correctly evaluate a
given position, but this one stood out in my mind
because of the involvement of Gary Kasparov and
of course, the similarity of the maneuver:
Q-d4-h4-g3 with how certain hyper-aggressive
types play on a routine basis in the opening.
Often as not, the ONLY way to "refute" such
moves is by meeting fire with fire, by sacrificing
material when necessary in order to highlight
the loss of time such moves entail.

-- help bot

On May 7, 12:09 pm, "Chess One" wrote:
"Chess One" wrote in message

news:P9m%h.1960$s7.772@trndny05...

Martin, there is one other point here to which I can add a note which is a
specific about playing chess rather than generalities about play or the
components aspects of the program which may or may not be useful

- I set up a Russian Chess journalist [a GM] to annotate his own game [in
our series, this mean he had to annotate a loss] and provide below his own
comments - there is a certain level of play where the engine doesn't see
enough ahead and in fact at this level of play are not taken too seriously
/because/ as you will see in this game tactical considerations count for
very little, and a deep positional move seems to win the game, and in fact
calls into question the entire line - here the program is not used as much
for its computational ability, but for its database of possibilities - this
GM criticises himself for over-reliance on using the db, even though he
plays out various lines against his computer - he is playing inferior lines.

Which engine is he using though? Fritz8 lacks discrimination in this
sort of position. I don't have Fritz9 or X.

All the engines I have tried on this position prefer the natural Be6
above all else (as does powerbooks2006). Nc6 is also playable. Rybka
briefly preferred the novelty Be7 over all other options by ply 20 it
is second to Be6.

What this example might illustrate for OUR conversation is if any chess
engine can find what Bezgogov's could not. Here is the key text:

"13. Nh4 . I underestimated this plan in my preparation. White's knight
arrives at the very strong g6-square and its almost impossible to drive it
away from this station. The knight distinctly complicates Black's defense.

Shredder10 sees this at ply 17. It might have seen it at ply 16 but I
was on the phone.
It clearly fears the knight move as it has also put Be7 in the running
(AFAIK a novelty).

Ply (scores in cp)
17 18 19 Line

22 18 24 10. ... Be6 11. Ng5 Nd7 12. Bf4 Nb6 13. a4
76 61 72 10. ... Nc6 11. Bxc4 Be6 12. Bxe6 fxe6 13. Rb1
80 85 71 10. ... Be7 11. Bxc4 Be6 12. Bxe6 fxe6 13. Ng5
104 92 96 10. ... h6 11. 11. Bxc4 Be6 12. Bxe6 fxe6 13. Nh4
96 92 103 10. ... Ke8 11. Bxc4 Be7 12. Ng5 Bxg5 13. Bxg5



I must admit one more serious mistake - I played this endgame against one of
the strongest chess programs. My 'inanimate' opponent didn't find White's
13th move and, against any other play Black stands well and I would defend
Black's position easily."

(( entire game and commentary is athttp://www.chessville.com/LessonsLearned/2004Jan.htm))

some other notes from the game:-

"GM Alexei Bezgodov: My loss to an excellent grandmaster from Moscow,
Evgeny Najer, one can say consists of two parts. First I prepared badly for
the game using an incomplete database and already from the opening got a
very difficult (maybe even lost) endgame."

A critical moment comes here
"10...h6?!

He is saying, to summarise the above, that his 10 ... h6 becomes refuted by
Whites 13. Nh4. And that, Martin, is where I think a current test might be
useful. Here are the preceeding moves:-

(1) Najer,E - Bezgodov,A [D24]
56th Russian Championship (3.13), 05.09.2003

1.d4 d5 2.c4 dxc4

3.Nf3 Nf6 4.Nc3! c5 5. d5

5...e6 6.e4 exd5 7.e5! 7...Ne4 8.Qxd5 8...Nxc3

9.Qxd8+ Kxd8 10.bxc3

10...h6?! ((Lets call this Alpha Point)

11.Bxc4 11....Be6 12.Bxe6 fxe6 13.Nh4! ((Lets call this Beta Point))

---

A) My assumptions are that (a) a program will play the Alpha point move at
9, or score it as a strong consideration if it (b) does not see the Beta
point move at 13.

Most engines prefer to avoid losing the pawn on c4 so h6 is well down
the list.

Shredder would probably see this at tournament rates of play.
It also thinks that 13. ... c4 or Nc6 would be better than Be7.

B) Can your chess engine (a) find the Beta point move on its own? and (b)
how does it score that move in you hand-enter it, compared with scores for
other White options at 13?

Nh4 is about 20-25cp ahead of all other options in Shredder10. Then
Rb1 or Be3.

The position is effective at detecting engines that can be blinded by
swapoff sequences - like Fritz8. Both Shredder10 and Rybka2.3.1 have
no real difficulty in seeing 13. Rh4 as a threat and offer other
alternatives to the line played.

Regards,
Martin Brown

PS I wonder if Google will swallow this without trace....

In article . com,
help bot wrote:
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, [...].

You have some problem with the notion that development
of Cray Blitz stopped when the Cray became unavailable, and that
RMH then wrote Crafty for PCs? Crafty is not CB. People move
on. So do programs.

As far as I know, the learning approach has stemmed
from neural networks, not Cray supercomputers.
There are several ways in which programs can learn, few
of which exploit neural networks.
Is that a challenge? :D
Give me a team of real programmers and a few
years, and we'll see if this brash dismissal was
justified.

What do you propose? To uninvent the other ways in
which programs can learn? These days, programs that don't
learn just get smashed on the servers. But neural networks
do not seem to be particularly successful at the sorts of
learning that chess needs. It's not for want of trying.
Temporal differencing, eg, seems to do much better.

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,

[Yes, "Harry", sorry!]

and Bob Hyatt and Bert Gower who wrote the chess and the algorithms.
Wrote the chess? What does that mean?

It means that a complete chess program includes a lot of
different aspects -- the user interface, chess knowledge, tree-
searching methods, use of heuristics, .... And exploitation of
particular hardware. This was especially true back then, when
there was a factor of 20+ between a generic C/Fortran/whatever
program and code tuned to a particular machine, even more so when
there were vector instructions available. HN, according to RMH,
knew nothing at all about chess, not even the moves, nor about
the search algorithms and heuristics being used; but he was an
expert on the Cray.

[...] 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?

Who has "dismissed" the "brute power" of the Cray?
And not many of the competition were in "ordinary cars".
Belle may have been a PDP 11/23, but it was coupled to the
best special-purpose hardware that Ken Thompson could lay
his hands on. Hitech, and so on, were just as "brute" as
Cray Blitz. You really have to wait until long after the
time of CB before "ordinary" competition got close. Even
now, the top programs are not winning the world championship
or beating Kramnik on the cheapest PCs from "PC World" --
there's a hint in names like "*Deep* Junior/Fritz/etc".

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.

*No-one* has managed to maintain a substantive lead
for any length of time. We don't know whether CB and DB
*could* have done so, because Cray and IBM los