yes, again achieved a new rating record
for standard play (mostly 16/0) on the
top (computer) chess server playchess.com !
Now in the top 10 %, on about 11th place,
with a rating of 2669, not bad with some
skulltrails or xeons (8 cpus') in higher places. And
on ICC, with standard play place nr5 has been achieved
(handle bookbuilder).
So my repertoire with 1.e4, found with Bookbuilder,
after lots of testing and fine-tuning seems to
be confirmed, and is becoming quite solid.
some recent changes:
in the Ruy Lopez Zaitsev (9..Bb7) with black, i
dont play 16..g6?! anymore, as it doesn't give
much active counterplay; better seems the old
well known variation with 16..Nd7, and after
17 Ra3 f5! 18. exf5=! (or 18.Nh2 Qe7!)
Rxe1 19. Qxe1 Nxd5!
and yes chess looks drawish again..
more news later
Friday, November 21, 2008
Thursday, March 27, 2008
Practical relevance/opening repertoire
Now after the previous ideas and perfect opening lines, whereby black can maintain a draw, one can (and indeed should) ask what's the practical relevance in chess..
Well first of all, by using such almost perfect opening lines, an opening book for a chess program becomes much better of course. Much better ? Well, at least slightly better. For example i've tested my Arena book against the main book by H.Schapp, and achieved improvements of about 60 Elo points. Not really very much, and indeed, the engine/tactics still are much more important.
So what's the relevance of such opening theory for human players ? Well, having thought about this a bit more, i suspect that the approach of finding 'best' opening
moves is not really suitable for human players, having not such perfect memory as computers of course.
Yet another approach can be worthwile, namely choosing opening lines which restrict the number of playable options
for the opponent. Does this mean choosing the sharpest lines like gambits etc.?
Well, not necessarily, when playing e.g. a gambit like Kings Gambit, after 2. f4?! black still has a lot of playable options (ie. resulting in = lines) options,
and probably 2. Nf3 with such an approach still fundamentally is better.
Yet.. when looking at opening theory in such a (new?) way, other lines than 2.Nf3 also can be played, e.g
2. Nc3, or even 2.d4!? And in the latter case, after 2.exd4, white can also play interesting gambit lines
with eg. 3. c3 (Danish gambit) !
In such a way opening play is a bit similar as middle game play, in the sense that it is useful to gain
mobility of your pieces for your own side, whereas simultaneously trying to restrict mobility of
your opponent(s) pieces.. (i suspect many chess programs still are not very good in evaluating mobility,
especially not in end games, eg. when looking at positions like 'fortresses'and so on..).
Anyway, using such an approach (of trying to restrict nr of playable moves for the opponent, already
in the opening stage) of course the human player would benefit a lot of practical knowledge of the
variations he would choose, especially in sharp lines, such as the above mentioned Danish gambit.
So the practical relevance of this idea is that knowing opening theory, having a repertoire aiming for
above strategy, and using for such purposes a program like Bookbuilder, can be useful for the human chess
player in practical (otb) play !
(and yes, i've experienced this myself as well ofcourse; my standard rating on playchess.com now is more
than 1700, and still increasing..)
So, enough for the moment about computer chess, and back to the more interesting human chess again, i would say.. :)
Well first of all, by using such almost perfect opening lines, an opening book for a chess program becomes much better of course. Much better ? Well, at least slightly better. For example i've tested my Arena book against the main book by H.Schapp, and achieved improvements of about 60 Elo points. Not really very much, and indeed, the engine/tactics still are much more important.
So what's the relevance of such opening theory for human players ? Well, having thought about this a bit more, i suspect that the approach of finding 'best' opening
moves is not really suitable for human players, having not such perfect memory as computers of course.
Yet another approach can be worthwile, namely choosing opening lines which restrict the number of playable options
for the opponent. Does this mean choosing the sharpest lines like gambits etc.?
Well, not necessarily, when playing e.g. a gambit like Kings Gambit, after 2. f4?! black still has a lot of playable options (ie. resulting in = lines) options,
and probably 2. Nf3 with such an approach still fundamentally is better.
Yet.. when looking at opening theory in such a (new?) way, other lines than 2.Nf3 also can be played, e.g
2. Nc3, or even 2.d4!? And in the latter case, after 2.exd4, white can also play interesting gambit lines
with eg. 3. c3 (Danish gambit) !
In such a way opening play is a bit similar as middle game play, in the sense that it is useful to gain
mobility of your pieces for your own side, whereas simultaneously trying to restrict mobility of
your opponent(s) pieces.. (i suspect many chess programs still are not very good in evaluating mobility,
especially not in end games, eg. when looking at positions like 'fortresses'and so on..).
Anyway, using such an approach (of trying to restrict nr of playable moves for the opponent, already
in the opening stage) of course the human player would benefit a lot of practical knowledge of the
variations he would choose, especially in sharp lines, such as the above mentioned Danish gambit.
So the practical relevance of this idea is that knowing opening theory, having a repertoire aiming for
above strategy, and using for such purposes a program like Bookbuilder, can be useful for the human chess
player in practical (otb) play !
(and yes, i've experienced this myself as well ofcourse; my standard rating on playchess.com now is more
than 1700, and still increasing..)
So, enough for the moment about computer chess, and back to the more interesting human chess again, i would say.. :)
Criticism, theory, etc.
In a recent discussion on Usenet, rec.games.chess.analysis
my 'conjecture' (hypothesis) that chess fundamentally is
a draw, was critized by some people. Trying to give
some evidence, i showed some standard comp-comp
games, eg. with Rl Zaitsev, and Queens Indian,
where black clearly could maintain a draw.
But people may still question such 'evidence'
of course, as there are many many other lines.
Yet, with perfect play for both sides, assuming
white tries to maintain some initial positional
advantage, i found that often Ruy Lopez, and
against d4 black Nimzo Indian or Queens Indian
are the result of such 'perfect'games.
Then again, one may wonder whats the relevance
of mentioning the RL Zaitsev variation as the
main drawing line for black after 1.e4 , as
white for example doesnt have to play 9.h3
in the Ruy Lopez. Well this doesnt matter,
otherwise i wouldnt have mentioned such
lines. Of course i also have analyzed other
possibilities, eg. with 9 d4, the Boguljubow
variation, but then i found even easier
drawing lines for black.
Same after 1.d4 Nf6, white could play instead
of 2.c4 lines as 2.c3, or Bg5, but also then
theoretically black has no problem in
achieving drawish positions. So indeed, its
my conviction that chess fundamentally
is a draw; in case of perfect play of
both sides, of course.
my 'conjecture' (hypothesis) that chess fundamentally is
a draw, was critized by some people. Trying to give
some evidence, i showed some standard comp-comp
games, eg. with Rl Zaitsev, and Queens Indian,
where black clearly could maintain a draw.
But people may still question such 'evidence'
of course, as there are many many other lines.
Yet, with perfect play for both sides, assuming
white tries to maintain some initial positional
advantage, i found that often Ruy Lopez, and
against d4 black Nimzo Indian or Queens Indian
are the result of such 'perfect'games.
Then again, one may wonder whats the relevance
of mentioning the RL Zaitsev variation as the
main drawing line for black after 1.e4 , as
white for example doesnt have to play 9.h3
in the Ruy Lopez. Well this doesnt matter,
otherwise i wouldnt have mentioned such
lines. Of course i also have analyzed other
possibilities, eg. with 9 d4, the Boguljubow
variation, but then i found even easier
drawing lines for black.
Same after 1.d4 Nf6, white could play instead
of 2.c4 lines as 2.c3, or Bg5, but also then
theoretically black has no problem in
achieving drawish positions. So indeed, its
my conviction that chess fundamentally
is a draw; in case of perfect play of
both sides, of course.
Wednesday, January 30, 2008
Black is ok, chess is a draw with perfect play..
With some more Rybka analysis, i'm convinced that
chess is a draw, with perfect play for black.
After e4 e5!, and with the Ruy Lopez it seems not
possible anymore to obtain a structural advantage
for white. Neither with the Bobuljubow Ruy Lopez
variation (8.d4) nor with the Closed RL with 9.h3,
as not only the Zaitsev (10..Bb7) but maybe also
the Breyer (10..Nb8) can maintain equality
for black, i.e. equally valued positions.
Of course careful play for black is needed,
to anticipate any winning plans for white,
but i've analyzed these lines deep enough i think
that a draw seems 99.999 % likely with perfect
play for black.
If one would talk about 'solving' chess, ie.
trying to find a forced win for white against any
defence for black (like eg. in four-in-a-row) then
proving chess would be a draw, also would be a
'solution' of chess. So, i now can say that
chess has been 'solved', its a draw.. :)
'Proof':
*if* there would be a forced win for white, then
there would be (deep) opening lines where a clear
advantage can be obtained against any defence for
black. As this is not the case (against 1.e4 e5!),
and against d4 eg. Nimzo-Indian or Queens Indian,
we can say with almost 100 % confidence that chess
is a draw.
And with end game rules like eg. the 50 move rule,
it also is not a suprising result, as with many
conventional 'main' opening lines it already appears
almost impossible to achieve a winning endgame
for white. Also it is known that the better GM
players or computers are playing, the higher the
chance of a draw will be; nevertheless its an
interesting result, now also verified with
computer analysis.
Then why on average is white winning more than
black ? Well this answer is not so difficult, in
general there are less good (non-losing) move-
possibilities for black after every white move.
Thus the chances that black makes a mistake is
larger, especially with lower-ranked/non
perfect playing people/computers.
And think about this: after white has made a
move, i.e. a certain choice about opening strategy,
(eg 2. Nf3 after 1.e4) he is gradually revealing
his plans to black, who can respond accordingly.
Another reason that we can be confident that black
always can equalize. Theoretical we can even think
about the concept of 'zugzwang' , white *has* to
move, and black can respond accordingly. So its
probably fair enough to say that white's (positional)
starting advantage is minimal, certainly less
than 1/3 of a pawn; this seems to be confirmed by
Rybka, if i let if freely analyze the starting
position, its positional value seems to converge
to a value of only about 1/10 of a pawn..
Some another questions:
--------------------------
can white win against the Sicilian (with 1.e4)
or against queens gambit (with d4) ?
Well i don't know, probably not, the Sicilian
certainly seems more difficult to play for black
(less correct moves after every 'best' move by white),
but this is not so important anyway in our theoretical
discussion, as 1..e5! is already 'ensures' a draw;
when after 2.Nf3 black aims for Ruy Lopez 2..Nc6 !
Whereas 2.Nc3 Vienna also doesn't impose much threat.
Would Petrov 2..Nf6?! also lead to a draw for black
with perfect play ? Well, again, irrelevant, RL
is fine, and it's confirmed by a book like
'The Ruy Lopez, a guide for black' :
http://www.amazon.com/Ruy-Lopez-Guide-Black/dp/1904600670
although i have found some improvements to this
book.. And yes, other moves like 1.c4 (e5!!) or
1.Nf3 (similar to d4) of course also don't lead to
advantage for white.
So what's the practical meaning of all this ?
Well, for me at least it indicates that opening
knowledge especially is useful for black,
as its always easier to remember the
correct/non-losing moves, than having
to analyze them Otb..
For white, well moves like 1.f3!? are not
recommended of course, but in general, depending
on personal style, any starting reasonable
move seems possible.
But for beginners, as they anyway often have
to face 1.e4 as black, i still would recommend
to often start with 1.e4, as described in my
e-book; which soon will be updated with the
latest variations; ie equalizing for black..
And for advanced players, well instead of
a draw you can always aim for a win with
black of course, choosing gambits like Benko,
Marshall, or other dynamic & well established
gamits which are hard (or impossible) to refute;
depending on style & preferences..
Books by A.Adorjan (Black is ok) confirm this:
http://www.amazon.com/Black-Forever-Chess-Andras-Adorjan/dp/0713489421
chess is a draw, with perfect play for black.
After e4 e5!, and with the Ruy Lopez it seems not
possible anymore to obtain a structural advantage
for white. Neither with the Bobuljubow Ruy Lopez
variation (8.d4) nor with the Closed RL with 9.h3,
as not only the Zaitsev (10..Bb7) but maybe also
the Breyer (10..Nb8) can maintain equality
for black, i.e. equally valued positions.
Of course careful play for black is needed,
to anticipate any winning plans for white,
but i've analyzed these lines deep enough i think
that a draw seems 99.999 % likely with perfect
play for black.
If one would talk about 'solving' chess, ie.
trying to find a forced win for white against any
defence for black (like eg. in four-in-a-row) then
proving chess would be a draw, also would be a
'solution' of chess. So, i now can say that
chess has been 'solved', its a draw.. :)
'Proof':
*if* there would be a forced win for white, then
there would be (deep) opening lines where a clear
advantage can be obtained against any defence for
black. As this is not the case (against 1.e4 e5!),
and against d4 eg. Nimzo-Indian or Queens Indian,
we can say with almost 100 % confidence that chess
is a draw.
And with end game rules like eg. the 50 move rule,
it also is not a suprising result, as with many
conventional 'main' opening lines it already appears
almost impossible to achieve a winning endgame
for white. Also it is known that the better GM
players or computers are playing, the higher the
chance of a draw will be; nevertheless its an
interesting result, now also verified with
computer analysis.
Then why on average is white winning more than
black ? Well this answer is not so difficult, in
general there are less good (non-losing) move-
possibilities for black after every white move.
Thus the chances that black makes a mistake is
larger, especially with lower-ranked/non
perfect playing people/computers.
And think about this: after white has made a
move, i.e. a certain choice about opening strategy,
(eg 2. Nf3 after 1.e4) he is gradually revealing
his plans to black, who can respond accordingly.
Another reason that we can be confident that black
always can equalize. Theoretical we can even think
about the concept of 'zugzwang' , white *has* to
move, and black can respond accordingly. So its
probably fair enough to say that white's (positional)
starting advantage is minimal, certainly less
than 1/3 of a pawn; this seems to be confirmed by
Rybka, if i let if freely analyze the starting
position, its positional value seems to converge
to a value of only about 1/10 of a pawn..
Some another questions:
--------------------------
can white win against the Sicilian (with 1.e4)
or against queens gambit (with d4) ?
Well i don't know, probably not, the Sicilian
certainly seems more difficult to play for black
(less correct moves after every 'best' move by white),
but this is not so important anyway in our theoretical
discussion, as 1..e5! is already 'ensures' a draw;
when after 2.Nf3 black aims for Ruy Lopez 2..Nc6 !
Whereas 2.Nc3 Vienna also doesn't impose much threat.
Would Petrov 2..Nf6?! also lead to a draw for black
with perfect play ? Well, again, irrelevant, RL
is fine, and it's confirmed by a book like
'The Ruy Lopez, a guide for black' :
http://www.amazon.com/Ruy-Lopez-Guide-Black/dp/1904600670
although i have found some improvements to this
book.. And yes, other moves like 1.c4 (e5!!) or
1.Nf3 (similar to d4) of course also don't lead to
advantage for white.
So what's the practical meaning of all this ?
Well, for me at least it indicates that opening
knowledge especially is useful for black,
as its always easier to remember the
correct/non-losing moves, than having
to analyze them Otb..
For white, well moves like 1.f3!? are not
recommended of course, but in general, depending
on personal style, any starting reasonable
move seems possible.
But for beginners, as they anyway often have
to face 1.e4 as black, i still would recommend
to often start with 1.e4, as described in my
e-book; which soon will be updated with the
latest variations; ie equalizing for black..
And for advanced players, well instead of
a draw you can always aim for a win with
black of course, choosing gambits like Benko,
Marshall, or other dynamic & well established
gamits which are hard (or impossible) to refute;
depending on style & preferences..
Books by A.Adorjan (Black is ok) confirm this:
http://www.amazon.com/Black-Forever-Chess-Andras-Adorjan/dp/0713489421
Monday, January 21, 2008
Opening theory and the art of bookmaking
Having looked more in detail into 'my' repertoire,
i.e. the opening book(s) i made for Arena it
appears that gradually i'm getting into new theory.
This is because conventional theory, and also
most 'conventional' opening books are based on (top)
GM games. But the deeper you get into a line, the
fewer games there are, after which the theory gets
very dependent on just one or a few games.
The statistics then get unreliable, and it the of
course is interesting to analyze some of the most important/frequent positions with the top engine
Rybka (in the past most computer chess programs,
i.e. 'engines' were positionally not good enough
to make good moves/ evaluations, but with the latest
Rybka 2.3.2 i believe it can be used to make
professional/GM quality evaluations); this sometimes
leads to interesting results, eg. more solid defences
for black in lines such as the closed Ruy Lopez,
Zaitsev or Breyer variations (doing a full minimax
with the old Bookbuilder 3.6 certainly was useful
in such cases, a/o because of transpositions;
but ofcourse much depends then on the
engine evaluation, and the subvariations which
are already in the book, whether played by GM's ,
having been analyzed myself, or being added
as result of computergames, either downloaded
or played myself in some Arena tournaments..)
In some cases i've checked the results against the
latest Rybka book (rybka2.ctg), and in the
Ruy Lopez i found some improvements for black;
with, i must confess, the conclusion that black is ok.
A similar exercise to check my repertoire against
d4 confirmed that black also is ok in this case,
although the defence with the QueensIndian
(against 1.d4 2.c4 3.Nf3!) was most difficult.
So although we cannot say that 1.e4 is the best
move, but for beginners of course it still is
highly recommended.
This also in some ways 'follows' the history
of chess, where initially (19th century) almost
only 1.e4 was played, whereas the more positional
1.d4 (with sometimes some highly complex play
, eg. against the Benko gambit) came
later, i.e. in the 20th century.
Some preliminary conclusions:
------------------
1) conventional chess is a draw (but not dead,
as RJ Fischer who passed away just recently,
claimed in his later years)
2) the repertoire given in my e-book basically
still is correct and can be used both by
beginners as well as more advanced players,
especially if they play against players of almost
equal ability
3) with Rybka analysis the lines have been improved
although i still have to write them down in detail
4) as a result some of the recommended subvariations
eg. against Caro Kann have changed again a bit
5) against weaker players it would be interesting
to develop a gambit repertoire, both for white
(then i still recommend e4) and with black (where
possible); currently i'm in the process of
researching such a repertoire and making a 2nd
Arena 'gambit' book
6) when becoming stronger chess players can
also develop a '3rd' repertoire with 1.d4 also
this will be a tedious task if it is done
in detail, but the spinoff could be it would
be useful against highly tactical (better?)
players; in such a way it might also be
useful to aim for a draw against stronger players
(maybe later i'll make a 3rd Arena book with
d4 for white but i like to do this as thoroughly
as i've done for e4, so it might take a few years..)
i.e. the opening book(s) i made for Arena it
appears that gradually i'm getting into new theory.
This is because conventional theory, and also
most 'conventional' opening books are based on (top)
GM games. But the deeper you get into a line, the
fewer games there are, after which the theory gets
very dependent on just one or a few games.
The statistics then get unreliable, and it the of
course is interesting to analyze some of the most important/frequent positions with the top engine
Rybka (in the past most computer chess programs,
i.e. 'engines' were positionally not good enough
to make good moves/ evaluations, but with the latest
Rybka 2.3.2 i believe it can be used to make
professional/GM quality evaluations); this sometimes
leads to interesting results, eg. more solid defences
for black in lines such as the closed Ruy Lopez,
Zaitsev or Breyer variations (doing a full minimax
with the old Bookbuilder 3.6 certainly was useful
in such cases, a/o because of transpositions;
but ofcourse much depends then on the
engine evaluation, and the subvariations which
are already in the book, whether played by GM's ,
having been analyzed myself, or being added
as result of computergames, either downloaded
or played myself in some Arena tournaments..)
In some cases i've checked the results against the
latest Rybka book (rybka2.ctg), and in the
Ruy Lopez i found some improvements for black;
with, i must confess, the conclusion that black is ok.
A similar exercise to check my repertoire against
d4 confirmed that black also is ok in this case,
although the defence with the QueensIndian
(against 1.d4 2.c4 3.Nf3!) was most difficult.
So although we cannot say that 1.e4 is the best
move, but for beginners of course it still is
highly recommended.
This also in some ways 'follows' the history
of chess, where initially (19th century) almost
only 1.e4 was played, whereas the more positional
1.d4 (with sometimes some highly complex play
, eg. against the Benko gambit) came
later, i.e. in the 20th century.
Some preliminary conclusions:
------------------
1) conventional chess is a draw (but not dead,
as RJ Fischer who passed away just recently,
claimed in his later years)
2) the repertoire given in my e-book basically
still is correct and can be used both by
beginners as well as more advanced players,
especially if they play against players of almost
equal ability
3) with Rybka analysis the lines have been improved
although i still have to write them down in detail
4) as a result some of the recommended subvariations
eg. against Caro Kann have changed again a bit
5) against weaker players it would be interesting
to develop a gambit repertoire, both for white
(then i still recommend e4) and with black (where
possible); currently i'm in the process of
researching such a repertoire and making a 2nd
Arena 'gambit' book
6) when becoming stronger chess players can
also develop a '3rd' repertoire with 1.d4 also
this will be a tedious task if it is done
in detail, but the spinoff could be it would
be useful against highly tactical (better?)
players; in such a way it might also be
useful to aim for a draw against stronger players
(maybe later i'll make a 3rd Arena book with
d4 for white but i like to do this as thoroughly
as i've done for e4, so it might take a few years..)
Sunday, January 06, 2008
A new start
First of all best wishes to all for the new year 2008.
For this year, i've got some plans to renew my chess repertoire.
After comprehensive analysis and computer games with
the top engine Rybka, it appears that 1 e4 still is
slightly better than d4. Games by Rybka with the Noomen book,
and also human games by Anand seem to confirm this idea.
In my own computer-repertoire, there will be some changes,
and i will write them down in the e-book 'better chess opening play',
so i'm planning a new version, probably ready in a few months.
Writing down the new variations will take less time,
probably only a few weeks.
Some noticeable changes: the Caro Kann defense has become
stronger for black, and the 'best' line for white now isnt
the advance line anymore, but more in line with the
classical lines, with a few improvements of course.
Against Sicilian, both in the Kalashnikov/Pelikan/Sveshnikov
as well as the Najdorf (English attack with Be3!) it appears white
can maintain a positional advantage, according to the Rybka
analysis. So also for advanced players, 1 ..e5 ! now seems the
best defense. Especially the Zaitsev defense (9..Bb7) seems
strong, especially when first 9.. Rfe8! is played, and then
later either 10 .. Bb7 (after 10.d4) or 10.. Bd7 (after 10.d3).
Yes i found some improvements, but i'll have to do some
more testing for a while before i will mention them here.
Until later,
Jef
For this year, i've got some plans to renew my chess repertoire.
After comprehensive analysis and computer games with
the top engine Rybka, it appears that 1 e4 still is
slightly better than d4. Games by Rybka with the Noomen book,
and also human games by Anand seem to confirm this idea.
In my own computer-repertoire, there will be some changes,
and i will write them down in the e-book 'better chess opening play',
so i'm planning a new version, probably ready in a few months.
Writing down the new variations will take less time,
probably only a few weeks.
Some noticeable changes: the Caro Kann defense has become
stronger for black, and the 'best' line for white now isnt
the advance line anymore, but more in line with the
classical lines, with a few improvements of course.
Against Sicilian, both in the Kalashnikov/Pelikan/Sveshnikov
as well as the Najdorf (English attack with Be3!) it appears white
can maintain a positional advantage, according to the Rybka
analysis. So also for advanced players, 1 ..e5 ! now seems the
best defense. Especially the Zaitsev defense (9..Bb7) seems
strong, especially when first 9.. Rfe8! is played, and then
later either 10 .. Bb7 (after 10.d4) or 10.. Bd7 (after 10.d3).
Yes i found some improvements, but i'll have to do some
more testing for a while before i will mention them here.
Until later,
Jef
Subscribe to:
Posts (Atom)