| |
Main
Date: 02 Apr 2005 04:53:20
From: Major Cat
Subject: Computers and opening theory
|
The opinion that computers are much better at calculating tactics as
opposed to evaluating positions in chess has enjoyed considerable cur-
rency over time. So be it. To this effect, have there been any "celeb-
rated" cases of respectable/promising opening lines that were found
glaringly wanting after having been subjected to in-depth tactical
analysis by powerful computers?
Major Cat
|
|
| |
Date: 22 May 2005 05:13:12
From: Mark Houlsby
Subject: Re: Backsolving
|
k Houlsby wrote:
>> Houlsby:
>> Above all: study tactics and endgames. *Never* study openings.
>well you would agree with me that a simple repertoire would
be useful for them, would you ?
Yes, a repertoire consisting entirely of systems. It's a case not of my
agreeing, I'm *proposing* that to the spamming troll Leahy and, because
*you're* replying to my post, *you're* doing the agreeing....or are
you?
>just 1.e4, maybe some
anti-sicilian like Alapin (c3) and QGD for example against
d4 and Reti ?
Definitely not. If you play 1.e4 you could face 1...Nf6 or 1...d5
or....
>at least that's what you suggested in your next line,
about 'systems'..
Wrong.
>>Play *systems* with respect to which you may hope to understand the
>> *tactics* better than your opponent does.
>agree
>> Until then, studying openings is a waste of time, and as a
> direct consequence, BookUp is a waste of time.
>nah it's probably useful to study some opening tactics,
like eg. refutation of Damiano's defence, and things like
that. some other products can do that as well.
Yes, that's exactly what I argued above. Learn the tactics pertaining
to the system you play.
>Nb theoretically it would be possible to generate
CAP data (chess analysis project) for *all* nodes
except those with deviating from abs(4) score or so
(tactical mistakes) at ply 12 or so. A subsequent
deep evaluation of the endnodes, with a top engine,
and then a complete minimax (that's in fact all
what's backsolving is all about) would reveil wat's
the best starting move, 1.e4! imho.yes,
that would be including novelties.
So, you're arguing that theoretically it's possible to solve chess. I
think I can agree with that. Chess is, after all, a closed system with
fixed internal laws.
1.e4 is an excellent choice of first move if one is prepared to learn
Alekhin's Defence, the Scandinavian, Nimzowitsch's Defence, the French
Defence, the Petroff Defence, Philidor's Defence,....
Systems are looking better, no?
>so.. there's still some work to be done.. :)
You said it (and put it rather well)!
>would the result be useful for beginners ?
probably not. maybe for correspondence
chess, and such kind of players.
Maybe, but only if the rules of the competition permit (c)heating.
Best regards
k
|
| |
Date: 20 May 2005 04:17:04
From: Mark Houlsby
Subject: Re: Backsolving
|
Leahy:
> > >I think this is one of the best uses of Bookup. As often as not,
when
> >the
> >contents of a popular opening book are backsolved, the starting
> >position of
> >chess (move 1) is often tagged as winning for White or winning for
> >Black.
Houlsby:
> > So how many lines have been worked out, by BookUp, from move one to
a
> draw or a loss for White or Black?
Leahy:
> As many lines as you put in.
Houlsby:
That being true, how many black defences have been refuted? Has BookUp
refuted, for example, the Gr=FCnfeld or the Sicilian? Any others?
Leahy
>Your general idea seems to be that a patzer should not prepare his
openings
and should not document his openings in Bookup and the assessments of a
patzer are not worth backsolving in Bookup.
Houlsby:
No, that's completely wrong. I've stated *only* that *patzers should
not study openings at all*. Rather than wasting our time studying
opening lines, we should study only the *tactical traps* which pertain
to the openings we play. That way, we're more likely both to emerge
from the opening unscathed, and to be ready to play a middlegame which
is to our liking. It *is* a good idea for us to study endgames, because
endgame theory is pretty static.
Leahy:
> You've said earlier that a
patzer should not play analysis from a grandmaster that he does not
understand.
Houlsby:
Correct. I tried it several times. I lost several times.
Leahy:
> You obviously know more about this than I do. What exactly
should a patzer do?
Once again: a patzer should study tactics and endgames. Ask any top
trainer, for example, Dvoretsky.
I've already told you this several times. Will you pay attention this
time, or just keep trolling?
Leahy:
>> >Backsolving makes the critical line in any analysis stand out like
a
> sore
> thumb. Then the unplayed moves that Al Tomalty (Komputer Korner)
> mentioned
> will scream to be added.
Houlsby:
> How many are to be added?
> <snip>
> Leahy wrote:
> >>>You've made a good case for the usefulness of Backsolving.
> Houlsby wrote:
> >>No, he hasn't, and I suspect that he may agree that he has not.
> Leahy:
> >I wasn't expecting his agreement. I tend to stick to topics on
chess
> software and not to topics on who likes/needs/understands chess
> software. :)
> Houlsby:
> Right, so your *understanding* of BookUp is crucial to your being
able
> to answer questions like KK's... agreed?
> Equally, your *understanding* of his (or my, or anyone's) misgivings
is
> crucial, too... agreed?
> Leahy:
> >Al is correct that adding a new move means "the preceding
backsolving
> process is now uselesss and has
> to be done again based on the new novelty."
> Houlsby:
> So, are you suggesting that BookUp has Backsolved *no* lines from a
> draw/forced loss back to move 1?
Leahy:
How do you get this suggestion from what Al and I wrote?
Houlsby:
If I backsolve a particular line, and play it, I lose because of a
novelty. DUH!
Once again: which black defences has BookUp refuted? Are you dumb?
> Leahy:
> > Fortunately you can and should
> do it again, and it can be done in real time just by moving backward
> through
> the tree from the novelty.
> Houlsby:
> Why should I, given the evident uselessness of the process, as you
have
> just explained?
Leahy:
How did you arrive at evident uselessness?
Houlsby:
Once again: if I play a backsolved line, I will lose because of a
novelty which hasn't been backsolved. Are you dumb?
Leahy:
That is akin to saying that
inserting a new number into a spreadsheet makes the spreadsheet
worthless.
Houlsby:
No, it's *nothing* like that. What a stupid analogy! All a spreadsheet
does is run the numbers. A spreadsheet does in seconds what it used to
take accountants a week to do.
Leahy:
Let the spreadsheet calculate the new cells already! That's what it is
designed to do.
Houlsby:
Right, it's designed to run the numbers. We're talking about whether or
not backsolving has refuted any black defences, which is *entirely
different*, qualitatively speaking. How many black defences has BookUp
refuted? The Caro-Kann, for example? The Dutch? Which ones?
Leahy:
And if GM Svidler was wicked fast and perfect at adding numbers I'll
bet he
could still use a spreadsheet to document his finances. Similarly if
he was
the best analyst on earth he could benefit from Bookup's Backsolving.
Houlsby:
Sure he could, on account of his being a super-GM who *understands* the
opening well enough to be able *to tell the difference* between the
Shinola that backsolving generates and the rest of its output. We've
already been over this.
Leahy:
Those that cannot add two numbers reliably would probably be bewildered
by a
spreadsheet. I imagine patzers are bewildered by backsolving. It is
the
human version of mini-maxing, the process used by all chess playing
programs
to "think." Ridiculing the process is not going to enhance one's
understanding of it.
Houlsby:
Once again: what I am ridiculing is not *the process itself*. I have
*already acknowledged*, several times, that the *process* of
backsolving is *fine* as far as it goes.
Unfortunately, it takes someone with the acuity of a Svidler, or at
least a master of some kind, to tell the difference between the Shinola
and the rest of backsolving's output. Why can't you understand this?
> Houlsby:
> So you're suggesting that BookUp is *as effective* at identifying and
> evaluating critical lines as *Svidler or any lesser GM* is? Even if
it
> is, what use is that to me, if I can't understand *why* a tabiya is
> evaluated as +/=3D or whatever?
Leahy:
You've hit on an important point. If you truly do not understand why
any
position (including a tabiya) is evaluated in any particular way then
it is
impossible to explain the importance of Backsolving.
Houlsby:
So you admit that backsolving is *completely useless* to anyone who is
not a master. You see, everyone who is not a master *definitely does
not understand* opening tabiyas, at all, which is why whenever a patzer
plays an opening which he/she understands less well than the other guy,
said patzer will *always* LOSE convincingly, for *tactical* reasons.
Leahy:
Backsolving can only be understood when you get that the "+/=3D or
whatever"
is based on the best line of play in a tree of analysis, and with huge
trees
of analysis Backsolving takes the guesswork out of calculating the
effect of
an improved line.
Houlsby:
Right, so you finally agree (after several weeks of your arguing the
contrary) that backsolving is useless to anyone who is not a master.
- Hide quoted text -
- Show quoted text -
> Houlsby:
> >>Backsolving is nothing more, and nothing less, than computer
> analysis,
> >>which is fine as far as it goes, but that may be only a couple of
> >>yards....
> Leahy:
> >Backsolving is much more often based on human analysis and can be
> contrasted
> with computer analysis.
> Houlsby:
> So how do I, a patzer:
> a) understand
> and
> b) reconcile any conflicting evaluations?
Leahy:
You tell me how a patzer does this.
Houlsby:
That's the *whole point*. No patzer *can* do it...AT ALL.
Leahy:
Computer analysis?
Houlsby:
Nope.
Leahy:
Reading a GM's notes?
Houlsby:
Nope.
Leahy:
Your original thought?
Houlsby:
Heaven forfend! That *always* loses *really* quickly... ;-)
Leahy:
What do you do?
Houlsby:
I play *systems* with respect to which I hope to understand the
*tactics* better than my opponent does. If I actually manage to get out
of the opening, I get a pretty good game, and it's down to *who
blunders last*.
Leahy:
What would you suggest for patzers?
Houlsby:
Above all: study tactics and endgames. *Never* study openings. Play
*systems* with respect to which you may hope to understand the
*tactics* better than your opponent does. If, and only if, you get
*good enough* at tactics and endgames to reach *master level*, you will
have become a good enough player to be able to discern the *types* of
opening which lead to the *types* of middlegame which suit your
*style*. Until then, studying openings is a waste of time, and as a
direct consequence, BookUp is a waste of time.
k Houlsby
|
| | |
Date: 22 May 2005 13:14:13
From: jef
Subject: Re: Backsolving
|
k Houlsby wrote:
> Houlsby:
> Above all: study tactics and endgames. *Never* study openings.
well you would agree with me that a simple repertoire would
be useful for them, would you ? just 1.e4, maybe some
anti-sicilian like Alapin (c3) and QGD for example against
d4 and Reti ?
at least that's what you suggested in your next line,
about 'systems'..
>Play *systems* with respect to which you may hope to understand the
> *tactics* better than your opponent does.
agree
> Until then, studying openings is a waste of time, and as a
> direct consequence, BookUp is a waste of time.
nah it's probably useful to study some opening tactics,
like eg. refutation of Damiano's defence, and things like
that. some other products can do that as well.
Nb theoretically it would be possible to generate
CAP data (chess analysis project) for *all* nodes
except those with deviating from abs(4) score or so
(tactical mistakes) at ply 12 or so. A subsequent
deep evaluation of the endnodes, with a top engine,
and then a complete minimax (that's in fact all
what's backsolving is all about) would reveil wat's
the best starting move, 1.e4! imho.yes,
that would be including novelties.
so.. there's still some work to be done..
:)
would the result be useful for beginners ?
probably not. maybe for correspondence
chess, and such kind of players.
best regards
jef
http://superchess.com
|
| |
Date: 19 May 2005 15:12:25
From: Mark Houlsby
Subject: Re: Backsolving
|
>>How effective is backsolving to move 1?
>Just as effective as it is to any other move.
So how many lines have been worked out, by BookUp, from move one to a
draw or a loss for White or Black?
>I think this is one of the best uses of Bookup. As often as not, when
the
contents of a popular opening book are backsolved, the starting
position of
chess (move 1) is often tagged as winning for White or winning for
Black.
So how many lines have been worked out, by BookUp, from move one to a
draw or a loss for White or Black?
>Backsolving makes the critical line in any analysis stand out like a
sore
thumb. Then the unplayed moves that Al Tomalty (Komputer Korner)
mentioned
will scream to be added.
How many are to be added?
<snip >
Leahy wrote:
>>>You've made a good case for the usefulness of Backsolving.
Houlsby wrote:
>>No, he hasn't, and I suspect that he may agree that he has not.
Leahy:
>I wasn't expecting his agreement. I tend to stick to topics on chess
software and not to topics on who likes/needs/understands chess
software. :)
Houlsby:
Right, so your *understanding* of BookUp is crucial to your being able
to answer questions like KK's... agreed?
Equally, your *understanding* of his (or my, or anyone's) misgivings is
crucial, too... agreed?
Leahy:
>Al is correct that adding a new move means "the preceding backsolving
process is now uselesss and has
to be done again based on the new novelty."
Houlsby:
So, are you suggesting that BookUp has Backsolved *no* lines from a
draw/forced loss back to move 1?
Leahy:
> Fortunately you can and should
do it again, and it can be done in real time just by moving backward
through
the tree from the novelty.
Houlsby:
Why should I, given the evident uselessness of the process, as you have
just explained?
Leahy:
> Clicking the button that takes you back to the
starting position will backsolve it in an instant. Take a look at the
video
mentioned above for an example.
Houlsby:
I think I get the idea, thanks.
Leahy:
>> The only way to get the certainty you're looking for is to make sure
those "unplayed" moves
are added and their resulting positions are accurately assessed .
Houlsby:
>>You mean, for *all* unplayed novelties, good and bad?! Whoa, baby,
>chess would be solved in a month at the most, if that were true.
>Unfortunately, it's horse manure.
Leahy:
>I agree. Adding all unplayed novelties would be horse manure.
Houlsby:
So you admit that Backsolving is basically useless?
Leahy:
>That's why I said "those" unplayed moves, as in those moves Al said
needed
to be added to improve upon the existing move(s)
Houlsby:
Ah, yes, but, don't you see...? Unless you *include* all unplayed
novelties, it would take someone with the talent of GM Svidler to
*figure out* which Backsolving output is useful, and which is horse
manure...
Leahy:
>>>You've just inferred that quite a few titled players are "complete
>>robot
>>like idiots."
Houlsby:
>>Has he? Does, for example, GM Svidler rely on backsolving *rather
than
>his own analysing capabilities*?
>Where is your evidence for this?
Leahy:
>You'd have to ask GM Svidler that.
Houlsby:
Well, do you have such evidence with respect to *anyone at all*?
Leahy:
>As far as discussing software, Bookup's Backsolving would be very
handy for
pointing out the critical line(s) in Svidler's or anyone's analysis as
well
as demonstrating the impact on the analysis when an improvement is
found.
Houlsby:
So you're suggesting that BookUp is *as effective* at identifying and
evaluating critical lines as *Svidler or any lesser GM* is? Even if it
is, what use is that to me, if I can't understand *why* a tabiya is
evaluated as +/= or whatever?
Leahy:
>>Next you'll be saying that Bookup's training wizard is only good for
players
>>who have burned up their long term memories with dissolute living. :)
Houlsby:
>Don't be silly.
Leahy:
You got my point.
Houlsby:
I beg your pardon, but I DID NOT GET YOUR POINT. KK has raised a
serious issue concerning BookUp's shortcomings, and you fob him off,
suggesting that his misgiving is the equivalent of his denigrating the
alleged lifestyle of a certain "type" of player, which may not even
exist in statistically significant numbers...
Houlsby:
>>Backsolving is nothing more, and nothing less, than computer
analysis,
>>which is fine as far as it goes, but that may be only a couple of
>>yards....
Leahy:
>Backsolving is much more often based on human analysis and can be
contrasted
with computer analysis.
Houlsby:
So how do I, a patzer:
a) understand
and
b) reconcile any conflicting evaluations?
Leahy:
>When I can beat the computers 9 out of 10 then I'll start ignoring
their
contribution
Houlsby:
How good are you at chess? How much has BookUp helped you... i.e. how
much has *your* rating increased?
If the answer is "not much", do you think that there might just
possibly be a reason for that?
k
|
| | |
Date: 20 May 2005 00:38:17
From: Mike Leahy
Subject: Re: Backsolving
|
"k Houlsby" <[email protected] > wrote in message
news:[email protected]...
> >I think this is one of the best uses of Bookup. As often as not, when
> >the
> >contents of a popular opening book are backsolved, the starting
> >position of
> >chess (move 1) is often tagged as winning for White or winning for
> >Black.
>
> So how many lines have been worked out, by BookUp, from move one to a
> draw or a loss for White or Black?
As many lines as you put in.
Your general idea seems to be that a patzer should not prepare his openings
and should not document his openings in Bookup and the assessments of a
patzer are not worth backsolving in Bookup. You've said earlier that a
patzer should not play analysis from a grandmaster that he does not
understand. You obviously know more about this than I do. What exactly
should a patzer do?
> >Backsolving makes the critical line in any analysis stand out like a
> sore
> thumb. Then the unplayed moves that Al Tomalty (Komputer Korner)
> mentioned
> will scream to be added.
>
> How many are to be added?
>
> <snip>
> Leahy wrote:
> >>>You've made a good case for the usefulness of Backsolving.
>
> Houlsby wrote:
> >>No, he hasn't, and I suspect that he may agree that he has not.
>
> Leahy:
> >I wasn't expecting his agreement. I tend to stick to topics on chess
> software and not to topics on who likes/needs/understands chess
> software. :)
>
> Houlsby:
> Right, so your *understanding* of BookUp is crucial to your being able
> to answer questions like KK's... agreed?
>
> Equally, your *understanding* of his (or my, or anyone's) misgivings is
> crucial, too... agreed?
>
> Leahy:
> >Al is correct that adding a new move means "the preceding backsolving
> process is now uselesss and has
> to be done again based on the new novelty."
>
> Houlsby:
> So, are you suggesting that BookUp has Backsolved *no* lines from a
> draw/forced loss back to move 1?
How do you get this suggestion from what Al and I wrote?
>
> Leahy:
> > Fortunately you can and should
> do it again, and it can be done in real time just by moving backward
> through
> the tree from the novelty.
>
> Houlsby:
> Why should I, given the evident uselessness of the process, as you have
> just explained?
How did you arrive at evident uselessness? That is akin to saying that
inserting a new number into a spreadsheet makes the spreadsheet worthless.
Let the spreadsheet calculate the new cells already! That's what it is
designed to do.
And if GM Svidler was wicked fast and perfect at adding numbers I'll bet he
could still use a spreadsheet to document his finances. Similarly if he was
the best analyst on earth he could benefit from Bookup's Backsolving.
Those that cannot add two numbers reliably would probably be bewildered by a
spreadsheet. I imagine patzers are bewildered by backsolving. It is the
human version of mini-maxing, the process used by all chess playing programs
to "think." Ridiculing the process is not going to enhance one's
understanding of it.
> Houlsby:
> So you're suggesting that BookUp is *as effective* at identifying and
> evaluating critical lines as *Svidler or any lesser GM* is? Even if it
> is, what use is that to me, if I can't understand *why* a tabiya is
> evaluated as +/= or whatever?
You've hit on an important point. If you truly do not understand why any
position (including a tabiya) is evaluated in any particular way then it is
impossible to explain the importance of Backsolving.
Backsolving can only be understood when you get that the "+/= or whatever"
is based on the best line of play in a tree of analysis, and with huge trees
of analysis Backsolving takes the guesswork out of calculating the effect of
an improved line.
> Houlsby:
> >>Backsolving is nothing more, and nothing less, than computer
> analysis,
> >>which is fine as far as it goes, but that may be only a couple of
> >>yards....
>
> Leahy:
> >Backsolving is much more often based on human analysis and can be
> contrasted
> with computer analysis.
>
> Houlsby:
> So how do I, a patzer:
>
> a) understand
>
> and
>
> b) reconcile any conflicting evaluations?
You tell me how a patzer does this. Computer analysis? Reading a GM's
notes? Your original thought? What do you do? What would you suggest for
patzers?
Mike Leahy
"The Database Man!"
www.bookup.com
|
| |
Date: 19 May 2005 13:51:10
From: Mark Houlsby
Subject: Re: Backsolving
|
>> Mike Leahy's claim for the miraculous effects of backsolving have
been
>> exaggerated for the last 10 years. They were hype 10 years ago and
the
>> backsolving claims are hype today. All backsolving does is
automatically
>> attach an evaluation to each node in the Bookup book. That
evaluation is
>> based on the evaluation at the node where you start the backsolving
Ex;
>> 30.Nxg6 mate and it backtracks up the tree to the node at the
beginning
>> where you want the backsolving to end Ex: 12.Ng5 .
>Thanks, Al.
>That's a quick sweep over the idea of backsolving. For a much more
in-depth
look at backsolving in a practical case I'd recommend watching this
movie:
http://www.bookup.com/chessvid=ADeo8.htm
>> I can't remember whether
>> you can specify the begginning node 12.Ng5 or whether you have to
allow
>> backsolving right back to the 1st move.
>You can. I'd have to wonder if you've used Backsolving in the last 10
years.
How effective is backsolving to move 1?
>> However there may be other unplayed moves not in the tree path.
Those good
>> unplayed moves are called novelties. If any one of those is added
to the
>> tree, then the preceding backsolving process is now uselesss and has
to be
>> done again based on the new novelty. Therefore you are never certain
if
the
>> present line of play is good or not because even overnight analysis
won't
be
>> the final decision on the playability of a line.
>You've made a good case for the usefulness of Backsolving.
No, he hasn't, and I suspect that he may agree that he has not.
> The only way to get the certainty you're looking for is to make sure
those "unplayed" moves
are added and their resulting positions are accurately assessed.
You mean, for *all* unplayed novelties, good and bad?! Whoa, baby,
chess would be solved in a month at the most, if that were true.
Unfortunately, it's horse manure.
>If you want to argue that leaving out moves and assessing positions
incorrectly will cause Backsolving problems then I'd have to agree.
And I'd
definitely use Backsolving to help me find those errors in analysis.
For *all* unplayed novelties? Do you actually *understand* what KK is
arguing?
>> Backsolving's questionable
>> merit is that it automates the process of actually attaching an
evaluation
>> at each node. However you don't really need to attach an evaluation
at
each
>> node. Simply manually attach an evaluation to the 12.Ng5 move based
on
the
>> overnight's analysis of the last move in the line 30.Nxg6.
>Imagine doing that task a hundred times or more and you'll see the
need for
the "power tools" now included with Bookup. They're described at
http://www.bookup.com/power_to=ADols.htm It almost sounds like you were
quoting the second paragraph from that page. :)
KK is arguing that it's necessary to do it for *all* unplayed lines.
> Backsolving is only beneficial to complete robot like idiots that
can't
> remember what line they are in at any particular moment of time.
These
robot
> like idiots would need to have an evaluation at every possible node
because
> they are like an Alzheimer patient who would't be able to remember
where
> they are in the tree. Backsolving adds an unnecessary amount of
bytes to
> the Bookup tree without any real merit.
> Komputer Korner
>You've just inferred that quite a few titled players are "complete
robot
like idiots."
Has he? Does, for example, GM Svidler rely on backsolving *rather than
his own analysing capabilities*?
Where is your evidence for this?
>Next you'll be saying that Bookup's training wizard is only good for
players
who have burned up their long term memories with dissolute living. :)
Don't be silly.
Backsolving is nothing more, and nothing less, than computer analysis,
which is fine as far as it goes, but that may be only a couple of
yards....
k Houlsby
"The man who *still* doesn't define himself in terms of the software he
uses!"
|
| | |
Date: 19 May 2005 21:41:42
From: Mike Leahy
Subject: Re: Backsolving
|
"k Houlsby" <[email protected] > wrote in message
news:[email protected]...
>> Mike Leahy's claim for the miraculous effects of backsolving have
been
>> exaggerated for the last 10 years. They were hype 10 years ago and
the
>> backsolving claims are hype today. All backsolving does is
automatically
>> attach an evaluation to each node in the Bookup book. That
evaluation is
>> based on the evaluation at the node where you start the backsolving
Ex;
>> 30.Nxg6 mate and it backtracks up the tree to the node at the
beginning
>> where you want the backsolving to end Ex: 12.Ng5 .
>Thanks, Al.
>That's a quick sweep over the idea of backsolving. For a much more
in-depth
look at backsolving in a practical case I'd recommend watching this
movie:
http://www.bookup.com/chessvid�eo8.htm
>How effective is backsolving to move 1?
Just as effective as it is to any other move.
I think this is one of the best uses of Bookup. As often as not, when the
contents of a popular opening book are backsolved, the starting position of
chess (move 1) is often tagged as winning for White or winning for Black.
Backsolving makes the critical line in any analysis stand out like a sore
thumb. Then the unplayed moves that Al Tomalty (Komputer Korner) mentioned
will scream to be added.
>> However there may be other unplayed moves not in the tree path.
Those good
>> unplayed moves are called novelties. If any one of those is added
to the
>> tree, then the preceding backsolving process is now uselesss and has
to be
>> done again based on the new novelty. Therefore you are never certain
if
the
>> present line of play is good or not because even overnight analysis
won't
be
>> the final decision on the playability of a line.
>>You've made a good case for the usefulness of Backsolving.
>No, he hasn't, and I suspect that he may agree that he has not.
I wasn't expecting his agreement. I tend to stick to topics on chess
software and not to topics on who likes/needs/understands chess software. :)
Al is correct that adding a new move means "the preceding backsolving
process is now uselesss and has
to be done again based on the new novelty." Fortunately you can and should
do it again, and it can be done in real time just by moving backward through
the tree from the novelty. Clicking the button that takes you back to the
starting position will backsolve it in an instant. Take a look at the video
mentioned above for an example.
> The only way to get the certainty you're looking for is to make sure
those "unplayed" moves
are added and their resulting positions are accurately assessed.
>You mean, for *all* unplayed novelties, good and bad?! Whoa, baby,
>chess would be solved in a month at the most, if that were true.
>Unfortunately, it's horse manure.
I agree. Adding all unplayed novelties would be horse manure.
That's why I said "those" unplayed moves, as in those moves Al said needed
to be added to improve upon the existing move(s).
> Backsolving is only beneficial to complete robot like idiots that
can't
> remember what line they are in at any particular moment of time.
These
robot
> like idiots would need to have an evaluation at every possible node
because
> they are like an Alzheimer patient who would't be able to remember
where
> they are in the tree. Backsolving adds an unnecessary amount of
bytes to
> the Bookup tree without any real merit.
> Komputer Korner
>>You've just inferred that quite a few titled players are "complete
>>robot
>>like idiots."
>Has he? Does, for example, GM Svidler rely on backsolving *rather than
>his own analysing capabilities*?
>Where is your evidence for this?
You'd have to ask GM Svidler that.
As far as discussing software, Bookup's Backsolving would be very handy for
pointing out the critical line(s) in Svidler's or anyone's analysis as well
as demonstrating the impact on the analysis when an improvement is found.
>>Next you'll be saying that Bookup's training wizard is only good for
players
>>who have burned up their long term memories with dissolute living. :)
>Don't be silly.
You got my point.
>Backsolving is nothing more, and nothing less, than computer analysis,
>which is fine as far as it goes, but that may be only a couple of
>yards....
Backsolving is much more often based on human analysis and can be contrasted
with computer analysis.
When I can beat the computers 9 out of 10 then I'll start ignoring their
contribution.
>k Houlsby
>"The man who *still* doesn't define himself in terms of the software he
>uses!"
Mike Leahy
"The Database Man!"
www.bookup.com
|
| | | |
Date: 19 May 2005 21:44:42
From: Mike Leahy
Subject: Re: Backsolving
|
I just realized that I posted the wrong movie. The one that explains
Backsolving in a real situation is at:
http://www.bookup.com/chessvideo9.htm
Mike Leahy
"The Database Man!"
www.bookup.com
|
| |
Date: 06 Apr 2005 06:58:58
From: [email protected]
Subject: Re: Computers and opening theory
|
Another application of computers to opening theory is storing up a vast
amount of information in an easy to access way. The best I have found
so far is the Chess Assistant direct tree, so you can see all the
statistics about which moves have been played in a given position, and
what the results were.
To improve on that, all the verbal explanations of opening positions
could also be collected and displayed in such a tree, I know that could
be done in Bookup as well as Chess Assistant, and probably other
database softwares could also do that. Ignoring the problem with
copyright laws, all the accumulated chess knowledge of centuries, which
has been written down, could eventually be collected and displayed in a
single such tree, so that might make a good project for someone
(someone else will have to figure out how the copyright laws would
affect such a project, that is not my area of expertise)
Back to your original question, the fun was taken out of some lines in
the Black-Diemer Gambit after computer chess software starting
playing at the grandmaster level.
|
| | |
Date: 07 Apr 2005 10:02:02
From: Major Cat
Subject: Re: Computers and opening theory
|
Thanks for pointing out the usefulness of the dedicated database approach
in providing quantitative _positional_ evaluation indicators, albeit
indirectly. Although not perfect, such statistics can encapsulate a fair bit
of over the board experience with any particular opening line. It appears,
then, that computers can be quite useful as powerful bookkeepers in their own
right and need not be thought of or used as _direct_ evaluators in every
instance.
Regards
Major Cat
[email protected] wrote:
>
> Another application of computers to opening theory is storing up a vast
> amount of information in an easy to access way. The best I have found
> so far is the Chess Assistant direct tree, so you can see all the
> statistics about which moves have been played in a given position, and
> what the results were.
>
> To improve on that, all the verbal explanations of opening positions
> could also be collected and displayed in such a tree, I know that could
> be done in Bookup as well as Chess Assistant, and probably other
> database softwares could also do that. Ignoring the problem with
> copyright laws, all the accumulated chess knowledge of centuries, which
> has been written down, could eventually be collected and displayed in a
> single such tree, so that might make a good project for someone
> (someone else will have to figure out how the copyright laws would
> affect such a project, that is not my area of expertise)
>
> Back to your original question, the fun was taken out of some lines in
> the Black-Diemer Gambit after computer chess software starting
> playing at the grandmaster level.
|
| | | |
Date: 07 Apr 2005 15:58:14
From: David Richerby
Subject: Re: Computers and opening theory
|
Major Cat <[email protected] > wrote:
> Thanks for pointing out the usefulness of the dedicated database
> approach in providing quantitative _positional_ evaluation indicators,
> albeit indirectly. Although not perfect, such statistics can encapsulate
> a fair bit of over the board experience with any particular opening
> line. It appears, then, that computers can be quite useful as powerful
> bookkeepers in their own right and need not be thought of or used as
> _direct_ evaluators in every instance.
You have to be careful with this kind of thing, though. A variation may
be popular and successful for one side until somebody refutes it in a
tournament. Assuming the refutation is sound, nobody will play the
variation again in serious play so the statistics will still look very
promising -- a reasonable number of games played and only one extra loss
from the refuting game -- but the variation itself is not good. You might
be able to detect this because the last game in your database featuring
the variation was some time ago but that might just mean that it went out
of fashion for no particular reason.
I think Mike Leahy said that his Bookup software has analysis features
that check for such refutations but I'm not sure how that works.
Dave.
--
David Richerby Indelible Erotic Peanut (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a roasted nut but it's genuinely
erotic and it can't be erased!
|
| | | | |
Date: 13 Apr 2005 15:38:14
From: Mike Leahy
Subject: Re: Computers and opening theory
|
"David Richerby" <[email protected] > wrote in message
news:wjC*[email protected]...
> Major Cat <[email protected]> wrote:
> > Thanks for pointing out the usefulness of the dedicated database
> > approach in providing quantitative _positional_ evaluation indicators,
> > albeit indirectly. Although not perfect, such statistics can encapsulate
> > a fair bit of over the board experience with any particular opening
> > line. It appears, then, that computers can be quite useful as powerful
> > bookkeepers in their own right and need not be thought of or used as
> > _direct_ evaluators in every instance.
>
> You have to be careful with this kind of thing, though. A variation may
> be popular and successful for one side until somebody refutes it in a
> tournament. Assuming the refutation is sound, nobody will play the
> variation again in serious play so the statistics will still look very
> promising -- a reasonable number of games played and only one extra loss
> from the refuting game -- but the variation itself is not good. You might
> be able to detect this because the last game in your database featuring
> the variation was some time ago but that might just mean that it went out
> of fashion for no particular reason.
>
> I think Mike Leahy said that his Bookup software has analysis features
> that check for such refutations but I'm not sure how that works.
That's more than I can put in a post, but a start would be to check out this
article:
http://www.bookup.com/backsolv.htm
To check for refutations, one does overnight analysis of the entire
repertoire and then uses one the "power tools" from Bookup 2000 Professional
to compare the backsolved Informant rate symbol with the backsolved numeric
assessment from the engine used to analyze it overnight. You pick how far
out of whack the symbol and assessment should be before the position is
tagged as a possible novelty/refutation.
Mike Leahy
"The Database Man!"
www.bookup.com
www.chessopeningspgn.com
|
| | | | | |
Date: 10 May 2005 13:40:23
From: Komputer Korner
Subject: Re: Backsolving
|
Mike Leahy's claim for the miraculous effects of backsolving have been
exaggerated for the last 10 years. They were hype 10 years ago and the
backsolving claims are hype today. All backsolving does is automatically
attach an evaluation to each node in the Bookup book. That evaluation is
based on the evaluation at the node where you start the backsolving Ex;
30.Nxg6 mate and it backtracks up the tree to the node at the beginning
where you want the backsolving to end Ex: 12.Ng5 . I can't remember whether
you can specify the begginning node 12.Ng5 or whether you have to allow
backsolving right back to the 1st move. If you have to allow backsolving
right back to the 1st move then backsolving is less than useless because you
may not want to change the evaluations of the moves before the 12.Ng5 move.
However there may be other unplayed moves not in the tree path. Those good
unplayed moves are called novelties. If any one of those is added to the
tree, then the preceding backsolving process is now uselesss and has to be
done again based on the new novelty. Therefore you are never certain if the
present line of play is good or not because even overnight analysis won't be
the final decision on the playability of a line. Backsolving's questionable
merit is that it automates the process of actually attaching an evaluation
at each node. However you don't really need to attach an evaluation at each
node. Simply manually attach an evaluation to the 12.Ng5 move based on the
overnight's analysis of the last move in the line 30.Nxg6. You won't be
investigating any of the moves after 12.Ng5 anyway because you have decided
that 12.Ng5 won't be in your repertoire. If you do decide to investigate
moves in the 12.Ng5 line, you don't need automatic evaluations beside them
because you already have attached an evaluation to the 12.Ng5 move.
Backsolving is only beneficial to complete robot like idiots that can't
remember what line they are in at any particular moment of time. These robot
like idiots would need to have an evaluation at every possible node because
they are like an Alzheimer patient who would't be able to remember where
they are in the tree. Backsolving adds an unnecessary amount of bytes to
the Bookup tree without any real merit.
Komputer Korner
"Mike Leahy" <[email protected] > wrote in message
news:[email protected]...
>
> "David Richerby" <[email protected]> wrote in message
> news:wjC*[email protected]...
> > Major Cat <[email protected]> wrote:
> > > Thanks for pointing out the usefulness of the dedicated database
> > > approach in providing quantitative _positional_ evaluation indicators,
> > > albeit indirectly. Although not perfect, such statistics can
encapsulate
> > > a fair bit of over the board experience with any particular opening
> > > line. It appears, then, that computers can be quite useful as powerful
> > > bookkeepers in their own right and need not be thought of or used as
> > > _direct_ evaluators in every instance.
> >
> > You have to be careful with this kind of thing, though. A variation may
> > be popular and successful for one side until somebody refutes it in a
> > tournament. Assuming the refutation is sound, nobody will play the
> > variation again in serious play so the statistics will still look very
> > promising -- a reasonable number of games played and only one extra loss
> > from the refuting game -- but the variation itself is not good. You
might
> > be able to detect this because the last game in your database featuring
> > the variation was some time ago but that might just mean that it went
out
> > of fashion for no particular reason.
> >
> > I think Mike Leahy said that his Bookup software has analysis features
> > that check for such refutations but I'm not sure how that works.
>
> That's more than I can put in a post, but a start would be to check out
this
> article:
>
> http://www.bookup.com/backsolv.htm
>
> To check for refutations, one does overnight analysis of the entire
> repertoire and then uses one the "power tools" from Bookup 2000
Professional
> to compare the backsolved Informant rate symbol with the backsolved
numeric
> assessment from the engine used to analyze it overnight. You pick how far
> out of whack the symbol and assessment should be before the position is
> tagged as a possible novelty/refutation.
>
>
> Mike Leahy
> "The Database Man!"
> www.bookup.com
> www.chessopeningspgn.com
>
>
|
| | | | | | |
Date: 19 May 2005 20:38:04
From: Mike Leahy
Subject: Re: Backsolving
|
"Komputer Korner" <[email protected] > wrote in message
news:[email protected]...
> Mike Leahy's claim for the miraculous effects of backsolving have been
> exaggerated for the last 10 years. They were hype 10 years ago and the
> backsolving claims are hype today. All backsolving does is automatically
> attach an evaluation to each node in the Bookup book. That evaluation is
> based on the evaluation at the node where you start the backsolving Ex;
> 30.Nxg6 mate and it backtracks up the tree to the node at the beginning
> where you want the backsolving to end Ex: 12.Ng5 .
Thanks, Al.
That's a quick sweep over the idea of backsolving. For a much more in-depth
look at backsolving in a practical case I'd recommend watching this movie:
http://www.bookup.com/chessvideo8.htm
> I can't remember whether
> you can specify the begginning node 12.Ng5 or whether you have to allow
> backsolving right back to the 1st move.
You can. I'd have to wonder if you've used Backsolving in the last 10
years.
> However there may be other unplayed moves not in the tree path. Those good
> unplayed moves are called novelties. If any one of those is added to the
> tree, then the preceding backsolving process is now uselesss and has to be
> done again based on the new novelty. Therefore you are never certain if
the
> present line of play is good or not because even overnight analysis won't
be
> the final decision on the playability of a line.
You've made a good case for the usefulness of Backsolving. The only way to
get the certainty you're looking for is to make sure those "unplayed" moves
are added and their resulting positions are accurately assessed.
If you want to argue that leaving out moves and assessing positions
incorrectly will cause Backsolving problems then I'd have to agree. And I'd
definitely use Backsolving to help me find those errors in analysis.
> Backsolving's questionable
> merit is that it automates the process of actually attaching an evaluation
> at each node. However you don't really need to attach an evaluation at
each
> node. Simply manually attach an evaluation to the 12.Ng5 move based on
the
> overnight's analysis of the last move in the line 30.Nxg6.
Imagine doing that task a hundred times or more and you'll see the need for
the "power tools" now included with Bookup. They're described at
http://www.bookup.com/power_tools.htm It almost sounds like you were
quoting the second paragraph from that page. :)
> Backsolving is only beneficial to complete robot like idiots that can't
> remember what line they are in at any particular moment of time. These
robot
> like idiots would need to have an evaluation at every possible node
because
> they are like an Alzheimer patient who would't be able to remember where
> they are in the tree. Backsolving adds an unnecessary amount of bytes to
> the Bookup tree without any real merit.
> Komputer Korner
You've just inferred that quite a few titled players are "complete robot
like idiots."
Next you'll be saying that Bookup's training wizard is only good for players
who have burned up their long term memories with dissolute living. :)
Mike Leahy
"The Database Man!"
www.bookup.com
> "Mike Leahy" <[email protected]> wrote in message
> news:[email protected]...
> >
> > "David Richerby" <[email protected]> wrote in message
> > news:wjC*[email protected]...
> > > Major Cat <[email protected]> wrote:
> > > > Thanks for pointing out the usefulness of the dedicated database
> > > > approach in providing quantitative _positional_ evaluation
indicators,
> > > > albeit indirectly. Although not perfect, such statistics can
> encapsulate
> > > > a fair bit of over the board experience with any particular opening
> > > > line. It appears, then, that computers can be quite useful as
powerful
> > > > bookkeepers in their own right and need not be thought of or used as
> > > > _direct_ evaluators in every instance.
> > >
> > > You have to be careful with this kind of thing, though. A variation
may
> > > be popular and successful for one side until somebody refutes it in a
> > > tournament. Assuming the refutation is sound, nobody will play the
> > > variation again in serious play so the statistics will still look very
> > > promising -- a reasonable number of games played and only one extra
loss
> > > from the refuting game -- but the variation itself is not good. You
> might
> > > be able to detect this because the last game in your database
featuring
> > > the variation was some time ago but that might just mean that it went
> out
> > > of fashion for no particular reason.
> > >
> > > I think Mike Leahy said that his Bookup software has analysis features
> > > that check for such refutations but I'm not sure how that works.
> >
> > That's more than I can put in a post, but a start would be to check out
> this
> > article:
> >
> > http://www.bookup.com/backsolv.htm
> >
> > To check for refutations, one does overnight analysis of the entire
> > repertoire and then uses one the "power tools" from Bookup 2000
> Professional
> > to compare the backsolved Informant rate symbol with the backsolved
> numeric
> > assessment from the engine used to analyze it overnight. You pick how
far
> > out of whack the symbol and assessment should be before the position is
> > tagged as a possible novelty/refutation.
> >
> >
> > Mike Leahy
> > "The Database Man!"
> > www.bookup.com
> > www.chessopeningspgn.com
> >
> >
>
>
|
|
