(Sterten) wrote in message ...


If one player permanently looses by time, this will be
taken in account, I don't ignore the result.
But if one player permanently looses on time and played
well, he should get a better rating than a player
who permanently looses on time and playes poorly.
This reflects the expectation of his future perform:
he once might learn how to manage his time-problems.

Sure, and a player who plays bad moves once might learn how to play
good moves. So why not give him a high rating, one that reflects the
expectation of his future performance.

Tobi

CeeBee wrote:

(Sterten) wrote in rec.games.chess.misc:

I don't see CeeBee's point:
"it's not based on differences alone, so it can't work."

No. You replace a system that is based on differences in strength from
_actual_ play with differences in strenght from a computer benchmark
crudely translated from differences in strenght in actual play. That
doesn't give more "absolute" results, but worse results.

sorry, I don't understand.
My system should also consider the ratings of the opponents
before that game , of course.
How can you know, that it's "crudely" translated ?

Your logical error is that you still want to translate strength
differences to "more objective values", which mixes up the basic idea of
absolute strength with differences in strength.

I'm probably too dumb to detect a logical error here.
How can the struggle for more objective values
or mixing up something be a logical error ?

Sure it's based on differences, you have to fix
a range.
You can't know how good it works, since
we've not discussed the details yet. If my system just
weight's the result with 1 and the move-quality by an infinitesam
amout, then it's almost equal to the existing
system. These parameters have to be optimised.

If one player permanently looses by time, this will be
taken in account, I don't ignore the result.
But if one player permanently looses on time and played
well, he should get a better rating than a player
who permanently looses on time and playes poorly.
This reflects the expectation of his future perform:
he once might learn how to manage his time-problems.


But you wanted a more absolute rating, and now you introduce the notion
about possible future performance. You suggest that knowing the moves is
enough to be qualified as a good player. This is not the case. If my
opponent -who is as strong as me- thinks longer that me in an actual game,
he finds better solutions, and might win.

if he "knows the moves" , then why should he reflect so long ?

However the clock prevents him
from doing that. So he loses on time. And now you state that shouldn't be
taken into account, and he's the stronger player.

No, I didn't.
I think that exceeding the time should be valued as just
making one very big mistake. This is the best we can do, since the
reflection times for each move are usually not reported.

In that case your system leads to a qualification "ceebee's opponent is
stronger than cee, because he found out better moves in an actual game,
yet not within his allotted time". But if we both stuck to our time, we
would have drawn, because we're equal in strength.

how do you know this ?
If you make worse moves 39 times in a game, and then your opponent
makes one big error in move 40, so you win ,
would you conclude that your opponent is good to make 39 moves
in 2 hours but not on 40 moves in 2 hours ?
Exceeding the time in move 40 is almost random, on move 35 you can't
know whether it will happen or not.

The Elo system values
that correctly, your system doesn't.

Frankly, the fact that players exceed the time at all shows, that there
is something wrong , IMO.
These players don't care so much about the result but more about
the quality of their moves in the game.
But the rating system assumes, that they only play for the result.

It's just unreasonable to loose by time permanently,
there is nothing you can do, if someone wants to
outfox the system by playing well but then resigning
or loosing by time.. But that's unreasonable.
It's just rated as a bad move.


suppose there only multi-game matches, no tournaments.
System A only rates the outcome of a match,
which player won. System B also takes in account the
score, so a 10:0 gives more ELO-points than a 5.5:4.5 .
Which system would you prefer ?


Neither of the systems is a correct description of the Elo system. The Elo
system takes into account both individual results _and_ opponent's
strength. A 10:0 victory by Kasparov over a patzer is less worth for his
Elo points than a 5,5:4,5 win in a 2700+ opponent match.

Of course, I was assuming that the same players were playing.

Let me explain in more lenght.

Your idea was:
"The computer analyses the positions and rates every single move of
the game and finally calculates a rating-number for both players
and that game based on the moves rather than the result."

sorry, it should be : "rather than the result alone"
as the example with time-exceeding shows.

Recapitulating you suggest to replace the current rating system with a
more objective benchmark by computers based om move valuing.

yes. And I was asking whether similar things are already being done
or experimented with.

I have explained that the current rating system is not about strength ,
but about strength differences.

and thus also about strength.
If you are stronger than anyone else, then you are "strong"
in usual language.

Even better: this is the prime objective
of the system. Why then are people so convinced that ratings tell you
something about absolute strength?

statistical indication for this is overwhelming.

It's because chess player pools are so fluently intermixing with each
other, both in time as in location. It won't surprise you that a 2200
player from say Australia is often on par with a 2200 player from say
France. Players mix with each other in worldwide tournaments, they mix
with players at home and de facto those players mix with players worldwide
- as a result ratings are often leveled. Older players with rating
established against retired players play younger players and thus transfer
those strength differences from one age pool to the other.

yes

But it does _not_ mean the rating system gives an accurate measure of
strenght.

then, what was the reason that you included the previous paragraph ?
You were showing, that it works quite well, although not accurate.
Nothing is 100% accurate.

Differences of the same order just mix throughout pools, but they stay
differences in strength. Often people don't understand that: they want to
compare Fischer with Kasparov. But to no avail: the playing pool of
Fischer is too much disconnected from Kasparov's pool of opponents. Their
rating difference has no meaning.

You mean "less meaning" . They are still "connected".
The reliability of the system spreads a bit over time.
Didn't I see here recently Jerry Spinrad with a rating system
for ancient players back to the 1830s ?
I'm sure, he would not have posted if it weren't somehow reliable.

Sometimes you'll read here that ratings are inflated, because in earlier
day you were a top grandmaster at 2600, while now you have to be 2700 to
be very strong.

no surprise. Players have computers available today for preparing,
learning,training,storing.
Also, when the number of players increases, then the spread :
best-weakest also (usually) increases.

Of course this is nonsense, the rating calculation hasn't been changed: a
difference of 200 still means the same in winning/losing chances as 30
years ago. Rating inflation would mean that a bigger difference is needed
to have the same winning chances than in the past.

depends, how you define it

That's only possible if
the calculation method has changed, which isn't the case.

Difference, not absolute strenght.


Now you want to develop a system that values chessplayers on individual
moves. That is not a new idea,

has it already been tried in practice ? References ?

but what does it mean? It means you have to
know what is important to be strong. Tactics? Positional knowledge?
Recognition of standard patterns and characteristics? Knowledge of opening
theory? Knowledge of endgames? Knowledge of games from past masters?
Knowledge to not find one crucial move, but calculate the actual and
correct move sequence? The ability to play a game without a losing move
after ten strong moves? Psychological strenght in a game? Physical fitness
during a tournament? The ability to think undisturbed in a noisy room? The
speed at which you solve a problem? The number of games you're able to
play at a constant level?
All these things determine the strenght of a chess player, and even more.

most of these are not open to computer measurement after the game.
And they are _not_ considered by the current ELO-system.
If you think, they should be considered, then you are just advocating
"my" rating system.

First problem: what is their comparative weight? You suggest that a
computer can tell you better how strong a chess player is in a 20 games
round robin tournament, at fast time controls, in a cold playing arena
with thousand spectators on an uncomfortable chair, against a very strong
and impressive opponent than the Elo-system.

The ELO-system can't. "My" system _maybe_ can handle this better,
but certainly not worse.

How would that be in a luxurious environment in one's hometown, against a
homesick opponent?

As important is the question of chosing and especially valuing those
moves.
Giving a value to a move is referring to another standard. Which standard
is that? What is the value of move one? And what of move two? Why would
move one be more valuable than move two?

as a start, just take the usual computer evaluations in pawn-units

In practice you arrive at determining differences in strenght between
moves. It means you suppose an arbitrary difference is better than a
difference based on actual play.

?

But we want to know strength in actual
play, and not in a theoretical situation.The proof of the pudding is in
the eating, not in knowing the recepy by heart.

only the actually played moves are evaluated, not any hypothetical
moves, or moves of another game.

Weird but true, but there is a system for just that: the Elo-system.

the ELO-system is only based on the result of a game and discards
their moves.

Your
computer system doesn't do that. It can give you a player that scores
better than Kasparov yet loses every game against him.

unlikely. And less likely than with the ELO-system , IMO.

Put them both back
around your computer test and the results will be the same: your player
is better in the computer test than Kasparov.

?

And worse, your system has
no way of dealing with that discrepancy.

But not with the current rating system you consider inferior: that system
takes into account evey win and loss. In the above example, your tested
player will see his rating drop rapidly below that of Kasparov, and _that_
tells you even more about his actual strenght than your benchmark. And
that while the Elo system wasn't even developed for that purpose.

You made two error in your idea: first of all you consider the current
rating system as a system to determine strength, which is not true, and
secondly you suppose that a computer valued rating without a pool of peers
and without proper understanding of all determining factors can give
_better_ information than the current rating system.

if we can't consider _all_ factors, then we should try to determine
_as many as possible_ , right ? Neglecting the moves is like ostrich
putting the head into the sand.

Maybe one day we will be able to establish all factors defining absolute
strength, but the current batch of computer programs is certainly no match
for the well estabshed and proven Elo rating system.

computers have developed a lot since that system was established.
Time to adapt it to the actual situation.


Guenter Stertenbrink

(Sterten) wrote in rec.games.chess.misc:


sorry, I don't understand.
My system should also consider the ratings of the opponents
before that game , of course.


So you just replaced one rating _difference_ system with another. That is
already happening in the Elo system.

You started out with a system with which "You can even rate two players
who both play their very first game". because " rating becomes more
absolute". But now the system should consider ratings of opponents before
that very first game. But these players didn't play opponents, as it is
their very first game.



Your logical error is that you still want to translate strength
differences to "more objective values", which mixes up the basic idea
of absolute strength with differences in strength.

I'm probably too dumb to detect a logical error here.
How can the struggle for more objective values
or mixing up something be a logical error ?


I fail to see the relevance. I haven't suggested that the struggle for
establishing more objective measurements is a logical error. The struggle
for more objective measurements leads you to an adapted system. The
_adapted system_ is based on logical errors, not the _wish_ to develop a
system.



if he "knows the moves" , then why should he reflect so long ?

He doesn't "know" the move, he has to think longer to determine the right
move. In chess that is very often as critical for the result as making a
blunder.



However the clock prevents him
from doing that. So he loses on time. And now you state that
shouldn't be taken into account, and he's the stronger player.

No, I didn't.
I think that exceeding the time should be valued as just
making one very big mistake. This is the best we can do, since the
reflection times for each move are usually not reported.


This has been taken care of in the Elo system as well. A very big mistake
resulting in a loss diminishes your rating.



In that case your system leads to a qualification "ceebee's opponent
is stronger than cee, because he found out better moves in an actual
game, yet not within his allotted time". But if we both stuck to our
time, we would have drawn, because we're equal in strength.

how do you know this ?

It is the consequence of your own proposed system.


Frankly, the fact that players exceed the time at all shows, that
there is something wrong , IMO.
These players don't care so much about the result but more about
the quality of their moves in the game.
But the rating system assumes, that they only play for the result.


Do you have a proper idea of the rules of chess in the first place? The
object of the game under the current rules is to mate the king in the
alotted time, not to make the most beautiful moves.

And why do you say "they only play for the result"? You started out this
thread with the assumption that the current rating system does not
calculate a player's chess strength, which you want to determine more
accurately, and now you suddenly discard this notion that strength, which
is per chess rules defined as the ability to win the game, is irrelevant?



sorry, it should be : "rather than the result alone"
as the example with time-exceeding shows.

You are comparing it with the current rating system. _That_ is the point.
You suggest that the current rating system tells us insufficient about
individual chess strength and want to replace it with a better system that
discards the sole determinant of chess strenght: winning a game.


Recapitulating you suggest to replace the current rating system with
a more objective benchmark by computers based om move valuing.

yes. And I was asking whether similar things are already being done
or experimented with.


Yes, there are several programs with rating features like that. E.g.
ChessMaster has it. Try to find out a move and see your estimated rating.
It's a bypass however, like I described before.


I have explained that the current rating system is not about strength
, but about strength differences.

and thus also about strength.
If you are stronger than anyone else, then you are "strong"
in usual language.


And what does that prove? That "strength" and "strength difference" have
the same meaning? They don't. They still have separate meanings.
The problem is that you don't understand this crucial difference in
relation to consequences for your rating system.




Even better: this is the prime objective
of the system. Why then are people so convinced that ratings tell you
something about absolute strength?

statistical indication for this is overwhelming.


This is not the case. There is no "overwhelming statistical indication"
that Elo rating is a measure of absolute strenght. Elo-rating is a
measure of strength difference. This is no opinion, but the _definition_
of Elo-ratings. There cannot be any indication at all that Elo ratings
indicate absolute strength.

It's like claiming that there is overwhelming statistical indication that
the _difference_ in height between a child and me tells me something about
my actual _absolute_ heigth.



But it does _not_ mean the rating system gives an accurate measure of
strenght.

then, what was the reason that you included the previous paragraph ?
You were showing, that it works quite well, although not accurate.
Nothing is 100% accurate.


The reason of my explanation is to clarify where your misconception about
ratings come from.

I didn't accuse you of proposing a system which is not 100% accurate, I
tried to explain that the system you started out with gives even worse
results than the current system, _even_ if that current system isn't
occupying itself with absolute strength measurements.

That does not depend on details of your system I'm not filled in about, it
is depending on a flaw in the systematic. Knowing the details doesn't
change the flawed systematic.


You mean "less meaning" . They are still "connected".
The reliability of the system spreads a bit over time.
Didn't I see here recently Jerry Spinrad with a rating system
for ancient players back to the 1830s ?
I'm sure, he would not have posted if it weren't somehow reliable.


I'm sorry, I'm losing you here. You mean that a faulty translation of
current ratings to historical players is a proof of reliability? You
suggest that someone who who makes the same wrong assumptions as you
proves you right?


Sometimes you'll read here that ratings are inflated, because in
earlier day you were a top grandmaster at 2600, while now you have to
be 2700 to be very strong.

no surprise. Players have computers available today for preparing,
learning,training,storing.
Also, when the number of players increases, then the spread :
best-weakest also (usually) increases.


After me explaining that the ratings are NOT inflated, you answer that the
inflation rating I _don't_ talk about is "no surprise" because of better
training facilities and bigger player pools.....

I just explained that the rating system has _not_ inflated; not by
"feeling" but by definition. There is no rating inflation. There are just
different ratings.


Of course this is nonsense, the rating calculation hasn't been
changed: a difference of 200 still means the same in winning/losing
chances as 30 years ago. Rating inflation would mean that a bigger
difference is needed to have the same winning chances than in the
past.

depends, how you define it


This is getting weirder and weirder. Are you suggesting the Elo rating
system is an arbitrary way of dealing out ratings, and an Elo strenght
difference of 200 means _this_ in part A of the world aor history and
_that_ in part B? Do you have proof that a difference of 200 Elo points
gives different winning/losing chances in 2003 if compared to 1973?

I'm sorry, but you are unable to prove that. You can only prove that if
you prove that the Elo rating algoritms have changed. And they haven't.

Basically you have come down to refining the current Elo system with
rating points added or substracted based on the ability to find certain
moves in a certain time. If that is implemented it _might_ show a
correlation with a better prediction of winning chances (and thus strength
differences), but it still doesn't tell you more about absolute strenght.
Still it needs the pool of players.


It is possible to develop a benchmark to test various chess skills based
on moves. Computer test suites are an example of that. And they all
compare strength. Strength in the ability to solve that particular test
set.

Chess is a game of two people trying to mate the king on the board. Those
are the established rules, this is not an interpretation of facts. The
strongest player is the player with the biggest chance to win the game,
because he has the biggest overal strength.

That strenght depends on very many fators, some even still unknown.
The system you propose won't give you extra information on the crucial
win/lose chances in chess. Even worse, it will distort that information
and replace it with values less relevant to determine the ability to win.


--
CeeBee


Uxbridge: "By God, sir, I've lost my leg!"
Wellington: "By God, sir, so you have!"


Google CeeBee @ www.geocities.com/ceebee_2

On 13 Oct 2003 10:49:50 GMT, CeeBee wrote:

No. You replace a system that is based on differences in strength from
_actual_ play with differences in strenght from a computer benchmark
crudely translated from differences in strenght in actual play. That
doesn't give more "absolute" results, but worse results.

It is fairly easy to see that the strength or quality of one's moves
in a single game bears little relation to one's overall chess
abilities. For instance I once, as white, played the following game,
or gamlet:

1.e4 e5 2.Nf3 d6 3.Bc4 h6?! 4.Nc3 Bg4? 5.Nxe5 Bxd1?? 6.Bxf7+ Ke7
7.Nd5#

Now, if I put this to our chess computer it will find that I, as
white, played pretty much perfectly and possibly give me a "rating" of
3000 or more. Of course the whole game had been played many times
before and I had basicly memorized it and so it required just about no
thought on my part at all. At the time I was rated in the 1800s.

And this is true of any single game. Sometimes I really play quite
well and sometimes I play like an idiot and my "rating" is a
prediction of my future *results* based on my overall past *results*.

However, if we get a very strong computer to go over a few thousand or
perhaps a few million games of all available players and then compute
and save in a database what we might call an "accuracy index" for each
game and an overall "accuracy index" for each player we will,
presumably, find some correlation between this index and the player's
actual rating.

If so we can then get the computer to rank players according to
overall "accuracy index" and rank all the players accordingly.

Now since this is a computer analysis what it will really be returning
is an assessment of each player's tactical accuracy. I suspect that
at the world championship level each player's index will be so high as
to produce quite a low correlation between index and rating and so we
might find that, so far as distinguishing between world championship
level players the "index" is pretty well useless. At that level we
might well discover that there are other factors than mere "accuracy"
that affect results more.

Take Lasker for instance. Always playing second best openings he
still manages to get massive results against the greatest players of
his day. Yet a computer might well give him a significantly lower
"accuracy index" than these other greats of his day based on his
relatively poor opening play. Or it might not.

The results would be extremely interesting, I think. It would be very
nice if someone did them and published the results. It would take a
long time and a computer dedicated to the task I imagine. Maybe the
results could give someone fodder for a doctorate thesis. Or maybe
not, I don't really know enough about such things to tell.

BUT the only thing a computer can get from analyzing my or anyone's
games move by move would be some kind of accuracy score. That such a
score would reflect my or anyone else's actual results in tournament
play is merely a hypothesis and at the present time I think it is a
hypothesis that is unsupported by sufficient evidence to accept
blindly.

Certainly the idea of computing one's rating and deciding one's
strength from a tactical analysis of a single game is naive at best.

In usual chess position there is no absolutely right move that is the only
way to win or to draw (not to lose).
Computer chooses one of them as the best only as "it seems not worse than
anything else in the nearest n-plies".
Only critical moves that human or computer can fail to recognize but that
lead to lose are important.
Some critical moves, e.g. oversight of some material, or move that leads to
lose of material, are able
to make the rest moves redundant. The result is known - it is a lose.
For computers this lose can be caused by necessity to sacrifice some figure
to save the nearest game.
So such rating system cannot say if player made perfect move in usual
position because it should
search the game to the end before say all of win is possible, draw is
possible, lose is possible.
The one case of win, draw, lose will come only after critical move. But
there exist psychological
moment for the critical move. If it is a game in a match, if human thinks
about 40th move of 2h game,
so tired, he/she can fail to make proper move and lose the game. But in case
you found
comfortable place and time to think about this only move you can concentrate
better and
find better move than some GM, that doesn't lead to lose. Is it worth to
think about you
success and rating against that GM?

Ed Seedhouse wrote in rec.games.chess.misc:


However, if we get a very strong computer to go over a few thousand or
perhaps a few million games of all available players and then compute
and save in a database what we might call an "accuracy index" for each
game and an overall "accuracy index" for each player we will,
presumably, find some correlation between this index and the player's
actual rating.

If so we can then get the computer to rank players according to
overall "accuracy index" and rank all the players accordingly.


This is another way to determine differences in strenght. Maybe it's a
more accurate way of determining those differences. Although
interesting, it might turn out to be an even more complicated way to
assess strength differences than the current Elo system. And it might
well add little extra knowledge to the meaning of those strenght
differences we already know.

The original poster wanted to use such a databaseto test people against
such a database (even without playing a single game before) to establish
a "more absolute" strength number.

However it will still be a translation of "how does this player
calculate moves in comparison with a lot of other players?" The result
will be "He is accurate at 57%." Now he will be placed in a pool with
players who scored 57% as well. How will he perform in real play? "He
will perform like people with a 57% score" say a 1700 player. Will he
win from a player that has a 54% score? If so, he's stronger.
However we do know that being stronger is not determined by 'winning one
game'. Oner can play hundred games against an opponent, win 60 and lose
40, and one is considered stronger, despite the fourty losses. His
winning chances are 60/40. This is the only practical result of your
calcuation.

So you still come to the same result.
His absolute strenght is not 57%, or 1700, his win/lose chances against
other players will depend on their rating difference with him.

In the end one has taken a long road to arrive at the same point the Elo
rating system took in a fraction of the time.


Certainly the idea of computing one's rating and deciding one's
strength from a tactical analysis of a single game is naive at best.

Not exactly naive I think, but very inaccurate.


--
CeeBee


Uxbridge: "By God, sir, I've lost my leg!"
Wellington: "By God, sir, so you have!"


Google CeeBee @ www.geocities.com/ceebee_2

If my 7 year old, my 4 year old and I each played a single game with
Kasparov, we would all lose. The 0-1 result would give us no information
about any of our strengths because losing a single game to Kasparov is not
an unlikely event for any chessplayer. But anyone who plays chess better
than
we do and looked at the moves of the games would easily see which of us was
stronger. For example, he could just count the frequency of errors at 1-ply
or 2-ply.

I don't think that anyone was proposing eliminating the result-based rating
system. The issue is a software tool using available information to assess
playing strength. Obviously the moves of a game contain more information
about playing strength than just the results of the game. It would be a
very useful function for software to have a rating function and it doesn't
seem like it would be difficult to implement. If results of a chess game
correlate with the quality of the moves played in the game, then analysis of
the moves can be used to predict results. For a fixed number of games, this
will be more accurate than a prediction based on results alone.

"CeeBee" wrote in message
. 6.83...
Ed Seedhouse wrote in rec.games.chess.misc:


However, if we get a very strong computer to go over a few thousand or
perhaps a few million games of all available players and then compute
and save in a database what we might call an "accuracy index" for each
game and an overall "accuracy index" for each player we will,
presumably, find some correlation between this index and the player's
actual rating.

If so we can then get the computer to rank players according to
overall "accuracy index" and rank all the players accordingly.


This is another way to determine differences in strenght. Maybe it's a
more accurate way of determining those differences. Although
interesting, it might turn out to be an even more complicated way to
assess strength differences than the current Elo system. And it might
well add little extra knowledge to the meaning of those strenght
differences we already know.

The original poster wanted to use such a databaseto test people against
such a database (even without playing a single game before) to establish
a "more absolute" strength number.

However it will still be a translation of "how does this player
calculate moves in comparison with a lot of other players?" The result
will be "He is accurate at 57%." Now he will be placed in a pool with
players who scored 57% as well. How will he perform in real play? "He
will perform like people with a 57% score" say a 1700 player. Will he
win from a player that has a 54% score? If so, he's stronger.
However we do know that being stronger is not determined by 'winning one
game'. Oner can play hundred games against an opponent, win 60 and lose
40, and one is considered stronger, despite the fourty losses. His
winning chances are 60/40. This is the only practical result of your
calcuation.

So you still come to the same result.
His absolute strenght is not 57%, or 1700, his win/lose chances against
other players will depend on their rating difference with him.

In the end one has taken a long road to arrive at the same point the Elo
rating system took in a fraction of the time.


Certainly the idea of computing one's rating and deciding one's
strength from a tactical analysis of a single game is naive at best.

Not exactly naive I think, but very inaccurate.


--
CeeBee


Uxbridge: "By God, sir, I've lost my leg!"
Wellington: "By God, sir, so you have!"


Google CeeBee @ www.geocities.com/ceebee_2

"Alexander Belov" wrote in message
...
In usual chess position there is no absolutely right move that is the only
way to win or to draw (not to lose).
Computer chooses one of them as the best only as "it seems not worse than
anything else in the nearest n-plies".
Only critical moves that human or computer can fail to recognize but that
lead to lose are important.
Some critical moves, e.g. oversight of some material, or move that leads
to
lose of material, are able
to make the rest moves redundant. The result is known - it is a lose.
For computers this lose can be caused by necessity to sacrifice some
figure
to save the nearest game.
So such rating system cannot say if player made perfect move in usual
position because it should
search the game to the end before say all of win is possible, draw is
possible, lose is possible.
The one case of win, draw, lose will come only after critical move. But
there exist psychological
moment for the critical move. If it is a game in a match, if human thinks
about 40th move of 2h game,
so tired, he/she can fail to make proper move and lose the game. But in
case
you found
comfortable place and time to think about this only move you can
concentrate
better and
find better move than some GM, that doesn't lead to lose. Is it worth to
think about you
success and rating against that GM?



Agreed

Regards

"David Kane" wrote in rec.games.chess.misc:

But anyone who
plays chess better than
we do and looked at the moves of the games would easily see which of
us was stronger. For example, he could just count the frequency of
errors at 1-ply or 2-ply.

That is true, yet it is not relevant for the current discussion.

I did't doubt that one is able to see strength differences in actual play,
I tried to explain that looking at actual moves is _not a better way_ to
determine _absolute strength_.


I don't think that anyone was proposing eliminating the result-based
rating system.

Again: it wasn't discussed. The original poster suggested adding a
_better_ rating system with "more absolute values". With that he makes two
mistakes: it suggests incorrectly that the current system gives
"inaccurate absolute values". The current Elo system might give
inaccurate values, but _no_ absolute values whatsoever; secondly he
supposes that the validation of individual moves can give a useful
official rating to even players who have never played a game before, which
is also not true, because it might tell you something (very limited) about
chess understanding, but doesn't tell you anything about _absolute_ chess
strenght in actual play.


The issue is a software tool using available
information to assess playing strength.

No, this is not the issue. The issue is the way to reach a correct
assessment of chess strength, disconnected from actual play.

Obviously the moves of a game
contain more information about playing strength than just the results
of the game.

Again: the original poster didn't suggest a _way_ to assess strength, he
suggested a _better_ way to determine playing strength. "Better" meant
"better than the current Elo rating". He then suggested a system that will
result in an even worse assessment of playing strength.


It would be a very useful function for software to have a
rating function and it doesn't seem like it would be difficult to
implement.

Many programs do have a rating function. The Fritz GUI rates you on basis
of results against a "rated" Fritz opponent.
ChessMaster even rates your play with puzzles in which you have to find
the right moves. This rating is however not based on objective strenght
determinants but on a comparison with a pool of players solving it with a
similar rating.
It's a derivative of the existing rating system.


If results of a chess game correlate with the quality of
the moves played in the game, then analysis of the moves can be used
to predict results. For a fixed number of games, this will be more
accurate than a prediction based on results alone.

It's a useful _bypass_. But this is not what's the discussion is about.
The discussion is not about the useability of such a bypass, but about the
question if such a system is _better_ than the current system, which isn't
the case.

Bottom line of the problem (or misunderstanding) is that chess is a
competitive sport. The opponent is not one of the many factors, she or
he's the determining factor. And as long as this is not the basis of the
calculation, every result of another calculation will always be a
substitute for the rating you get based on play against opponents, like in
the Elo-rating system.

--
CeeBee


Uxbridge: "By God, sir, I've lost my leg!"
Wellington: "By God, sir, so you have!"


Google CeeBee @ www.geocities.com/ceebee_2

On 13 Oct 2003 16:32:50 GMT, CeeBee wrote:

Ed Seedhouse wrote in rec.games.chess.misc:

However, if we get a very strong computer to go over a few thousand or
perhaps a few million games of all available players and then compute
and save in a database what we might call an "accuracy index" for each
game and an overall "accuracy index" for each player we will,
presumably, find some correlation between this index and the player's
actual rating.

If so we can then get the computer to rank players according to
overall "accuracy index" and rank all the players accordingly.

This is another way to determine differences in strenght.

Well, I don't know if it is. The only real coherent measurement of
strength I know of is how well you actually perform against actual
players.

Does an "accuracy index" reflect performance? I imagine it does, up
to a point. But the evidence is not in since the exercise has never
(to my knowlege) been done. So I'd prefer not to jump to conclusions
and simply say that the results of such an analysis would be very
interesting at least to me.

Maybe it's a
more accurate way of determining those differences. Although
interesting, it might turn out to be an even more complicated way to
assess strength differences than the current Elo system. And it might
well add little extra knowledge to the meaning of those strenght
differences we already know.

Yes, I agree. That doesn't necessarily mean it isn't worth doing at
some point. Maybe we should wait a few years until the micro programs
are vastly stronger than all human grandmasters and so can be agreed
to provide an objective evaluation of moves.

Nor can one just assume that computing an "accuracy index" will be
simple and straightforward. For one thing, just what we mean by
"accuracy" will have to be defined in numerical terms.

The original poster wanted to use such a databaseto test people against
such a database (even without playing a single game before) to establish
a "more absolute" strength number.

In this he is being very naive, in my opinion. First get the evidence
and *then* draw some conclusions from it, if possible.

In the end one has taken a long road to arrive at the same point the Elo
rating system took in a fraction of the time.

Well possibly, but until we actually do some such project we really
don't know if it will give us useful added information. I suspect it
will, but it won't solve all problems by any means of course.