GetClub.com rankings

Main

Date: 22 Sep 2007 20:43:46
From: help bot
Subject: GetClub.com rankings

It seems to me that Zebediah has established an 800 point
advantage over the "Master" level, having systematically
pushed it downward as his own rating s-l-o-w-l-y rose into
the 1300s.

Let's assume, for this purpose, that old Zeb is a USCF
2200: that would put the Master level at around 1400, if we
could just map one ratings system to the other. But I don't
think that is accurate; IMO, the spread is probably much
lower, and GetClub's Master level might well be far above
1400, in spite of its tendency to occasionally allow pawn
forks or toss pieces away for no apparent reason.

In one recent game, I decided to try something like:
1. f3, 2. Kf2, and moving quickly, I missed a simple
fork where I played p-d4? and Black could have replied
...exd4, exd4 Qh4+, winning the pawn for free -- but it
missed it. It has also consistently missed tricks and
traps which might have netted it material or saved it from
losing material, and just one example was a recently
posted game where it almost got back-rank mated, but
was able to interpose -- and lose -- its Queen instead.

In sum, it has serious tactical problems if the game
goes for any length of time, but on occasion appears to
be playing reasonably well. I have had games where
the program's play is suspiciously akin to that of a
human player, but I have also had endings where the
program moves its King about at random, unlike any
human I have ever seen. In fact, while others may pan
the program's openings bungling, I see the endgame as
particularly weak, on account of the fact that programs
ought to see much deeper here, yet Sanny's doesn't.

There ought to be a battery of tests to determine the
equivalent rating of a given program, but it is not quite
fair to say this one is, say, 800 because it fails to see
a mate-in-two on itself. When I am playing the lower
levels, I am anticipating a horrible piece-hang; but when
I play the s-l-o-w levels, I feel considerable resistance
(except for when I don't, *wink*).

The last time I recall playing a USCF 1400, he was
obviously (a weakness in itself) planning to castle on
the Queen side, and I made a move which prevented
this because it tied his King to the defense of his
f-pawn. He castled Queen side anyway and I chopped
the pawn, forking his Rooks. He then got angry at
himself, and played the rest of the game very poorly.

Sanny's program doesn't do that; when it hangs
material, it goes "ha! you lucky dog" and then
proceeds to play just the same as before. That's
why I can't see it as only 1400 strength, except for
the lowest levels, perhaps. In fact, the ratings as
they exist now are ass-backwards: Advance and
Master are the lowest around, while the weakest
of them all somehow have the highest numbers.

-- help bot

Date: 25 Sep 2007 01:00:35
From: Sanny
Subject: Re: GetClub.com rankings

On Sep 25, 11:45 am, help bot <nomorech...@hotmail.com > wrote:
> On Sep 25, 12:42 am, Sanny <softta...@hotmail.com> wrote:
>
> > > But the most glaring problem is the one I already noted:
> > > the strongest levels are by far the lowest-rated; and the
> > > strength of thechessprogram has varied wildly over time,
> > > while being rated as a single, fixed entity. In sum, these
> > > ratings mean very little. I still believe the program was
> > > much, much weaker when I first started playing there,
> > > which was back when the Master level had a rating of
> > > somewhere near 1000 -- nearly twice its current rating.
>
> > To overcome it, If You Win a game with 500 Rated and you are 1000
> > Rated you only get +2 While if you loose The 500 Rated gets +20
> > Rating. So once more people play with higher levels the ratings will
> > be corrected.
>
> What is really needed is a better balance between the
> wins and losses; in other words, getting more weak
> players and more average players to play atGetClub,
> so things are not so lopsided.
>
> Sanny keeps asking for ideas on how to improve the
> program; well, how about making the program play its
> move fairly quickly when there is just one legal move?
>
> I call that an improvement because the user, which
> here is the human opponent, could become impatient
> if he realizes there is only one legal move, yet the
> program may take a very long time to play it so the
> game can continue. Tonight I am in an ending where
> I have a Queen and two Bishops versus a Queen and
> a Rook, and I played a move which offered up one of
> the Bishops for the Rook; the program took my Bishop,
> I immediately took his Rook, giving check, and there
> was but one legal response: K-h2, yet the program
> did its usual long think.
>
> Now, let's suppose that the program starts off by
> generating a list of every conceivable move of every
> man on the board (for the side to move), and then
> passes this list to another subroutine which checks
> each move on the list for legality. We are still on
> ply #1 here, and it should come out with a "list"
> with just a single entry: K-h2. Noting that there is
> only one possibility, there is no need to waste time
> with deep calculations: just play the ("best") move,
> K-h2. Whatever the response, the program will be
> able to see deeper *after* the opponent responds to
> this forced move, so why not just play it?
>
> -- help bot

Now on Beginner Level will always make move in 20 Seconds
Easy level in 80 seconds
Normal in 320 seconds
Master in 1280 seconds.

Even if it has not found any good move it will play the move as soon
as time has arrived.

I hope this will make everyone happy. As Program will make its move as
soon as Time Arrives.

Play a few game and see it is always playing in time. So no more waits
and time Proiblem.

Bye
Sanny

Play Chess at: http://www.GetClub.com/Chess.html

Date: 24 Sep 2007 23:45:19
From: help bot
Subject: Re: GetClub.com rankings

On Sep 25, 12:42 am, Sanny <softta...@hotmail.com > wrote:

> > But the most glaring problem is the one I already noted:
> > the strongest levels are by far the lowest-rated; and the
> > strength of thechessprogram has varied wildly over time,
> > while being rated as a single, fixed entity. In sum, these
> > ratings mean very little. I still believe the program was
> > much, much weaker when I first started playing there,
> > which was back when the Master level had a rating of
> > somewhere near 1000 -- nearly twice its current rating.
>
> To overcome it, If You Win a game with 500 Rated and you are 1000
> Rated you only get +2 While if you loose The 500 Rated gets +20
> Rating. So once more people play with higher levels the ratings will
> be corrected.

What is really needed is a better balance between the
wins and losses; in other words, getting more weak
players and more average players to play at GetClub,
so things are not so lopsided.

Sanny keeps asking for ideas on how to improve the
program; well, how about making the program play its
move fairly quickly when there is just one legal move?

I call that an improvement because the user, which
here is the human opponent, could become impatient
if he realizes there is only one legal move, yet the
program may take a very long time to play it so the
game can continue. Tonight I am in an ending where
I have a Queen and two Bishops versus a Queen and
a Rook, and I played a move which offered up one of
the Bishops for the Rook; the program took my Bishop,
I immediately took his Rook, giving check, and there
was but one legal response: K-h2, yet the program
did its usual long think.

Now, let's suppose that the program starts off by
generating a list of every conceivable move of every
man on the board (for the side to move), and then
passes this list to another subroutine which checks
each move on the list for legality. We are still on
ply #1 here, and it should come out with a "list"
with just a single entry: K-h2. Noting that there is
only one possibility, there is no need to waste time
with deep calculations: just play the ("best") move,
K-h2. Whatever the response, the program will be
able to see deeper *after* the opponent responds to
this forced move, so why not just play it?

-- help bot

Date: 24 Sep 2007 22:42:45
From: Sanny
Subject: Re: GetClub.com rankings

> But the most glaring problem is the one I already noted:
> the strongest levels are by far the lowest-rated; and the
> strength of thechessprogram has varied wildly over time,
> while being rated as a single, fixed entity. In sum, these
> ratings mean very little. I still believe the program was
> much, much weaker when I first started playing there,
> which was back when the Master level had a rating of
> somewhere near 1000 -- nearly twice its current rating.

To overcome it, If You Win a game with 500 Rated and you are 1000
Rated you only get +2 While if you loose The 500 Rated gets +20
Rating. So once more people play with higher levels the ratings will
be corrected.

Bye
Sanny

Play Chess at: http://www.getclub.com/Chess.html

Date: 24 Sep 2007 18:46:31
From: help bot
Subject: Re: GetClub.com rankings

On Sep 24, 9:32 am, David Richerby <dav...@chiark.greenend.org.uk >
wrote:

> Sanny isn't using an Elo system so you just can't do that. What you
> can do, though, is observe that Zebediah has played 135 games, with
> 133 wins and two losses (98.5%). Interpolating FIDE's chart[1], this
> corresponds to a rating difference of about 630 points.

Interesting. However, you have missed a subtle point,
which is that in 135 games there were no draws; my best
guess is that, just like me, Zebediah has in fact drawn
but because GetClub has no clue how to handle draws,
those games were relegated to the dust bin, disappearing
without a trace. In this example, the existence of such
draws would tend to drag Zebediah's performance rating
sharply downward, narrowing your calculated gap of 630
points a bit.

By the same token, it is possible that any given loss
could be the result of some freak computer error; take a
look at the many games which cannot be replayed at
GetClub: clearly, an illegal move or other irregularity has
created a glitch which his Web site deals with by locking
up like a turnip. I know that playing as Nomorechess, I
had several draws which were tossed out, a multitude of
easy wins which were "reset" before I could finish them,
and some games where the program simply became
confused as to which side was to move.

But the most glaring problem is the one I already noted:
the strongest levels are by far the lowest-rated; and the
strength of the chess program has varied wildly over time,
while being rated as a single, fixed entity. In sum, these
ratings mean very little. I still believe the program was
much, much weaker when I first started playing there,
which was back when the Master level had a rating of
somewhere near 1000 -- nearly twice its current rating.

-- help bot

Date: 24 Sep 2007 15:32:55
From: David Richerby
Subject: Re: GetClub.com rankings

help bot <nomorechess@hotmail.com > wrote:
> It seems to me that Zebediah has established an 800 point advantage
> over the "Master" level, having systematically pushed it downward as
> his own rating s-l-o-w-l-y rose into the 1300s.
>
> Let's assume, for this purpose, that old Zeb is a USCF 2200: that
> would put the Master level at around 1400, if we could just map one
> ratings system to the other.

Sanny isn't using an Elo system so you just can't do that. What you
can do, though, is observe that Zebediah has played 135 games, with
133 wins and two losses (98.5%). Interpolating FIDE's chart[1], this
corresponds to a rating difference of about 630 points.

Dave.

PS. Death to all wasps.

[1] http://www.fide.com/official/handbook.asp?level=B0210

--
David Richerby Sadistic Swiss Hi-Fi (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a music system but it's made in
Switzerland and it wants to hurt you!

Date: 23 Sep 2007 01:09:11
From: help bot
Subject: Re: GetClub.com rankings

On Sep 23, 2:03 am, Sanny <softta...@hotmail.com > wrote:

> The Ratings are untouched let them be where the winds take them. I
> think The reasion High level has low rating is that High level is
> being played by 2200+ players And they beat 2000 Master Level every
> time.

Good point. I imagine that weaker players would prefer
to play the quick-moving levels, and so the program gets
to win occasionally.

> While Beginner 1700 is played by 1600 Rated Players and they get lost
> a couple of time. But I see you are playing a lot with Beginner & Easy
> level So they will also get low rating. Players with your level should
> play with Normal & Master Level.

Yes, well that would mean letting my notebook run
hot all night to *maybe* finish a single game, if I'm
lucky. Every disconnect means loss of perhaps an
hour of thinking time, since the program's just-played
move is not registered, and must be recalculated.
BTW, this is obviously a bug, since I tend to get
disconnected just as the program makes its move.
While some other sites may be very slow on my
connection, none of them "disconnects" and makes
me log back in, reload the Java applet, and so forth.

Another point is that there is always the possibility
that any given game could go, say, 100 moves; do
you have any idea how old I would be by the time
such a game would finally finish? (Now that I think
about it, there could be much faster computers by
that time and things might go quicker, since we
would all have Apple Computers running at around
a gazillion megahertz... .)

> As beginner & Easy are for 1600 and below rated players.

The truth is, Sanny, that most people don't want to
take all night or all day (or both, for the Advance level)
to play a single game against a computer.

You seem to have made a lot of progress in the
openings; my game where I played 1.h3 had the
program setting up quite a nice pawn center for
me to try and undermine; I won a piece, as usual,
but it was a tough defense since I was behind in
development from attacking and taking the Knight,
plus it got a center pawn in the deal and I could
not castle right away.

One thing you might want to do when the
complaints about time controls die down is look
at testing the Web site for bugs; I seem to recall
that someone posted the address for a free test
site here a long while back. Hey, if it's free, what
have you got to lose?

-- help bot

Date: 23 Sep 2007 00:03:02
From: Sanny
Subject: Re: GetClub.com rankings

On Sep 23, 8:43 am, help bot <nomorech...@hotmail.com > wrote:
> It seems to me that Zebediah has established an 800 point
> advantage over the "Master" level, having systematically
> pushed it downward as his own rating s-l-o-w-l-y rose into
> the 1300s.
>
> Let's assume, for this purpose, that old Zeb is a USCF
> 2200: that would put the Master level at around 1400, if we
> could just map one ratings system to the other. But I don't
> think that is accurate; IMO, the spread is probably much
> lower, and GetClub's Master level might well be far above
> 1400, in spite of its tendency to occasionally allow pawn
> forks or toss pieces away for no apparent reason.
>
> In one recent game, I decided to try something like:
> 1. f3, 2. Kf2, and moving quickly, I missed a simple
> fork where I played p-d4? and Black could have replied
> ...exd4, exd4 Qh4+, winning the pawn for free -- but it
> missed it. It has also consistently missed tricks and
> traps which might have netted it material or saved it from
> losing material, and just one example was a recently
> posted game where it almost got back-rank mated, but
> was able to interpose -- and lose -- its Queen instead.
>
> In sum, it has serious tactical problems if the game
> goes for any length of time, but on occasion appears to
> be playing reasonably well. I have had games where
> the program's play is suspiciously akin to that of a
> human player, but I have also had endings where the
> program moves its King about at random, unlike any
> human I have ever seen. In fact, while others may pan
> the program's openings bungling, I see the endgame as
> particularly weak, on account of the fact that programs
> ought to see much deeper here, yet Sanny's doesn't.
>
> There ought to be a battery of tests to determine the
> equivalent rating of a given program, but it is not quite
> fair to say this one is, say, 800 because it fails to see
> a mate-in-two on itself. When I am playing the lower
> levels, I am anticipating a horrible piece-hang; but when
> I play the s-l-o-w levels, I feel considerable resistance
> (except for when I don't, *wink*).
>
> The last time I recall playing a USCF 1400, he was
> obviously (a weakness in itself) planning to castle on
> the Queen side, and I made a move which prevented
> this because it tied his King to the defense of his
> f-pawn. He castled Queen side anyway and I chopped
> the pawn, forking his Rooks. He then got angry at
> himself, and played the rest of the game very poorly.
>
> Sanny's program doesn't do that; when it hangs
> material, it goes "ha! you lucky dog" and then
> proceeds to play just the same as before. That's
> why I can't see it as only 1400 strength, except for
> the lowest levels, perhaps. In fact, the ratings as
> they exist now are ass-backwards: Advance and
> Master are the lowest around, while the weakest
> of them all somehow have the highest numbers.
>
> -- help bot

The Ratings are untouched let them be where the winds take them. I
think The reasion High level has low rating is that High level is
being played by 2200+ players And they beat 2000 Master Level every
time.

While Beginner 1700 is played by 1600 Rated Players and they get lost
a couple of time. But I see you are playing a lot with Beginner & Easy
level So they will also get low rating. Players with your level should
play with Normal & Master Level.

As beginner & Easy are for 1600 and below rated players.

Bye
Sanny

Play Chess at: http://www.GetClub.com/Chess.html

Date: 23 Sep 2007 12:11:58
From: Kenneth Sloan
Subject: Re: GetClub.com rankings

Sanny wrote:
>
> The Ratings are untouched let them be where the winds take them. I
> think The reasion High level has low rating is that High level is
> being played by 2200+ players And they beat 2000 Master Level every
> time.
>

So, it's not just your "chess" program that's broken - your "rating"
system is also broken.

--
Kenneth Sloan KennethRSloan@gmail.com
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

Date: 23 Sep 2007 19:13:58
From: David Richerby
Subject: Re: GetClub.com rankings

Kenneth Sloan <KennethRSloan@gmail.com > wrote:
> Sanny wrote:
>> The Ratings are untouched let them be where the winds take them. I
>> think The reasion High level has low rating is that High level is
>> being played by 2200+ players And they beat 2000 Master Level every
>> time.
>
> So, it's not just your "chess" program that's broken - your "rating"
> system is also broken.

I think you missed the inverted commas around `program', `system' and
possibly other words.

Dave.

--
David Richerby Natural Dictator (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a totalitarian leader but it's
completely natural!

Date: 23 Sep 2007 13:26:33
From: Kenneth Sloan
Subject: Re: GetClub.com rankings

David Richerby wrote:
> Kenneth Sloan <KennethRSloan@gmail.com> wrote:
>> Sanny wrote:
>>> The Ratings are untouched let them be where the winds take them. I
>>> think The reasion High level has low rating is that High level is
>>> being played by 2200+ players And they beat 2000 Master Level every
>>> time.
>> So, it's not just your "chess" program that's broken - your "rating"
>> system is also broken.
>
> I think you missed the inverted commas around `program', `system' and
> possibly other words.
>
>
> Dave.
>
I don't "think" so.

--
Kenneth Sloan KennethRSloan@gmail.com
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

Date: 23 Sep 2007 20:27:15
From: David Richerby
Subject: Re: GetClub.com rankings

Kenneth Sloan <KennethRSloan@gmail.com > wrote:
> David Richerby wrote:
>> Kenneth Sloan <KennethRSloan@gmail.com> wrote:
>>> So, it's not just your "chess" program that's broken - your "rating"
>>> system is also broken.
>> I think you missed the inverted commas around `program', `system' and
>> possibly other words.
> I don't "think" so.

O`K'.

Dave`.'

--
David Richerby Broken Incredible Sword (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a razor-sharp blade but it'll
blow your mind and it doesn't work!