On Jul 6, 11:39 am, Sanny wrote:

Help Bot complained me that lower levels are getting higher rating So
he do not play with Higher Levels.

In reality, I was merely explaining why I preferred to
play the lower levels, and not the higher ones Sanny
kept requesting I play.


For real Comparision.

I played 3 games against IVAN rated 1800+ and GetClub easily defeated
IVAN So I guessed the correct rating.

Easily? What about the first game-- the one
played out here in rgcm, which did not go like
a strawberry smoothie, but more like broken
glass.


Zebediah is using 2900+ rated program against higher levels and
winning all the games So it is very difficult to assess correct rating
based on that.

I don't know which program Zeb is using, but when
I looked at his moves using Rybka, she was not very
impressed. As always, numbers are just being pulled
out of a hat.


So people use brains only when playing with Beginner & Easy Levels.
While those playing with Normal/ Master take help of other Commercial
programs rated 3000+.

Perhaps they believed the "stories" they saw
posted here in rgc, and thought it was hopeless?


I would have agreed But GetClub Easy Level has beaten IVAN with 1800+
rating 3 times. So Beginner Level is atleast 2100+ as the game has
been improved 5 times by then.

I am patiently waiting for all these "improvements"
to get the program up to my own level-- roughly
47,000,000 USCF.


-- help bot

In one message Sanny writes:

"3 months back GetClub was twice improved.
2 months back GetClub was Twice improved.
1 monrh back GetClub was twice improved
2 weeks back GetClub was Twice improved
3 days back GetClub was twice improved.
So GetClub was improve 2*2*2*2*2 = 32 times. "

And in the very next message, Sanny writes:

"There was a bug that was removed.
That bug was causing it to make inferior moves.
This update will slow the GetClub but make good moves."

32 times improved, but making inferior moves due to a bug. Of course
now it will make good moves but be slower.

Have we heard this before? At least 32 times?

On Jul 7, 2:38 pm, "Alessandro J." wrote:

I tried to finish the game, but GC refuses to concede regardless of
the fact that I've brough to an endgame

Hmm-- so you couldn't beat GC in the
opening or in the middle-game, eh?


R+Ps v. Ps with an
unassailable passer. Suddenly, after having logged in for the 5th
time, my PC screen went completely black.

That is just the latest in a long line of
"improvements".


Ratings are for people and programs that don't cheat. GetClub is rated
zero.

That's a pity, with my 2/2 my " Sanny rating " would be around 2350.

Your second games was banished to
Oblivion-- it doesn't count for rating
purposes. A few quick calculations...
your provisional rating at GC is zero,
based on one win against an opponent
rated zero. (Note that even if we were
to "give you" the other game, it would
still come out to zero.)


Hey, isn't that nearly ....

A record? No. Sanny himself holds
the all-time-low record, with a score of
0-for-238, as of Dec. 12, 2007, which
beats your zero-rating on tie-breaks.
You came close, mister, but no cigar.


-- help bot

And in the very next message, Sanny writes:

"There was a bug that was removed.
That bug was causing it to make inferior moves.
This update will slow theGetClubbut make good moves."

32 times improved, but making inferior moves due to a bug. *Of course
now it will make good moves but be slower.

Have we heard this before? *At least 32 times?

Bug comes without giving invitation. Whats the use of depth of search
in increased while that makes it dumb at strategy.

However whenever we find a bug we remove that immidietely even if that
reduce the performance. As correct move is our first priority.

Bye
Sanny

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

I tried to finish the game, but GC refuses to concede regardless of
the fact that I've brough to an endgame R+Ps v. Ps with an
unassailable passer. Suddenly, after having logged in for the 5th
time, my PC screen went completely black.
I think that's enough for me.

Try again and Login & Play and tell me exactly what happen when you
login?

I think the game is over and it declared Game Over.

Bye
Sanny

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

Sanny wrote:
And in the very next message, Sanny writes:

"There was a bug that was removed.
That bug was causing it to make inferior moves.
This update will slow theGetClubbut make good moves."

32 times improved, but making inferior moves due to a bug. Of course
now it will make good moves but be slower.

Have we heard this before? At least 32 times?

Bug comes without giving invitation. Whats the use of depth of search
in increased while that makes it dumb at strategy.

However whenever we find a bug we remove that immidietely even if that
reduce the performance. As correct move is our first priority.


You remove bugs immediately?

OK - here's a bug report: you program does not observe the time limit.

Please advise when this is fixed.

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

On Tue, 8 Jul 2008 08:11:15 -0700 (PDT), Sanny
wrote:

And in the very next message, Sanny writes:

"There was a bug that was removed.
That bug was causing it to make inferior moves.
This update will slow theGetClubbut make good moves."

32 times improved, but making inferior moves due to a bug. *Of course
now it will make good moves but be slower.

Have we heard this before? *At least 32 times?

Bug comes without giving invitation. Whats the use of depth of search
in increased while that makes it dumb at strategy.

Simple little test which will help you out there.

If Rybka has a position scoring mode, use it. If not, any old chess
program will do. Play a game against it. Analyze the game, and
determine how far program X determines you are ahead.
I know the earlier versions of Chessmaster allowed you to adjust the
weights of pieces (which you could make it prefer knights over
bishops, or how many pieces it would trade for a queen).

Put the same position into GC, and see how close it comes to that.

Best place to start actually is the endgame (P+K v. K is probably the
simplest, the king opposing the pawn needs to remain within as many
squares as the pawn has to go to get promoted). Fewer pieces, and it's
very tactical play (you actually should be able to cheat there,
increasing ply because you don't have as many pieces). Work your way
backwards to the middle game. This also works pretty well for
verifying the correct moves of pieces.
A program often written by students is the knight's tour. You might
want to find out what that is. That also helps pave the way for game
mechanics.
You might also want to check out some chess books. They have
diagrams, and are a faster source of information than the internet.
Don't tell me you can't afford them, or don't have time. If you're
trying to make a business out of this, you'll invest some time and
money into learning exactly what in the hell you're trying to do.

You should have some scoring system, consisting of at least
- Value of pieces (e.g. pawn 1, queen 9, king MAX)
And also probably
- Increase the value of a trade when you are ahead on material.
Trading material when ahead increases how far you are ahead
proportionally (e.g. if the score is 100-90, taking 10 points from
both to 90-80 takes the advantage from .11 to .125).
- Value of pieces en prise, and a penalty if you take a piece out of
en prise.
- Value of squares controlled. You might vary the value based on how
many times a square is controlled (read Pandolfini to learn more, if
black attacks a sqaure 4 times, and white only 3, black controls the
square), and where the square is located (center game values center
squares more). A controlled square ISN'T one that has a piece on it.
You generally can tell programs that value this because they love
closed games.
- Value of a square with a piece on it that is protected by other
pieces.
More nuanced checks (which might be fractions of a pawn):
- Value of connected versus unconnected pawns. Slight penalty for
doubled pawns.
- Value of sqaures controlled when you only have one bishop, and the
square is the same color as the bishop.

If you can't do this because the program is written in a fashion where
you can't do this, you're an arrogant idiot (and help bot calls me
arrogant... I'm not the one thinking I could write a chess program
correct ON THE FIRST TRY. Maybe he thinks he can).

It's called UNIT TESTING. The more programs you use to evaluate the
position, the better GC should be. That's called REVERSE ENGINEERING
(black-box because you're not looking at the source).

And programmers make bugs. Don't tell me otherwise. They generally
appear because somebody didn't think or understand. The better you
know what you're doing, the fewer bugs you will have.


However whenever we find a bug we remove that immidietely even if that
reduce the performance. As correct move is our first priority.

Bull****. If a correct move was truly your first priority, it would
have made correct moves before we ever saw it. One of the biggest
excuses programmers and managers make is there wasn't enough time to
test.

The other thing that is downright annoying, is you don't really go
into any detail of the bugs.

Also, your 'human' play may hide it's tactical blind spots at the
moment, but if someone plays a few games, it will become pretty
obvious (c.f. security by obscurity).
And, you should be thankful people are playing your program, and not
be upset that people are using chess programs to beat it. Even people
on ICC complain about that.

Finally, a lot of bugs are linear. It's like a water pipe with a
bunch of leaks in it. Fix one, and the next one down the line becomes
obvious. Fixing the one doesn't make the program any better.
In reality the bug shows that the programs wasn't half as much as it
was claimed to be. Better get proactive, or you will be able to see
how many bugs people tolerate before they give up (one catastrophic
bug every 2 hours... probably every other game of chess).
Don't think for a moment you're really convincing anyone that bugs
come out of the blue. People don't plan to fail, they fail to plan.

But hey, help bot says I don't know what I'm talking about...

If your car breaks down, to you go to a strangers house to ask them to
fix it?


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

On Jul 7, 5:50 pm, " wrote:

In one message Sanny writes:

"3 months back GetClub was twice improved.
2 months back GetClub was Twice improved.
1 monrh back GetClub was twice improved
2 weeks back GetClub was Twice improved
3 days back GetClub was twice improved.
So GetClub was improve 2*2*2*2*2 = 32 times. "

And in the very next message, Sanny writes:

"There was a bug that was removed.
That bug was causing it to make inferior moves.
This update will slow the GetClub but make good moves."

32 times improved, but making inferior moves due to a bug. Of course
now it will make good moves but be slower.

Have we heard this before? At least 32 times?

The thinking time has been changed so that
the program will think longer on difficult moves,
and shorter on easier ones. That's the idea,
anyway. It's /feels/ a bit more like playing a
human opponent.

In one of my most recent games at GetClub,
I had K&R against K, and the program saw
checkmate coming several moves in advance
on Easy level, which is better than in the old
days. On the other hand, in that game I was
disgusted at how many times I made crude
tactical blunders, yet I still managed to win,
by a hair:

http://www.getclub.com/playgame.php?...991&game=Chess

The Normal level now seems to take quite
some time, as many, if not most of the
positions are judged to be worthy of long
thinks. On the other hand, my computer's
cooling fan no longer goes crazy, suggesting
a possible "improvement" of some sort which
I cannot fathom.

The ratings on Sanny's Web site would
themselves tell us something about these
many "improvements", except that some
folks seem to feel they can't compete
without the aid of their own chess engines.
Zebediah has manipulated the ratings, and
even a cursory glance at his games reveals
an engine versus engine battle, where poor
Sanny is simply outclassed.

I still believe that a stupid but tactically-
correct engine would do better than this
one, because correctly calculating tactics
is (or should be) the computer's strength
and our weakness.


-- help bot

Hey, isn't that nearly ....

* A record? *No. *Sanny himself holds
the all-time-low record, with a score of
0-for-238, as of Dec. 12, 2007, which
beats your zero-rating on tie-breaks.
You came close, mister, but no cigar.

Today again game was improved. So game will play Twice better moves.

Now, Help Bot has 2 choices either improve hig game or use some
computers help to win?

Help Bot, now easy will play very strong moves so be careful while you
play against easy. It is no more the carrot you used to eat.

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


Yesterday 10, games were played at GetClub I was surprised to see all
included chrisf was playing with Beginner Level. No one trying Easy
Level as Beginner was good tough for them.

What do Help Bot say, Can he still compete the Easy Level?

Bye
Sanny

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

But hey, help bot says I don't know what I'm talking about...

If your car breaks down, to you go to a strangers house to ask them to
fix it?

Thanks, for your good advices.

Bye
Sanny

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