| |
Main
Date: 18 Nov 2007 23:04:38
From: Sanny
Subject: Nomorechess Suggestions implemented at GetClub.
|
Nomorechess suggested me that GetClub Chess should look for Defence
moves and also look ahead when something is captured or King gets
Check.
His technique was implemented. But it looks like its depth has gone
very down. Earlier it used to go upto 6 level deep now it goes till
only 2-3 level deep.
Play Chess at: http://www.GetClub.com/Chess.html
Lets test with your games and see if this helps the gameplay.
If it makes even weaker moves this modification will be removed.
However, If this improvement gives more wins to GetClub Chess then it
will be kept as it is.
So Help Bot, how are you feeling the gameplay after this improvement.
Are you finding GetClub Chess Stronger or it has become weaker by this
improvement. As depth of search has reduced from 5-6 to 2-3.
Play a few games with Beginner, Easy & Normal Levels and let me know
your opinion. Should I remove this Check & Capture Extension or keep
it. As the depth has gone down from 5-6 to only 2-3 {on Easy Level}
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
|
| |
Date: 24 Nov 2007 02:13:26
From: help bot
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
On Nov 24, 4:34 am, "Wlodzimierz Holsztynski (Wlod)"
<[email protected] > wrote:
> Finally GetClub has given up on chess. This
> decision was long overdue.
Most observers predicted that Sanny's contraption would
never make it this far, but alas, they have now been proved
wrong in terms of playing strength. All that remains is to
make the confounded machine play chess according to
Hoyle: recognition of draws, an end to King captures, and
no more randomly-generated "resignations" of its human
opponents.
One area which is difficult to comprehend is the hit-or-miss
approach to opening play; if a sequence of moves is in the
book, the program now responds instantly, but sometimes
the program is out of its book at move one. Another oddity
is that when the program has but one legal move, it has no
clue and considers its option -- singular -- deeply. As far
as I can determine, the legality of moves takes last -- not
first -- priority; it's like when you're buying a new car, and the
dealer asks if you want to "add on" floor mats.
-- help bot
|
| |
Date: 24 Nov 2007 01:34:24
From: Wlodzimierz Holsztynski (Wlod)
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
"Momorechess Suggestions implemented at GetClub".
Finally GetClub has given up on chess. This
decision was long overdue.
Wlod
|
| |
Date: 22 Nov 2007 00:18:25
From: help bot
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
On Nov 22, 12:48 am, "Ray Gordon, creator of the \"pivot\""
<[email protected] > wrote:
> > takes the fun out of it when you spend all that time and
> > then POOF, are awarded a loss for your efforts. That
> > was against the Master level, I think.
>
> Any player that can resign for its opponent should be a master.
It only "resigns me" on occasion; the majority of the
time it resigns itself, just one or two moves before
checkmate.
Right now I am stuck like a Panzer tank in the mud;
the position is my King vs. GetClub's K & N, and of
course the program does not know this is a draw by
lack of mating material; even so, I valiantly defended
until... the program went brain-dead, refusing to ever
move. I cannot resign; I cannot claim a draw; I cannot
move, because it is not my turn. Time ches on.
The program seems to be getting tougher and
tougher. I had a lock on its Queen side counter-play
but was unable to break through quickly on the other
wing. So I took risks, rather than watch myself die
of old age by creeping forward, ever so carefully --
and the result was a Queen intrusion, then gross
simplification to a legitimate draw. The last time
that happened to me against a human opponent,
it was in a time scramble and I had been losing on
the board.
-- help bot
|
| |
Date: 19 Nov 2007 04:35:05
From: Sanny
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
On Nov 19, 1:32 pm, help bot <[email protected] > wrote:
> On Nov 19, 2:04 am, Sanny <[email protected]> wrote:
>
> > Nomorechess suggested me thatGetClubChessshould look for Defence
> > moves and also look ahead when something is captured or King gets
> > Check.
>
> > His technique was implemented. But it looks like its depth has gone
> > very down. Earlier it used to go upto 6 level deep now it goes till
> > only 2-3 level deep.
>
> That's quite a drop.
>
> > So Help Bot, how are you feeling the gameplay after this improvement.
> > Are you findingGetClubChessStronger or it has become weaker by this
> > improvement. As depth of search has reduced from 5-6 to 2-3.
>
> Well, a couple days ago I was watching "Smallville" on
> DVD when theGetClubprogram resigned for me, where
> I was clearly winning. I was not surprised, but it sort of
> takes the fun out of it when you spend all that time and
> then POOF, are awarded a loss for your efforts. That
> was against the Master level, I think.
>
> More recently, I played one game against Beginner
> level and demolished the program, a miniature I think.
>
> My perspective is that there is a huge difference in
> strength between the Beginner level and the higher
> levels, but I can't say if or when the tactical search
> extensions were implemented.
>
> What I can say is that even if Java applets are very
> slow compared to normalchessprograms, there is no
> good reason the program should be reaching only 2 or
> 3 plies even with all the search extensions in the world.
> Something must be wrong with the implementation of
> this idea:
>
> 1) You must have a limit, a cut-off point so that when
> there are perpetual checks, you don't keep going like
> the Energizer Bunny.
>
> 2) Even programs which normally have no trouble
> searching deeply, sometimes bog down when it comes
> to pawn promotions. Others suffer no significant
> problems, so it is all in how you implement the idea,
> and the same principle may apply to other tricky moves
> like castling and /en passant/ captures.
>
> My commentaries regarding the program's sharp
> attacks yet weak defense were intended to convey
> information which might be used to eradicate
> weaknesses; but it is unwise to play purely
> defensively, for then even a weak human opponent
> may be able to slowly overwhelm the program
> strategically. It is essential that the program get
> into tactical complications often enough to allow
> the computer to do what computers do best:
> calculate more accurately than any human can!
>
> -- help bot
Yes only after 10-20 Games are played by players, I will know that
this improvement is fruitful or not. I tried one game with Jester. The
program was in advantage but later got under Mate Threath. Then I
played a Game myself with Beginner Level It easily beat me. So I think
It is unclear whether the game has improved or detoriated.
Play Chess at: http://www.GetClub.com/Chess.html
It is only after strong players beat it then I will know how it is
performing.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| |
Date: 19 Nov 2007 00:32:33
From: help bot
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
On Nov 19, 2:04 am, Sanny <[email protected] > wrote:
> Nomorechess suggested me that GetClub Chess should look for Defence
> moves and also look ahead when something is captured or King gets
> Check.
>
> His technique was implemented. But it looks like its depth has gone
> very down. Earlier it used to go upto 6 level deep now it goes till
> only 2-3 level deep.
That's quite a drop.
> So Help Bot, how are you feeling the gameplay after this improvement.
> Are you finding GetClub Chess Stronger or it has become weaker by this
> improvement. As depth of search has reduced from 5-6 to 2-3.
Well, a couple days ago I was watching "Smallville" on
DVD when the GetClub program resigned for me, where
I was clearly winning. I was not surprised, but it sort of
takes the fun out of it when you spend all that time and
then POOF, are awarded a loss for your efforts. That
was against the Master level, I think.
More recently, I played one game against Beginner
level and demolished the program, a miniature I think.
My perspective is that there is a huge difference in
strength between the Beginner level and the higher
levels, but I can't say if or when the tactical search
extensions were implemented.
What I can say is that even if Java applets are very
slow compared to normal chess programs, there is no
good reason the program should be reaching only 2 or
3 plies even with all the search extensions in the world.
Something must be wrong with the implementation of
this idea:
1) You must have a limit, a cut-off point so that when
there are perpetual checks, you don't keep going like
the Energizer Bunny.
2) Even programs which normally have no trouble
searching deeply, sometimes bog down when it comes
to pawn promotions. Others suffer no significant
problems, so it is all in how you implement the idea,
and the same principle may apply to other tricky moves
like castling and /en passant/ captures.
My commentaries regarding the program's sharp
attacks yet weak defense were intended to convey
information which might be used to eradicate
weaknesses; but it is unwise to play purely
defensively, for then even a weak human opponent
may be able to slowly overwhelm the program
strategically. It is essential that the program get
into tactical complications often enough to allow
the computer to do what computers do best:
calculate more accurately than any human can!
-- help bot
|
| | |
Date: 22 Nov 2007 00:48:39
From: Ray Gordon, creator of the \pivot\
Subject: Re: Nomorechess Suggestions implemented at GetClub.
|
> takes the fun out of it when you spend all that time and
> then POOF, are awarded a loss for your efforts. That
> was against the Master level, I think.
Any player that can resign for its opponent should be a master.
--
Ray Gordon, The ORIGINAL Lifestyle Seduction Guru
http://www.cybersheet.com/library.html
Includes 29 Reasons Not To Be A Nice Guy
Ray's new "Project 5000" is here:
http://groups.yahoo.com/group/project-5000
Don't rely on overexposed, mass-keted commercial seduction methods which
no longer work.
Thinking of taking a seduction "workshiop?" Read THIS:
http://www.dirtyscottsdale.com/?p=1187
Beware! VH-1's "The Pickup Artst" was FRAUDULENT. Six of the eight
contestants were actors, and they used PAID TARGETS in the club. The paid
targets got mad when VH-1 said "there are no actors in this club" and ruined
their prromised acting credit. What else has Mystery lied about?
|
|
