On Sep 17, 3:10 am, Sanny wrote:

1. Nf3 d5
2. d4 Nf6
3. Bg5 Ne4

This is the Trompowski, and it was only when GetClub
played the horrible 4. Qd3 that it got into trouble.

So the Master was playing correct Opening moves for itself. Thats
good.

Even so, the sac' on f7 was completely un-called for;
the Knight should have just retreated. Something is
horribly wrong if the program sac's a piece for spite
checks, when there is no substantive attack. Hint:
you cannot conduct an attack by sacrificing the only
piece you have developed, because after the sacrifice,
that piece can attack no more. Duh!

Here Kinight had no place to move. If it comes back Knight & Queen get
into fork by pawn.

Stop! You have made a simple tactical error in your
analysis. First, although 7. Nh3 is much better for
Black, this is still better than hanging a piece. But the
real flaw in your thinking is much too shallow calculation:

1. Nf3 d5
2. d4 Nf6
3. Bg5 Ne4
4. Qd3 Nxg5
5. Nxg5 e5
6. h4 h6
7. Nf3 e4 (forking two pieces)
8. Qe3!

Both White pieces escape intact. This is a perfect
example of why I keep telling you to focus on tactics.

Do you recall the odds matches where Rybka gave
up a random pawn for nothing against grandmasters
and still managed to win? Tactics; they couldn't
quite handle her tactical skill, even in clearly better
positions where the possibility of counterplay was
severely limited due to the gambit pawn.


bIf it goes at h3, Then pawn at h4 can be captured
by queen. So always there was a loss of a Pawn.

Let's look at this for a moment:

6. h4 h6
7. Nh3 Qxh4
8. dxe5

Now, the tactics are far from over as Black can try
....Qb4+ or ...Nc6, but White is not down anything and
he may be able to castle Queenside very soon, with
counterplay.


By taking Pawn at f7, It was 2 points loss. But It took 1 pawn
advantage for King Moved and King cannot do chastling. And the King
defence was broken.

I see; you give a full pawn for loss of castling rights.
But how do you explain the program *not* taking on e5
after ...Nxf7, Kxf7? That counts as a point, no?


Since for all moves there was a loss of 1 pawn. It descided to
sacrifice knight for a pawn + weakening King ~ 1+1=2 Points having a
loss of 1 point.

So trhe move 4. Q-d3 was wrong. I think Master Level did 8 (half
moves) calculations and could not see the Fork, Incase it has seen
till 9 (half moves) It would not have played Q-d3.

Okay, what we have here is a situation where your chess
program is, perhaps, the only one in existence which does
not use tactical search extensions. I wish I could explain
precisely how they can be implemented, but that is beyond
my abilities. In essence, when in a "quiet position", which
is one where there are no legal checks or captures, you
evaluate positional factors; but when there are checks and/or
captures, you *must* look deeper until you reach a quiet
position, and pass back the position score along the line
to the first move.

If best play for both sides in one line ends in checkmate,
the score for the first move in that line is +1 bazillion points.
If best play leads to the forced win of a pawn, the score for
the first move in that line is +1 point.

As far as I know, the first commercial chess computer,
a portable tabletop model, which made this leap to light
speed was probably the Fidelity Chess Challenger series.
It had by-rote opening moves, tactical search extensions,
and selectable levels from very fast to exceedingly slow,
and its primary weakness was that in simple endgames
the play sped up dramatically because of a fixed depth of
search (a poor idea). That was perhaps in the late 1970s;
you are in this sense, a quarter of a century behind the
curve.


-- help bot