Main
Date: 22 May 2008 23:07:42
From: Sanny
Subject: Help Bot struggled to beat the Advance Level.
Help Bot played against Advance Level and to my surprise win the game.

I find Help Bot created a wall of pawns So that Advance Level which
thinks 18 depth deep could not peneterate the wall. And in the end
Help Bot was able to get one good Combination and win the Game.

Looks like GetClub is still a baby for Help Bot. He played moves in
just 2 min / move while GetClub was thinking 30 min / move. Still Help
Bot was able to find a victory.

(See the Recorded game at Bottom)
----------------------------------------------------

The game was even (equal) till 60 Moves. But after the Queens were
exchanged Help Bot managed to fork Two pawns with rooks and kill one
of the pawn and then in a few moves He got the Queen.

So help Bot win the game in just 68 moves.

There were no Tactical Mistake in this game as Both had equal pieces
till end. But I think there were strategical mistakes by GetClub.

But since this was Advance Level which thinks 18 moves ahead it is
very difficult to win by taking its pieces. So Help Bot was able to
use good strategy in beating the Advance Level.

Game Played between help bot and advance at GetClub.com

--------------------------------------------------------------------------------
help bot: (White)
advance: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM20209&game=Chess
--------------------------------------------------------------------------------

White -- Black
(help bot) -- (advance)

1. c2-c3{26} e7-e5{0}
2. g2-g3{16} Ng8-f6{1548}
3. Bf1-g2{18} d7-d5{1792}
4. d2-d4{34} e5-e4{1724}
5. f2-f3{12} Bc8-f5{2352}
6. Ng1-h3{90} e4-f3{2180}
7. e2-f3{16} Qd8-e7{2268}
8. Ke1-f2{46} Qe7-d7{2420}
9. Nh3-f4{74} Rh8-g8{1344}
10. Rh1-e1{60} Bf8-e7{3106}
11. Nf4-d3{52} Nb8-c6{1464}
12. Nd3-c5{84} Qd7-c8{2474}
13. Bg2-f1{98} Ke8-f8{1502}
14. Nc5-d3{40} Nf6-d7{2138}
15. b2-b3{372} g7-g5{2812}
16. Nb1-a3{184} h7-h5{1852}
17. Na3-c2{92} h5-h4{1862}
18. g3-g4{18} Bf5-e6{1520}
19. Kf2-g2{94} Ra8-b8{2722}
20. Nc2-e3{136} Be7-d6{3254}
21. Kg2-h1{72} Nd7-b6{3970}
22. a2-a4{162} Rg8-g6{1782}
23. Bc1-a3{78} Bd6-a3{1470}
24. Ra1-a3{10} Rg6-f6{1736}
25. Ra3-a2{92} Kf8-g8{2204}
26. Ra2-f2{136} Qc8-f8{1642}
27. f3-f4{64} Qf8-d8{1576}
28. f4-f5{268} Be6-d7{1694}
29. Bf1-g2{112} Rf6-d6{1690}
30. Kh1-g1{120} Qd8-f6{1872}
31. Nd3-c5{44} Qf6-h6{2288}
32. h2-h3{98} Qh6-f6{1536}
33. Rf2-e2{98} a7-a5{1452}
34. Ne3-f1{110} Kg8-h7{1806}
35. Nf1-h2{138} Qf6-d8{2114}
36. Nh2-f3{96} f7-f6{1386}
37. Re2-e3{152} Kh7-h8{3110}
38. Bg2-f1{54} Nb6-a8{2004}
39. Bf1-b5{88} b7-b6{1732}
40. Nc5-d7{38} Qd8-d7{2220}
41. Qd1-e2{138} Rb8-g8{2976}
42. Kg1-f2{102} Qd7-f7{1810}
43. Re3-e6{290} Nc6-a7{1328}
44. Re6-e7{430} Qf7-f8{3652}
45. Re7-e8{74} Qf8-g7{1900}
46. Bb5-d3{132} Rd6-d8{2334}
47. Re8-e6{124} Rd8-d7{1486}
48. Kf2-f1{210} c7-c6{1816}
49. Qe2-f2{122} Qg7-f8{1688}
50. Kf1-g2{260} Rd7-h7{2590}
51. Re1-e2{112} Na8-c7{1938}
52. Re6-e3{2218} c6-c5{2018}
53. Qf2-e1{134} Na7-c6{1702}
54. Bd3-b5{726} c5-d4{2534}
55. Nf3-d4{254} Nc6-d4{1588}
56. c3-d4{32} Qf8-d6{2438}
57. Re3-c3{1130} Nc7-b5{2912}
58. a4-b5{18} Qd6-f4{1476}
59. Qe1-f2{50} Qf4-f2{2484}
60. Kg2-f2{14} Rh7-b7{1764}
61. Re2-e6{66} Rg8-f8{1890}
62. Rc3-c6{82} Rf8-f7{2150}
63. Rc6-b6{102} Rb7-b6{1576}
64. Re6-b6{10} Kh8-g7{2306}
65. Rb6-c6{90} Rf7-e7{2146}
66. Pb5-b6{38} Re7-e4{1832}
67. Pb6-b7{52} Re4-e2{1594}
68. Kf2-e2{8} Kg7-f7{0}
--------------------------------------------------------------------------------
help bot: (White)
advance: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM20209&game=Chess

Since I am very weak player and even cant handle the Beginner Level
which plays in 5-6 sec / move. I do not know much about chess playing
techniques.

But looking at this game can you find any weakness in GetClub Game
that need to be corrected. Since Advance level thinks 18 depth deep it
is very difficult to say a move is wrong by Tactics but still Strategy
mistakes can be found in the game.

Please tell what could have been done by GetClub to win the Game. As
Help Bot created a Wall of Pawns not allowing Advance Level to cross
the board.

Bye
Sanny

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






 
Date: 25 May 2008 03:24:05
From: Sanny
Subject: Re: Help Bot struggled to beat the Advance Level.
> > At the momentGetClubis 40 times weaker than Rybka the worlds best
> > program. That isGetClubtakes 40 seconds to think a move which Rybka
> > can search in 1 second.
>
> =A0 Part of that may be the Java applet's inherent
> slowness.

Java applet is 2-3 times slower than a program written in native
language.

Secondly servers are atleast 10 times faster than my old Desktop. So
GetClub is playing 20 times slow because of Hardware and my not
implementing the Parallel Processing.

Currently GetClub do not use the processing power of dual core and
Quad core.

> > OnceGetClubis improved 40 times It will be as strong as worlds
> > fastest Engine. (RYBKA)
>
> =A0 I don't know if Rybka is the fastest engine...
> but it seems to be the strongest.

Yes many say Rybka is the fastest. Does anyone have Rybka Standalone?
Can you play on a Computer having Single core and tell how much slow
GetClub is to the Rybka.

Here is the way one may test.

Play a game setting 8 sec / move and play with Beginner If Game is
equal then Rybka and GetClub speed are same

Then play with Easy Level while Rybka with 8 sec / move. If game is
equal then Rybka is 4 times stronger.

Then play with Normal Level while Rybka with 8 sec / move. If game is
equal then Rybka is 16 times stronger.

Then play with Master Level while Rybka with 8 sec / move. If game is
equal then Rybka is 64 times stronger.


And I think Master Level will play as good as Rybka as acording to my
estimate Rybka is 40 times Stronger.

But the game I play that does calculation on a server. And servers are
10 times faster. Incase On same desktop Rybka gets 40 times stronger
that will be great for me to know.


Bye
Sanny

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


 
Date: 25 May 2008 00:39:30
From: help bot
Subject: Re: Help Bot struggled to beat the Advance Level.
On May 25, 2:05 am, Sanny <[email protected] > wrote:

> Remember earlier Beginner Level used to think for 10 min / move and
> Master Level used to think for 5-10 hours / move.
>
> Now Beginner Plays in just 10 seconds / Move.
>
> So When the game was improved 60 times That was only used to shorten
> the time from 10 min/ move to 10 sec / move.
>
> Once It started playing fast Then It really increased in strength. Now
> no one complains it is taking long time. But last 6 months people were
> complaining it is very weak.

True enough. I stopped playing the lower levels
since I am now much higher-rated than they are,
and one careless move could cost me a lot of
rating points.


> But last 2 month no one said it is weak and most of them are now using
> other Commercial program to win higher Levels. Only you are capable of
> beating without Commercial program help.

Very few great players -- besides me -- know of
the existence of GetClub, so that leaves mainly
the numskulls... .


> And may be Taylor Kingston

You read my mind.


> too win the higher Levels without using help from Computers.

Mr. Kingston is the sort of guy who might use
a computer, then report here his "peak" GetClub
rating, to compare with others' average ratings.


> At the moment GetClub is 40 times weaker than Rybka the worlds best
> program. That is GetClub takes 40 seconds to think a move which Rybka
> can search in 1 second.

Part of that may be the Java applet's inherent
slowness.


> Once GetClub is improved 40 times It will be as strong as worlds
> fastest Engine. (RYBKA)

I don't know if Rybka is the fastest engine...
but it seems to be the strongest.


> Everytime you win a game against Higher Level I learn the mistakes in
> GetClub Chess and it is improved further.

The wins are now taking longer, because I am
playing the higher levels and having a tougher
time winning material (hey, I'm no GM Alekhine
who can whip up a mating attack from out of
nowhere).


> Your chances of winning Master Level is same as what it was against
> winning the Advance Level 2 days back.

About 99%? ; >D


> I hope you may Draw the game. Then lets see how you face the Advance
> Level.

Most of my draws have been the result of my
having a clearly losing position. I don't allow
many draws, even against humans, except
when the alternative is near-certain death. In
fact, I lost a game this way not long ago, and
had another game where I ought to have lost
as a result of refusing a draw offer in a
tournament; that guy gave me a draw even
after I made further blunders.


-- help bot




 
Date: 24 May 2008 23:05:15
From: Sanny
Subject: Re: Help Bot struggled to beat the Advance Level.
> > Game has been further improved taking your advices nowGetClubwill
> > play much stronger game.
>
> =A0 It is amazing just how many times the program has
> been greatly improved while still not exceeding my
> own, rather limited strength. =A0Looking back, I would
> have to guess that it started somewhere around a
> negative five thousand rating, to account for all these
> vast improvements. =A0;>D

Not actually so.

Remember earlier Beginner Level used to think for 10 min / move and
Master Level used to think for 5-10 hours / move.

Now Beginner Plays in just 10 seconds / Move.

So When the game was improved 60 times That was only used to shorten
the time from 10 min/ move to 10 sec / move.

Once It started playing fast Then It really increased in strength. Now
no one complains it is taking long time. But last 6 months people were
complaining it is very weak.

But last 2 month no one said it is weak and most of them are now using
other Commercial program to win higher Levels. Only you are capable of
beating without Commercial program help. And may be Taylor Kingston
too win the higher Levels without using help from Computers.

At the moment GetClub is 40 times weaker than Rybka the worlds best
program. That is GetClub takes 40 seconds to think a move which Rybka
can search in 1 second.

Once GetClub is improved 40 times It will be as strong as worlds
fastest Engine. (RYBKA)

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

> =A0 Right now, I am playing the Master level, and
> I have Black. =A0Pieces are getting traded off, and
> nobody has any significant advantage. =A0But I
> hope to kick butt in the endgame, using my
> vastly superior human understanding to
> overcome a dumb-but-fast calculating machine.

Now, Master level will play as strong as Advance Level played 2 days
ago. Depth of search will be same but the strategy will be much better
making it play as good as Advance Level in overall strength.

Everytime you win a game against Higher Level I learn the mistakes in
GetClub Chess and it is improved further.

Your chances of winning Master Level is same as what it was against
winning the Advance Level 2 days back.

I hope you may Draw the game. Then lets see how you face the Advance
Level.

Bye
Sanny

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




 
Date: 24 May 2008 20:18:00
From: help bot
Subject: Re: Help Bot struggled to beat the Advance Level.
On May 24, 2:01 am, Sanny <[email protected] > wrote:

> Game has been further improved taking your advices now GetClub will
> play much stronger game.

It is amazing just how many times the program has
been greatly improved while still not exceeding my
own, rather limited strength. Looking back, I would
have to guess that it started somewhere around a
negative five thousand rating, to account for all these
vast improvements. ; >D


> With every new improvement GetClub is becoming stronger and stronger.
> Now on you will have to beat it tactically as now I have improved its
> strategy a bit. So it will not allow ro form a wall of pawns.

It appears that the wall of pawns resulted in *me*
missing several tactical blows. Another way for me
to handle the computer's superior tactical strength
is by a combination of simple, solid play, and
simplification. I may yet be able to win, say, a
Rook and pawns ending, from a dead-even position
or one where I have just a miniscule edge.

As I recall, an old game between DeepThought
and GM Karpov had the former gobbling pawns
while the human player finagled a pair of
connected passed pawns, to win. It's not so
easy for programmers to get everything just right,
and sometimes a chess program is calculating
millions of variations which a strong human
player "just knows" are irrelevant.

Right now, I am playing the Master level, and
I have Black. Pieces are getting traded off, and
nobody has any significant advantage. But I
hope to kick butt in the endgame, using my
vastly superior human understanding to
overcome a dumb-but-fast calculating machine.


-- help bot





 
Date: 23 May 2008 23:01:28
From: Sanny
Subject: Re: Help Bot struggled to beat the Advance Level.
Game has been further improved taking your advices now GetClub will
play much stronger game.

With every new improvement GetClub is becoming stronger and stronger.
Now on you will have to beat it tactically as now I have improved its
strategy a bit. So it will not allow ro form a wall of pawns.

Lets see how it plays in new games and if anything else needs change.

Bye
Sanny

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



 
Date: 23 May 2008 16:28:08
From: help bot
Subject: Re: Help Bot struggled to beat the Advance Level.
On May 23, 4:43 am, Sanny <[email protected] > wrote:

> > In this game, the recent problem with Rook to Knight-one
> > manifested itself. In general, such moves only make sense
> > if preparing for a wild no-holds-barred pawn-storm of the
> > castled position, sacrificing to open lines if necessary. In
> > that case, the Black King would need to find shelter on the
> > other wing-- which did not happen here. This R-N1 problem
> > has appeared in several of my most recent games.

> Yes I have seen R-N1 is played many times when no other good move is
> found.
>
> Is placing Rook to Knight place a weak move? Why is this move
> considered weak move I think there is no piece loss?

In this game the computer played R-N1 on both sides
of the board, which left the King unable to castle either
way. In general, the timing of these moves in my
games has been such that a passive, do-nothing
"attitude" manifests itself. Meanwhile, I might be
completing my piece development or maneuvering to
control a vital open file, giving me a winning advantage.


> Should Rook not allowed to move in 1st Rank? How much penalty should
> be added to this move?

Instead of applying the stick to errant Rooks, why
not just dangle a carrot in front of *good* moves?


> > Another error was the spite check, ...Q-e7+. This was
> > obviously a bad move, but I will say that even grandmasters
> > -- such as Yasser Seirawan, for instance -- have made the
> > very same mistake!

> It played 7..... Qe7 So tyhat White King move and unable to chastle.

That's what GM Seirawan thought as well, in his
famous game against Boris Spassky in the King's
Gambit!


> > The Queen incursion (i.e. ...Qf4) / trade was fatal, as I
> > controlled both open files and only a desperate pawn sac'
>
> I see 58 ....Qf4 As there was no way to stop Rook from taking the
> pawns.

Even if, as you say, there was no way to stop White
from taking the pawn(s), when losing, the proper
approach (as I have learned from *vast* experience in
this area) is to obtain counter-play. Otherwise, even
a relatively weak player can finish you off without too
much trouble.


> > The finish was a bit shocking,
> > though; instead of setting up a potential stalemate,
> > theGetClubprogram gave away its last Rook and then
> > instantaneously "resigned", as if it somehow knew that I
> > was not a seventh grader, who might fall for such a trick.

> Actually Advance Level seeing at 18 ply was able to see a Forced Mate
> for all other moves.

I never got to see the final listed move, ...Kf7; on
my screen, it appeared as though the instant I
captured the Rook, the program "resigned". Any
decent human player would have either delayed
the Queening of the pawn as long as possible, or
else set up a potential stalemate (Black King on
h6, White Rook on the seventh rank, etc.) in hope
of swindling a draw. I realize this is not something
which can easily be programmed into a computer,
since it requires a huge blunder by the opponent
to work. It relates to the psychology of a player
who knows he is winning easily, and who then
becomes careless (which never happens to me,
of course). : >D


> In end game Advance may be thinking 22 ply deep. And I think it had
> seen a Mate in 11 Moves or Mate in 22 Ply.

Not bad, but the moves actually played do not
reflect this depth, since it plays very-sub-optimal
defense at the end of each game it loses.


> > The idea of locking up pawns and avoiding trades was
> > to cramp Black, and then figure out a way to overwhelm
> > his defenses at some point. However, what actually
> > happened was that I was unable to penetrate on the
> > e-file by force, and only via a risky sacrifice of my
> > Queen-side pawns was I able to sneak my way in--
> > with considerable help from my opponent.

> I do not see any sacrifice of your Queen side pawns. At the end number
> of pawns were same.

The program did not take them, but they were
offered. (Instead, it went in for a Queen trade with
...Q-f4, losing like a cucumber.) After ...Nxb5, my
Queen-side pawns are defenseless, but if they
are taken, I get to chomp the Black pawn on f6
and then attack the enemy King.


> > Black was quite cramped in this game, and the
> > tactical wizardry of the Advance level was, perhaps,
> > rendered invisible. When Sanny says it sees 18
> > moves ahead, I suspect he means 18 plys-- or very
> > roughly 9 moves ahead.

> Yes it sees 18 ply in Middle game and upto 22 ply in end game. Or 9
> Moves in Middle game and 11 moves in end game.
>
> Great Analysis.

I ran this through another chess program and
sure enough, I missed several different ways to
win material in tricky combinations, and one
which was quite obvious. One example was
where I could have played Nxg5, and if the Rook
recaptures, I win the Black Queen with R-e8.
But here, if the pawn captures the Knight, it is
not quite so simple. At the time, I was thinking
about another sacrifice (Nxh4), which had
carefully been circumvented by Black.

The worst mistake in the entire game may well
have been mine: 51. R/1-e2 shows the position
score falling off a cliff, because I failed to find the
crushing blow 51. Nxg5!. From about moves 45
through 50, my advantage nearly evaporated as
I prepared to sacrifice on h4 or triple on the open
file. Clearly, when it comes to tactics, I'm just
not in the same "class" as computers (not even
in the same school district).


-- help bot


 
Date: 23 May 2008 05:11:13
From: Sanny
Subject: Re: Help Bot struggled to beat the Advance Level.
> GetClubwas never even close to winning. The only time he had a chance
> for a fully acceptable position was on move 7 with 7...Be7. After that
> the white advantage fluctuated between completely winning and just a bit
> better, but the computer never did anything active himself, so there was
> never any danger to white and he was just going to win eventually.

A few improvements were done based on your suggestions Lets see if it
plays better.

I played a game against Rybka. And found Rybka easily won the game
against Easy Level.

Easy level fall in a trap and lost the game.

So despite improvement made Rybka was way ahead in tactics.

Game Played between sanjay11 and easy at GetClub.com

--------------------------------------------------------------------------------
sanjay11: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM20227&game=Chess
--------------------------------------------------------------------------------

White -- Black
(sanjay11) -- (easy)

1. d2-d4{8} d7-d5{0}
2. Ng1-f3{16} Ng8-f6{0}
3. c2-c4{12} d5-c4{34}
4. e2-e3{14} Nf6-e4{22}
5. Bf1-c4{16} e7-e6{20}
6. Ke1-g1{18} Nb8-c6{46}
7. Qd1-c2{14} f7-f5{32}
8. Nb1-c3{12} Ne4-c3{28}
9. b2-c3{12} h7-h6{44}
10. e3-e4{14} Bf8-e7{42}
11. e4-f5{18} e6-f5{38}
12. Rf1-e1{18} Qd8-d6{22}
13. d4-d5{20} Nc6-d8{26}
14. a2-a4{20} c7-c5{36}
15. d5-c6{14} Nd8-c6{24}
16. Bc1-a3{18} Qd6-c7{52}
17. Re1-e7{20} Nc6-e7{24}
18. Ra1-e1{18} Bc8-e6{40}
19. Bc4-e6{22} Qc7-f4{20}
20. Be6-f5{20} Ke8-g8{22}
21. Bf5-e6{16} Kg8-h8{24}
22. Ba3-e7{12} Qf4-c7{48}
23. Be7-f8{14} Ra8-f8{20}
24. Nf3-e5{40} Rf8-f6{26}
25. Ne5-g6{16} Rf6-g6{26}
26. Qc2-g6{36} Qc7-e7{30}
27. Kg1-f1{14} Qe7-f6{20}
28. Qg6-e8{12} Kh8-h7{36}
29. Be6-g8{12} Kh7-h8{0}
--------------------------------------------------------------------------------
sanjay11: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM20227&game=Chess

After 16th move Rybka captured a lot of pieces and Easy level could
not do anything.

Do you think this game was lost because of weaker Tactics of Easy
Level. Easy level plays in just 20 sec/ move. It is much faster than
the Advance level which takes 30 min / move.

Bye
Sanny

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



 
Date: 23 May 2008 01:43:35
From: Sanny
Subject: Re: Help Bot struggled to beat the Advance Level.
> =A0 In this game, the recent problem with Rook to Knight-one
> manifested itself. =A0In general, such moves only make sense
> if preparing for a wild no-holds-barred pawn-storm of the
> castled position, sacrificing to open lines if necessary. =A0In
> that case, the Black King would need to find shelter on the
> other wing-- which did not happen here. =A0This R-N1 problem
> has appeared in several of my most recent games.

Yes I have seen R-N1 is played many times when no other good move is
found.

Is placing Rook to Knight place a weak move? Why is this move
considered weak move I think there is no piece loss?

Should Rook not allowed to move in 1st Rank? How much penalty should
be added to this move?



> =A0 Another error was the spite check, ...Q-e7+. =A0 This was
> obviously a bad move, but I will say that even grandmasters
> -- such as Yasser Seirawan, for instance -- have made the
> very same mistake!

It played 7..... Qe7 So tyhat White King move and unable to chastle.


> =A0 The Queen incursion (i.e. ...Qf4) / trade was fatal, as I
> controlled both open files and only a desperate pawn sac'

I see 58 ....Qf4 As there was no way to stop Rook from taking the
pawns.


>=A0The finish was a bit shocking,
> though; instead of setting up a potential stalemate,
> theGetClubprogram gave away its last Rook and then
> instantaneously "resigned", as if it somehow knew that I
> was not a seventh grader, who might fall for such a trick.

Actually Advance Level seeing at 18 ply was able to see a Forced Mate
for all other moves.

In end game Advance may be thinking 22 ply deep. And I think it had
seen a Mate in 11 Moves or Mate in 22 Ply.



> =A0 The idea of locking up pawns and avoiding trades was
> to cramp Black, and then figure out a way to overwhelm
> his defenses at some point. =A0However, what actually
> happened was that I was unable to penetrate on the
> e-file by force, and only via a risky sacrifice of my
> Queen-side pawns was I able to sneak my way in--
> with considerable help from my opponent.

I do not see any sacrifice of your Queen side pawns. At the end number
of pawns were same.

> =A0 Black was quite cramped in this game, and the
> tactical wizardry of the Advance level was, perhaps,
> rendered invisible. =A0When Sanny says it sees 18
> moves ahead, I suspect he means 18 plys-- or very
> roughly 9 moves ahead.

Yes it sees 18 ply in Middle game and upto 22 ply in end game. Or 9
Moves in Middle game and 11 moves in end game.

Great Analysis.

Bye
Sanny

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




 
Date: 23 May 2008 09:43:08
From: Bjoern
Subject: Re: Help Bot struggled to beat the Advance Level.
Sanny wrote:
> There were no Tactical Mistake in this game as Both had equal pieces
> till end. But I think there were strategical mistakes by GetClub.

There were plenty of tactical mistakes and strategic ones, too. Looking
at it with a slightly stronger computer program + human insight makes
that quite obvious.

> But since this was Advance Level which thinks 18 moves ahead it is
> very difficult to win by taking its pieces.

Actually there were some such chances e.g. 42.Re8.

> So Help Bot was able to
> use good strategy in beating the Advance Level.

That too. In fact he missed quite a few winning tactics, but he didn't
really need them, because of the strategic inferiority of the program.

> Game Played between help bot and advance at GetClub.com
>
> --------------------------------------------------------------------------------
> help bot: (White)
> advance: (Black)
> Game Played at: http://www.GetClub.com/Chess.html
> View Recorded Game: http://www.getclub.com/playgame.php?id=DM20209&game=Chess
> --------------------------------------------------------------------------------
>
> White -- Black
> (help bot) -- (advance)
>
> 1. c2-c3{26} e7-e5{0}
> 2. g2-g3{16} Ng8-f6{1548}
> 3. Bf1-g2{18} d7-d5{1792}
> 4. d2-d4{34} e5-e4{1724}
> 5. f2-f3{12} Bc8-f5{2352}
> 6. Ng1-h3{90} e4-f3{2180}
> 7. e2-f3{16} Qd8-e7{2268}

Just wrong 7.Be7 is the right move.

> 8. Ke1-f2{46} Qe7-d7{2420}
> 9. Nh3-f4{74} Rh8-g8{1344}

The last 3 moves moves by black were just tactically and strategically
wrong (well, okay 8...Qd7 is alright, but both 7...Qe7 and especially
9...Rg8 are not moves one should play). I suppose the computer thinks
there is some inherent value in preventing white from castling, when
really white doesn't need to and black's king is in more trouble here
than white's.

White could have played even more energetically to exploit it (e.g. 13.g4).

> 10. Rh1-e1{60} Bf8-e7{3106}
> 11. Nf4-d3{52} Nb8-c6{1464}
> 12. Nd3-c5{84} Qd7-c8{2474}
> 13. Bg2-f1{98} Ke8-f8{1502}
> 14. Nc5-d3{40} Nf6-d7{2138}
> 15. b2-b3{372} g7-g5{2812}
> 16. Nb1-a3{184} h7-h5{1852}
> 17. Na3-c2{92} h5-h4{1862}
> 18. g3-g4{18} Bf5-e6{1520}
> 19. Kf2-g2{94} Ra8-b8{2722}

Not sure what else the computer should play and the move might not be so
bad in itself, but this move is a sign of a buggy positional evaluation
algorithm.

> 20. Nc2-e3{136} Be7-d6{3254}
> 21. Kg2-h1{72} Nd7-b6{3970}

The black knight has no prospects here.

> 22. a2-a4{162} Rg8-g6{1782}
> 23. Bc1-a3{78} Bd6-a3{1470}

The last couple of moves by black were just pointless and strategically
poor, so now white manages to swap off the good bishop of black and thus
makes f3-f4 possible, which 3 moves later gains him an advantage.

I wonder Sanny, do you have an algorithm that gives pieces different
values depending on the pawn structure? Here for example the black
bishop on d6 is better than the bishop on e6. The tricky bit of course
is to assign the right kind of value to this difference compared to
other elements in the game.

> 24. Ra1-a3{10} Rg6-f6{1736}
> 25. Ra3-a2{92} Kf8-g8{2204}
> 26. Ra2-f2{136} Qc8-f8{1642}
> 27. f3-f4{64} Qf8-d8{1576}

The 27.f4 move is just great for white and he is probably just winning now.

> 28. f4-f5{268} Be6-d7{1694}
> 29. Bf1-g2{112} Rf6-d6{1690}
> 30. Kh1-g1{120} Qd8-f6{1872}
> 31. Nd3-c5{44} Qf6-h6{2288}

What are all those queen moves meant to be good for? They achieve
nothing and black could be trying to do something more constructive
instead. As it is black is just moving his queen forth and back while
white can calmly improves his position.

I think your computer currently just cannot handle such closed (i.e. the
pawn chains are locked with each other and the pieces are not directly
tangling with each other). While these positions are sometimes quite
even, I think you should program your computer so it avoids them and
heads for positions with more immediate tactics.

> 32. h2-h3{98} Qh6-f6{1536}
> 33. Rf2-e2{98} a7-a5{1452}
> 34. Ne3-f1{110} Kg8-h7{1806}
> 35. Nf1-h2{138} Qf6-d8{2114}
> 36. Nh2-f3{96} f7-f6{1386}
> 37. Re2-e3{152} Kh7-h8{3110}
> 38. Bg2-f1{54} Nb6-a8{2004}
> 39. Bf1-b5{88} b7-b6{1732}

This position illustrates clearly what is wrong with blacks play. All of
black's pieces are uselessly hanging around on the queenside and can
hardly move (just look at the knight on a8!). Here black's situation is
hopeless.

> 40. Nc5-d7{38} Qd8-d7{2220}
> 41. Qd1-e2{138} Rb8-g8{2976}

Actually now 42.Re8 would have won the knight on a8, but white does not
take that option and will obviously win anyway.

> 42. Kg1-f2{102} Qd7-f7{1810}
> 43. Re3-e6{290} Nc6-a7{1328}
> 44. Re6-e7{430} Qf7-f8{3652}
> 45. Re7-e8{74} Qf8-g7{1900}
> 46. Bb5-d3{132} Rd6-d8{2334}
> 47. Re8-e6{124} Rd8-d7{1486}

Now white should play 48.Bb5 to not allow 48...c6.

> 48. Kf2-f1{210} c7-c6{1816}

This is a strategically good move by the computer (did I just say
that?!) and now white is only clearly better and maybe not completely
winning any more - unless he spots such a strong move as 49.Ne5 (wins).

> 49. Qe2-f2{122} Qg7-f8{1688}
> 50. Kf1-g2{260} Rd7-h7{2590}

What kind of move is that? Obviously one needs to play 50.Nc7, I would
have thought that was part of the point of 48...c6 (which I praised
above), but clearly the computer played the right move for the wrong
reasons 3 moves ago. Regarind 50...Rh7: strategically it's insane and
tactically it's wrong, even if white doesn't punish it (as he could
have) with 51.Nxg5 fxg5 52.f6 Rf7 53.Re7!.

Now white actually makes a few strategic mistakes and black gets a
nearly playable position again.

> 51. Re1-e2{112} Na8-c7{1938}
> 52. Re6-e3{2218} c6-c5{2018}
> 53. Qf2-e1{134} Na7-c6{1702}
> 54. Bd3-b5{726} c5-d4{2534}
> 55. Nf3-d4{254} Nc6-d4{1588}
> 56. c3-d4{32} Qf8-d6{2438}

Compared to before black has recovered quite well.

> 57. Re3-c3{1130} Nc7-b5{2912}
> 58. a4-b5{18} Qd6-f4{1476}
> 59. Qe1-f2{50} Qf4-f2{2484}

However after being nearly okay 3 moves ago the computer throws it all
away again with three bad moves in a row. If it allegedly looks 18 moves
ahead, then it should somehow suspect that it will loose some pawns now.
Or maybe it sees too many moves inbetween, in which case your position
evaluation algorithm should have some warning signs flashing: White
controls both open files, white has a space advantage, the pawns b6, f6
and d5 are weak and black has no active play. This position is once more
completely winning for white.

> 60. Kg2-f2{14} Rh7-b7{1764}
> 61. Re2-e6{66} Rg8-f8{1890}
> 62. Rc3-c6{82} Rf8-f7{2150}

So, the computer was unable to spot that he was going to loose pawns now???

> 63. Rc6-b6{102} Rb7-b6{1576}
> 64. Re6-b6{10} Kh8-g7{2306}
> 65. Rb6-c6{90} Rf7-e7{2146}
> 66. Pb5-b6{38} Re7-e4{1832}

This is a huge tactical blunder. 66...Re8 keeps the game going for longer.

> 67. Pb6-b7{52} Re4-e2{1594}

Even more bizarre for a computer, this is a helpmate in 4, as opposed to
playing on for much longer after (the equally hopeless) 67...Re8.

> 68. Kf2-e2{8} Kg7-f7{0}

> But looking at this game can you find any weakness in GetClub Game
> that need to be corrected. Since Advance level thinks 18 depth deep it
> is very difficult to say a move is wrong by Tactics but still Strategy
> mistakes can be found in the game.

There's plenty of both and if he had played against an opponent that he
takes more seriously helpbot might have spotted some of the tactics he
missed, but I guess he played this game while doing some other work on
his computer (as I usually do when playing this program) and only
switched the the program window when the computer had moved, thought for
a moment and then went back to his work.

> Please tell what could have been done by GetClub to win the Game. As
> Help Bot created a Wall of Pawns not allowing Advance Level to cross
> the board.

GetClub was never even close to winning. The only time he had a chance
for a fully acceptable position was on move 7 with 7...Be7. After that
the white advantage fluctuated between completely winning and just a bit
better, but the computer never did anything active himself, so there was
never any danger to white and he was just going to win eventually.

--B


 
Date: 23 May 2008 00:01:40
From: help bot
Subject: Re: Help Bot struggled to beat the Advance Level.
On May 23, 2:07 am, Sanny <[email protected] > wrote:

> Help Bot played against Advance Level and to my surprise win the game.
>
> I find Help Bot created a wall of pawns So that Advance Level which
> thinks 18 depth deep could not peneterate the wall. And in the end
> Help Bot was able to get one good Combination and win the Game.
>
> Looks like GetClub is still a baby for Help Bot. He played moves in
> just 2 min / move while GetClub was thinking 30 min / move. Still Help
> Bot was able to find a victory.
>
> (See the Recorded game at Bottom)
> ----------------------------------------------------
>
> The game was even (equal) till 60 Moves. But after the Queens were
> exchanged Help Bot managed to fork Two pawns with rooks and kill one
> of the pawn and then in a few moves He got the Queen.
>
> So help Bot win the game in just 68 moves.
>
> There were no Tactical Mistake in this game as Both had equal pieces
> till end. But I think there were strategical mistakes by GetClub.
>
> But since this was Advance Level which thinks 18 moves ahead it is
> very difficult to win by taking its pieces. So Help Bot was able to
> use good strategy in beating the Advance Level.


> help bot: (White)
> advance: (Black)
> Game Played at:http://www.GetClub.com/Chess.html
> View Recorded Game:http://www.getclub.com/playgame.php?id=DM20209&game=Chess
>
> Since I am very weak player and even cant handle the Beginner Level
> which plays in 5-6 sec / move. I do not know much about chess playing
> techniques.
>
> But looking at this game can you find any weakness in GetClub Game
> that need to be corrected. Since Advance level thinks 18 depth deep it
> is very difficult to say a move is wrong by Tactics but still Strategy
> mistakes can be found in the game.
>
> Please tell what could have been done by GetClub to win the Game. As
> Help Bot created a Wall of Pawns not allowing Advance Level to cross
> the board.

In this game, the recent problem with Rook to Knight-one
manifested itself. In general, such moves only make sense
if preparing for a wild no-holds-barred pawn-storm of the
castled position, sacrificing to open lines if necessary. In
that case, the Black King would need to find shelter on the
other wing-- which did not happen here. This R-N1 problem
has appeared in several of my most recent games.

Another error was the spite check, ...Q-e7+. This was
obviously a bad move, but I will say that even grandmasters
-- such as Yasser Seirawan, for instance -- have made the
very same mistake!

The Queen incursion (i.e. ...Qf4) / trade was fatal, as I
controlled both open files and only a desperate pawn sac'
(...a4) would allow for any counter-play whatsoever; such a
sacrifice would have placed me up two pawns-- surely a
good thing for White. Still, it was better than passive
(non-)defense, as in the game.

Apparently, the program saw my prospective Knight sac'
(on h4) coming, for it defended with ...R-h7 at just the right
time. It also managed to cover penetration squares on the
open e-file after I tripled my heavy pieces, intending to
brute-force my way in. The finish was a bit shocking,
though; instead of setting up a potential stalemate, the
GetClub program gave away its last Rook and then
instantaneously "resigned", as if it somehow knew that I
was not a seventh grader, who might fall for such a trick.

The idea of locking up pawns and avoiding trades was
to cramp Black, and then figure out a way to overwhelm
his defenses at some point. However, what actually
happened was that I was unable to penetrate on the
e-file by force, and only via a risky sacrifice of my
Queen-side pawns was I able to sneak my way in--
with considerable help from my opponent.

Black was quite cramped in this game, and the
tactical wizardry of the Advance level was, perhaps,
rendered invisible. When Sanny says it sees 18
moves ahead, I suspect he means 18 plys-- or very
roughly 9 moves ahead.


-- help bot