Sanny writes:

So the reasion of this horrible play is that Crafty is thinking when
GetClubProgram is also thinking. So the strength ofGetClubBeginner
level reduces to half. Since Beginner Level has to play in 5-10
seconds So if other program is also running simultaniously then the
strength of theGetClubProgram reduced to half.

I'm using a quad core CPU andgetclubis getting one core all to
itself. Crafty is using the other 3 cores. It doesn't look likegetclub
can use more than one core, so turning off crafty's thinking on
opponents time would not makegetclubstronger. (But would make crafty
weaker.)

How do you know Crafty is not using all the Four Processors. And
GetClub is getting Complete one Core. May Be Crafty is taking 3.5
Cores and GetClub is using just 0.5 Core So Crafty is actually using 7
times more processing power when GetClub is moving.

I "reniced" the crafty threads so that the GetClub thread is getting
priority, and I verified with "top" that GetClub is getting 100% of
one core. (I'm using linux, I'm aware that "renice" and "top" may not
be familiar to windows users.)

And in its own move it is taking 8 times more power than GetClub is
taking.

So a total of 7+8=15 times for CPU usage. So if Crafty is using 15
times more power than GetClub then I can imagine such poor games.

I'll try some more games with crafty using only one thread and not
thinking on the opponents time to see if it makes a difference.

--
Peter Osterlund -
http://web.telia.com/~u89404340

I Played a game between GetClub Easy Level which takes 20-40 sec /
move and Jester which always takes 40 seconds/ move.

I played both with Higher level So that there is less calculation
mistake and we can see what real mistake GetClub Chess is making.

Please Analyze and let me know which were the wrong moves that must be
avoided by Black. As Black lost the game in just 56 moves despite
playing with Easy level which is one step ahead of Beginner Level.

Since Both took atleast 40 seconds per move a strong analysis will be
needed to find the mistakes in the game.

Game Played between Jester and easy at GetClub.com

--------------------------------------------------------------------------------
Jester: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?...493&game=Chess
--------------------------------------------------------------------------------

White -- Black
(Jester) -- (easy)

1. e2-e4{16} e7-e5{0}
2. Ng1-f3{6} Nb8-c6{0}
3. Bf1-b5{6} a7-a6{0}
4. Bb5-c6{6} d7-c6{82}
5. Ke1-g1{8} Ng8-f6{30}
6. d2-d3{48} Qd8-d6{22}
7. Nb1-c3{46} Bc8-g4{62}
8. h2-h3{50} Bg4-h5{82}
9. Bc1-g5{52} Ke8-c8{82}
10. g2-g4{48} Bh5-g6{42}
11. Nf3-h4{46} Kc8-b8{38}
12. Qd1-c1{48} Bf8-e7{82}
13. Nc3-e2{74} b7-b5{84}
14. a2-a4{50} h7-h6{20}
15. Nh4-g6{260} f7-g6{28}
16. Bg5-e3{50} b5-b4{50}
17. b2-b3{50} c6-c5{48}
18. Qc1-b2{44} Nf6-h7{22}
19. Kg1-h2{46} Be7-h4{72}
20. f2-f4{50} Bh4-f6{20}
21. Qb2-b1{46} e5-f4{42}
22. Be3-f4{48} Qd6-e6{22}
23. Ra1-a2{46} Bf6-e5{24}
24. Qb1-e1{50} Rh8-f8{20}
25. Bf4-e5{46} Qe6-e5{28}
26. Qe1-g3{44} Qe5-g3{68}
27. Ne2-g3{46} Rf8-f1{20}
28. Ng3-f1{60} Nh7-g5{28}
29. Kh2-g3{48} c5-c4{52}
30. b3-c4{50} Rd8-f8{26}
31. Nf1-d2{46} Kb8-a8{82}
32. Ra2-b2{44} c7-c5{72}
33. a4-a5{88} Rf8-f6{62}
34. Rb2-b1{64} h6-h5{50}
35. Rb1-b3{48} Ng5-e6{44}
36. e4-e5{48} Rf6-f8{22}
37. Nd2-e4{78} h5-g4{24}
38. h3-g4{50} Rf8-d8{26}
39. Ne4-d6{50} Ne6-d4{58}
40. Rb3-b2{96} g6-g5{28}
41. Nd6-e4{50} Nd4-e6{82}
42. Rb2-b1{46} Ka8-b7{24}
43. Ne4-d6{48} Kb7-a7{48}
44. Rb1-f1{48} Rd8-h8{52}
45. Rf1-f7{46} Ka7-a8{82}
46. Rf7-e7{50} Rh8-h6{22}
47. Nd6-e4{48} Ka8-b8{20}
48. Ne4-f6{48} Ne6-d4{26}
49. Nf6-d7{46} Kb8-a8{34}
50. Nd7-c5{44} Nd4-c2{32}
51. Re7-e8{44} Ka8-a7{20}
52. e5-e6{44} g7-g6{56}
53. e6-e7{46} Rh6-h7{40}
54. Re8-a8{44} Ka7-a8{36}
55. Qe7-e8{Q}{20} Ka8-a7{78}
56. Qe8-g6{18} Rh7-d7{52}
--------------------------------------------------------------------------------
Jester: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?...493&game=Chess

In the End game GetClub played weak moves and lost. It was equal game
in middle game. What are the serious wrong moves played by GetClub
Easy Level.

Bye
Sanny

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

In message , Peter Osterlund
writes
Sanny writes:

On Feb 2, 1:20*pm, Peter Osterlund wrote:
Sanny writes:
Now it is very difficult to win even the Beginner Level, So it is very
difficult to get higher ratings.

A typical game played by Peter with Beginner Level which he won in
just 22 moves.

I should add that I'm using a computer, mainly to test the playing
strength of theGetClubengine. I'm nowhere near that good myself.

That Solves the Mystry, Thats what I was wondering others are taking
80 Moves to win and you are winning in just 20-30 moves. Which Engine
you are testing with is it Fritz/ Rybka/ Chess Master ?

I'm testing with crafty, set to an average thinking time of 5
seconds. However, it does think also when it's the oppoenents turn to
move, so the actual thinking time gets longer in the cases where
crafty correctly predicts which move the opponent will make.

With these settings for Crafty and your hardware I would hazard a guess
that when it guesses the principle variation correctly you will have a
full depth search to something like 12ply with extensions out to 24 or
so and if it is restricted to 5s around 10ply with extensions to ~20.

This will mean that Crafty will see things that are beyond the search
horizon for GetClubs engine. You could also try halving the allowed
thinking time to say 2s or even 1s to make the search depths more evenly
matched. Even then I suspect that Crafty's selective extensions will
give it the edge. A diagnostic that Getclub could add to the thinking
time in the final logged game is the machine evaluation and depth
searched.

What is the rating of Beginner Level as per the games played by it.

I can't get a rating number out of the games I've played, but here are
some subjective thoughts: (Others have said similar things before.)

If the application you are running Crafty under allows nominal fixed
depth searches it might be interesting to try 8ply or 9ply and then make
the search gradually shallower until Beginner can just win.

* It seems to prefer playing knights to a3/h3/a6/h6 too often. In
general, knights are stronger the closer they are to the center of
the board, so it should prefer c3/f3/c6/f6 instead.

* There seems to be something wrong in the quiescence search which
sometimes makes it do very bad things when the position gets
complicated. Someone gave the good advice that the quiescence search
should not stop as long as its possible to capture a piece that is
worth more than the piece used for the capture.

For a general explanation of quiescence search, see this:

http://en.wikipedia.org/wiki/Quiescence_search

* Another possibility is that the program is vulnerable to the
"horizon effect". I've seen it do bad moves that give away more
material than necessary just to delay the loss of other material.
"Singular extensions", "capture extensions" and/or "check
extensions" might be able to solve some of these problems.

For a general explanation of the horizon effect, see this:

http://en.wikipedia.org/wiki/Horizon_effect

* All in all it plays pretty decent chess for beginner level. If you
were to correctly implement all suggestions given by all people, the
program would probably be so strong that most people would not be
interested in playing against it.

However, it would be nice if the higher levels could be made
stronger without increasing the thinking time so much. The reason I
haven't tested the higher levels yet is that it would simply take
more time than I have available.

Also, regarding the "killer heuristic" asked in another post,
wikipedia has a pretty good explanation he

http://en.wikipedia.org/wiki/Killer_heuristic

He would probably be best off getting a copy of the fairly old Computer
Chess Compendium (ed David Levy) for its detailed description of the
internal workings of the Northwestern University Chess 4.5. It is quite
dated now but it would probably still include tricks that are apparently
missing from GetClub.

Index of contents is online at:
http://portal.acm.org/citation.cfm?i...DE&CFID=148 0
7911&CFTOKEN=15688007

The hardback version seems to go for around $25 now. It is interesting
since my original softback copy cost me £25 donkeys years ago.

Regards,
--
Martin Brown

--
Posted via a free Usenet account from http://www.teranews.com

Peter Osterlund writes:

I'll try some more games with crafty using only one thread and not
thinking on the opponents time to see if it makes a difference.

I played one game with crafty using only one thread, not thinking on
the opponents time, and using one second per move thinking time. This
should give GetClub a factor 10 advantage in terms of CPU time.

Crafty still got into a winning position but then GetClub cheated and
declared the game a draw after the same position was repeated two
times. (It must be repeated three times before you are allowed to
claim a draw.)

--
Peter Osterlund -
http://web.telia.com/~u89404340

Martin Brown writes:

If the application you are running Crafty under allows nominal fixed
depth searches it might be interesting to try 8ply or 9ply and then
make the search gradually shallower until Beginner can just win.

I set the search depth in crafty to 5 ply, which reduces its thinking
time to about 0.01 seconds. A game against GetClub beginner level was
much more even than before, but crafty still managed to win in the end
game, mostly because GetClub played very weakly.

(The opening book I'm using for crafty seems to contain some really
bad lines. Crafty evaluated the position to -1.5 the first move after
the opening book, which was move 6. Ne5.)

[Event "Computer chess game"]
[Site "quad.localdomain"]
[Date "2008.02.03"]
[Round "-"]
[White "Crafty-20.14 5 ply search depth"]
[Black "GetClub beginner level"]
[Result "1-0"]
[TimeControl "60/120"]

1. e4 e5 2. f4 exf4 3. Nf3 g5 4. Bc4 Nc6 5. d4 g4 6. Ne5 Nxe5 7. dxe5 Qh4+
8. Kd2 f3 9. gxf3 Qf2+ 10. Kc3 Qxf3+ 11. Qxf3 gxf3 12. Rf1 Bg7 13. Bf4 d6
14. Rxf3 dxe5 15. Bg3 Be6 16. Bxe6 fxe6 17. Nd2 Nf6 18. Raf1 O-O 19. a3
Nd5+ 20. Kb3 Rxf3+ 21. Rxf3 Nf4 22. Bxf4 exf4 23. Rxf4 c6 24. Rf1 Rf8 25.
Rd1 Kh8 26. Nc4 b5 27. Na5 c5 28. Rd7 a6 29. c4 Rf2 30. Rd8+ Rf8 31. Rxf8+
Bxf8 32. e5 Bh6 33. cxb5 axb5 34. Nc6 Bg7 35. Kc3 h6 36. b3 c4 37. bxc4
bxc4 38. a4 Bf8 39. a5 Kh7 40. a6 Bc5 41. Kxc4 Bg1 42. a7 Bxa7 43. Nxa7 Kh8
44. Nc6 Kh7 45. Nd4 Kg6 46. Nxe6 h5 47. Nf4+ Kg5 48. Nxh5 Kxh5 49. Kd5 Kg5
50. e6 Kf6 51. Kd6 Kg6 52. h4 Kf6 53. e7 Kf7 54. Kd7 Kg8 55. e8=Q+ Kh7 56.
Ke6
{Black resigns} 1-0

--
Peter Osterlund -
http://web.telia.com/~u89404340

On Feb 3, 4:15*pm, Peter Osterlund wrote:
Martin Brown writes:
If the application you are running Crafty under allows nominal fixed
depth searches it might be interesting to try 8ply or 9ply and then
make the search gradually shallower until Beginner can just win.

I set the search depth in crafty to 5 ply, which reduces its thinking
time to about 0.01 seconds. A game againstGetClubbeginner level was
much more even than before, but crafty still managed to win in the end
game, mostly becauseGetClubplayed very weakly.

(The opening book I'm using for crafty seems to contain some really
bad lines. Crafty evaluated the position to -1.5 the first move after
the opening book, which was move 6. Ne5.)

[Event "Computerchessgame"]
[Site "quad.localdomain"]
[Date "2008.02.03"]
[Round "-"]
[White "Crafty-20.14 5 ply search depth"]
[Black "GetClubbeginner level"]
[Result "1-0"]
[TimeControl "60/120"]

1. e4 e5 2. f4 exf4 3. Nf3 g5 4. Bc4 Nc6 5. d4 g4 6. Ne5 Nxe5 7. dxe5 Qh4+
8. Kd2 f3 9. gxf3 Qf2+ 10. Kc3 Qxf3+ 11. Qxf3 gxf3 12. Rf1 Bg7 13. Bf4 d6
14. Rxf3 dxe5 15. Bg3 Be6 16. Bxe6 fxe6 17. Nd2 Nf6 18. Raf1 O-O 19. a3
Nd5+ 20. Kb3 Rxf3+ 21. Rxf3 Nf4 22. Bxf4 exf4 23. Rxf4 c6 24. Rf1 Rf8 25.
Rd1 Kh8 26. Nc4 b5 27. Na5 c5 28. Rd7 a6 29. c4 Rf2 30. Rd8+ Rf8 31. Rxf8+
Bxf8 32. e5 Bh6 33. cxb5 axb5 34. Nc6 Bg7 35. Kc3 h6 36. b3 c4 37. bxc4
bxc4 38. a4 Bf8 39. a5 Kh7 40. a6 Bc5 41. Kxc4 Bg1 42. a7 Bxa7 43. Nxa7 Kh8
44. Nc6 Kh7 45. Nd4 Kg6 46. Nxe6 h5 47. Nf4+ Kg5 48. Nxh5 Kxh5 49. Kd5 Kg5
50. e6 Kf6 51. Kd6 Kg6 52. h4 Kf6 53. e7 Kf7 54. Kd7 Kg8 55. e8=Q+ Kh7 56.
Ke6
{Black resigns} 1-0

--
Peter Osterlund - /~u89404340

It looks like Crafty keeps running when GetClub is thinking. And it is
not allowing GetClub Program to take enough resources.

I suggest use Crafty on some other computer and then test so that
Crafty do not spoil GetClubs Calculations.

On my Computer GetClubs Beginner Level thinks 5-6 Depth deep. How much
is shown on your Computer.

Two things need toi be done play both on Single Core computer with
Thinking in Opponents move not allowed Else Crafty gets 15 times more
processing power.

I am also thinking to have it utilize the power of Quad Cores. So that
it plays better and also to think on opponents move.

Bye
Sanny

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

In message , Peter Osterlund
writes
Martin Brown writes:

If the application you are running Crafty under allows nominal fixed
depth searches it might be interesting to try 8ply or 9ply and then
make the search gradually shallower until Beginner can just win.

I set the search depth in crafty to 5 ply, which reduces its thinking
time to about 0.01 seconds. A game against GetClub beginner level was
much more even than before, but crafty still managed to win in the end
game, mostly because GetClub played very weakly.

Yes. Getclub really needs a tweak to stop its king from cowering in one
corner of the board once the total enemy material is down to endgame
levels (say 8-10 pawns equivalent). Unless the King comes into more
active play it is always going to be hammered in the endgames.

(The opening book I'm using for crafty seems to contain some really
bad lines. Crafty evaluated the position to -1.5 the first move after
the opening book, which was move 6. Ne5.)

If it is any consolation Powerbooks6 has the same questionable line in
it. Sometimes it is a mismatch between the engines numerical view of the
situation and the ideas or positional features that it cannot see.

However, none of the engines I tried liked this line very much.

[White "Crafty-20.14 5 ply search depth"]
[Black "GetClub beginner level"]
[Result "1-0"]
[TimeControl "60/120"]

1. e4 e5 2. f4 exf4 3. Nf3 g5 4. Bc4 Nc6 5. d4 g4 6. Ne5 Nxe5 7. dxe5 Qh4+

I am not very familiar with the King's gambit so I could be hopelessly
wrong here. But to my eyes 5. ... g4 (which Getclub found) looks a lot
better than the standard book line of Bg7. Strangely 6. O-O seems OK in
theory.
Even Rybka agrees that 5. ... g4 looks pretty good here (and it isn't
easily fooled). It gives Bg7 and h6 joint second place (transposition
table effect).

And 6. Ne5 should be refuted by Qh4+
(which Crafty 19.19 here found with a 5 ply search on 2 cpus @ 3GHz)

The two engines seemed to be pretty well matched though with near
equality for most of the match. Until the endgame where Getclubs
inherent weaknesses become far more visible.

25.
Rd1 Kh8

This is a classic GetClub losing move. King cowers in the corner and it
loses a tempo. There were good attacking moves possible here and the
king is not in any danger. b5, Rf2 or a6 look reasonable.

25. ... Kh8 is actually so bad that the only moves which evaluate worse
than it are the ones which lead to immediate material loss namely:

Bxb2, Bc3, Rf1, Rf3, Rf5

The king protection logic has to be fixed if the program is to survive
in the endgame. It also needs to recognise pawn can run situations
rather than finding them as a complete surprise.

White missed a trick here with 32. Nc6 or h3

34. Nc6 Bg7

34. .... Bg7 is not good. Bf4 or Be3 might still hold on.

35. Kc3 h6

Bh6 might still hold out, but blocking h6 with your own pawn is
unfortunate.

36. b3 c4

Again the program seems to have a death wish. This invites more trouble.
Only b4+ is really playable here and it is still pretty dire. But it
might be enough to stop another weak endgame player from winning.

41. Kxc4 Bg1

Bb6 would slow things down but defeat is now inevitable.

42. a7 Bxa7 43. Nxa7 Kh8

This is a surprise. My copy of Crafty saw Nd4 at 5 ply search presumably
through selective extensions. The principal line it had was:

42. Nd4 Bxh2 43. a7 Bxe5 44. Nxe6 h5 45. a8Q

Seems to me like Kg6 or Bg3 is needed to prolong the agony.
But Crafty here saw the right best move (v19.19) of Nd4.

Which version of Crafty were you running?

The challenge now is to find a win with a 4 ply search!

Regards,
--
Martin Brown

--
Posted via a free Usenet account from http://www.teranews.com

Sanny writes:

On Feb 3, 4:15Â*pm, Peter Osterlund wrote:
Martin Brown writes:
If the application you are running Crafty under allows nominal fixed
depth searches it might be interesting to try 8ply or 9ply and then
make the search gradually shallower until Beginner can just win.

I set the search depth in crafty to 5 ply, which reduces its thinking
time to about 0.01 seconds. A game againstGetClubbeginner level was
much more even than before, but crafty still managed to win in the end
game, mostly becauseGetClubplayed very weakly.

(The opening book I'm using for crafty seems to contain some really
bad lines. Crafty evaluated the position to -1.5 the first move after
the opening book, which was move 6. Ne5.)

[Event "Computerchessgame"]
[Site "quad.localdomain"]
[Date "2008.02.03"]
[Round "-"]
[White "Crafty-20.14 5 ply search depth"]
[Black "GetClubbeginner level"]
[Result "1-0"]
[TimeControl "60/120"]

1. e4 e5 2. f4 exf4 3. Nf3 g5 4. Bc4 Nc6 5. d4 g4 6. Ne5 Nxe5 7. dxe5 Qh4+
8. Kd2 f3 9. gxf3 Qf2+ 10. Kc3 Qxf3+ 11. Qxf3 gxf3 12. Rf1 Bg7 13. Bf4 d6
14. Rxf3 dxe5 15. Bg3 Be6 16. Bxe6 fxe6 17. Nd2 Nf6 18. Raf1 O-O 19. a3
Nd5+ 20. Kb3 Rxf3+ 21. Rxf3 Nf4 22. Bxf4 exf4 23. Rxf4 c6 24. Rf1 Rf8 25.
Rd1 Kh8 26. Nc4 b5 27. Na5 c5 28. Rd7 a6 29. c4 Rf2 30. Rd8+ Rf8 31. Rxf8+
Bxf8 32. e5 Bh6 33. cxb5 axb5 34. Nc6 Bg7 35. Kc3 h6 36. b3 c4 37. bxc4
bxc4 38. a4 Bf8 39. a5 Kh7 40. a6 Bc5 41. Kxc4 Bg1 42. a7 Bxa7 43. Nxa7 Kh8
44. Nc6 Kh7 45. Nd4 Kg6 46. Nxe6 h5 47. Nf4+ Kg5 48. Nxh5 Kxh5 49. Kd5 Kg5
50. e6 Kf6 51. Kd6 Kg6 52. h4 Kf6 53. e7 Kf7 54. Kd7 Kg8 55. e8=Q+ Kh7 56.
Ke6
{Black resigns} 1-0

--
Peter Osterlund - /~u89404340

It looks like Crafty keeps running when GetClub is thinking. And it is
not allowing GetClub Program to take enough resources.

It wasn't. In that match crafty used in total 1 second of CPU time
(about 10ms per move.) And all of that time was when GetClub was not
thinking.

I suggest use Crafty on some other computer and then test so that
Crafty do not spoil GetClubs Calculations.

On my Computer GetClubs Beginner Level thinks 5-6 Depth deep. How much
is shown on your Computer.

Two things need toi be done play both on Single Core computer with
Thinking in Opponents move not allowed Else Crafty gets 15 times more
processing power.

I already did this match were GetClub gets about 1000 times more
computation time. (10s versus .01s.)

--
Peter Osterlund -
http://web.telia.com/~u89404340

On my Computer GetClubs Beginner Level thinks 5-6 Depth deep. How much
is shown on your Computer.

Two things need toi be done play both on Single Core computer with
Thinking in Opponents move not allowed Else Crafty gets 15 times more
processing power.

I already did this match wereGetClubgets about 1000 times more
computation time. (10s versus .01s.)

What was the depth shown by Beginner Level was it 5/6.

It shows depth of thinking like "Thinking on Bishop(5)" Meaning It is
thinking at 5 depth.

On my Computer it plays with 5-6 depth What was the depth on your
Computer?

Bye
Sanny

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

Sanny writes:

On my Computer GetClubs Beginner Level thinks 5-6 Depth deep. How much
is shown on your Computer.

Two things need toi be done play both on Single Core computer with
Thinking in Opponents move not allowed Else Crafty gets 15 times more
processing power.

I already did this match wereGetClubgets about 1000 times more
computation time. (10s versus .01s.)

What was the depth shown by Beginner Level was it 5/6.

It shows depth of thinking like "Thinking on Bishop(5)" Meaning It is
thinking at 5 depth.

On my Computer it plays with 5-6 depth What was the depth on your
Computer?

Unfortunately I don't know. I was running GetClub chess in firefox on
linux and for some reason the font size gets so big that the text
doesn't fit in the box.

I'll play another game where I let GetClub chess play on my laptop in
windows XP instead (3.07 GHz Pentium 4 CPU), and report what happens.

--
Peter Osterlund -
http://web.telia.com/~u89404340