| |
Main
Date: 27 Jul 2008 04:49:19
From: Sanny
Subject: GetClub improved by 30% today.
|
GetClub will play 30% stronger moves. So it will be a lot of challenge
to win the Beginner Level.
Now Beginner Level will not be as easy as it used to earlier. A lot of
tough competition at GetClub.
Heres a game against Jester with GetClub Beginner Level. Game was
equal till 40 moves but later GetClub lost a Rook for Knight and pawn.
Game Played between sanjay11 and beginner at GetClub.com
--------------------------------------------------------------------------------
sanjay11: (White)
beginner: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM23521&game=Chess
--------------------------------------------------------------------------------
White -- Black
(sanjay11) -- (beginner)
1. e2-e4{2} d7-d6{0}
2. d2-d4{6} Ng8-f6{0}
3. Nb1-c3{8} g7-g6{0}
4. f2-f3{6} Nb8-c6{8}
5. Bf1-b5{10} Bc8-d7{20}
6. Bc1-g5{10} Bf8-g7{12}
7. Ng1-e2{14} Ke8-g8{14}
8. Ke1-g1{16} e7-e5{24}
9. d4-d5{18} Nc6-e7{18}
10. Bb5-d7{12} Qd8-d7{8}
11. Qd1-d2{12} c7-c5{36}
12. Bg5-h6{16} Bg7-h6{10}
13. Qd2-h6{10} Kg8-h8{6}
14. Qh6-h4{14} Nf6-h5{18}
15. g2-g4{16} Nh5-g7{10}
16. Ne2-g3{14} f7-f5{24}
17. g4-f5{12} g6-f5{10}
18. Ra1-e1{12} f5-f4{8}
19. Ng3-e2{14} Ne7-g6{34}
20. Qh4-h6{10} Rf8-f7{16}
21. Kg1-h1{12} Kh8-g8{38}
22. Ne2-g1{10} Ra8-f8{24}
23. Ng1-h3{12} Ng7-e8{32}
24. Nh3-g5{10} Rf7-f6{12}
25. Ng5-e6{16} Rf8-f7{6}
26. Rf1-g1{14} Ne8-c7{22}
27. Ne6-c7{10} Qd7-c7{6}
28. Rg1-g5{20} Qc7-c8{0}
29. Nc3-b5{12} c5-c4{12}
30. Re1-f1{24} Qc8-c5{14}
31. Nb5-c3{12} Qc5-e3{8}
32. Qh6-h3{14} Rf7-c7{6}
33. Nc3-b5{24} Rc7-c5{10}
34. Qh3-d7{16} Qe3-e2{22}
35. Rg5-g1{12} Qe2-c2{50}
36. Qd7-d8{12} Rf6-f8{6}
37. Qd8-d7{10} Rf8-f7{32}
38. Qd7-e6{12} Rc5-b5{10}
39. Qe6-e8{16} Kg8-g7{18}
40. Qe8-b5{16} Kg7-f8{8}
41. Qb5-a5{24} Rf7-d7{14}
42. Qa5-a7{58} Kf8-g7{0}
43. Rf1-f2{14} Qc2-d3{14}
44. Qa7-b8{12} Kg7-h6{8}
45. Qb8-c8{12} Qd3-e3{20}
46. Rf2-g2{18} Rd7-e7{24}
47. Qc8-f5{14} Kh6-g7{6}
48. Rg2-g6{10} h7-g6{0}
--------------------------------------------------------------------------------
sanjay11: (White)
beginner: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM23521&game=Chess
Are you seeing that GetClub playing much stronger in this game?
So play a game with Beginner Level and it will give you a good
Challenge.
Help Bot play a few more games and tell me how you find it now?
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
|
| |
Date: 28 Jul 2008 20:08:09
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 10:39=A0pm, Patrick Volk <[email protected] > wrote:
> >The improvement is judged by playing GetClub against other commercial
> >programs.
>
> >By looking at number of moves needed to win a game by stronger
> >opponent I can judge the improvement.
>
> Umm, not how it works.
Take the following two scenarios:
1. In a freak occurrence, a strong player
hangs a piece in the first ten moves, then
reverts to playing at his "normal" level.
2. In a routine occurrence, a weak player
hangs a piece, then plays a bit stronger
for the rest of the game.
In scenario (1), the game might last for
a very long time. In scenario (2), probably
not. Given a reasonable sample size, the
strong player will generally win quicker,
the weak player, much more slowly. (This
assumes little or no difference in style.)
I can still recall the time when I was often,
though not always, miniaturizing Sanny's
program, some time back. It was
particularly pleasing to score a mini versus
one of the higher levels back then, but now
this is a rarity. In fact, most of my games
go thirty moves or more. I see this not as
a mere coincidence, but as confirmation of
the program's very real improvement (not
to be confused with Sanny's ridiculous
claims).
Now, testing against Rybka is far from
ideal in that there is too great a gap in
playing strength. In this scenario, lasting
longer might only indicate that GC is now
playing a bit less aggressively in the early
part of the game, less hari kiri style chess
than before.
But look at the tail end of games: the
program used to just allow itself to be
checkmated, but now it sees checkmate
coming and "resigns", often well in
advance. That's a real improvement,
even if it still loses the game. Indeed, I
am reminded of a game against a certain
grandmaster, in which I resigned
because the endgame was hopeless; it
could not possibly be held. Weaker
players complained that I ought to have
played on a few more moves, to get to
where this hopelessness was evident to
them (and I can only note the titanic gap
in chess understanding between us).
Sanny's multitude of claims are best
taken with a half-kilo of sodium chloride
(or better still, potassium chloride). But
there has been considerable real
improvement over a considerable length
of time.
-- help bot
|
| |
Date: 28 Jul 2008 16:16:56
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 6:05=A0pm, Kenneth Sloan <[email protected] > wrote:
> >> PI * i
>
> > =A0 Um, it looks like you need to turn off your
> > shift key or something; all those characters
> > came out as letters or special characters,
> > instead of /numbers/. =A0;>D
> You do know the difference between "number" and "numeral", don't you?
The small i, I believe, represents some
imaginary number, as might be used in
calculus (or hardened tartar). As for the
pea and the el, I can only guess.
A numeral is one of these: 0, 1, 2, 3, 4,
um... 5, 6, 7, 8, 9 which we stole from
the Arabs (long before we knew about
their light, sweet crude). It could also
be one of these: I, II, III, IV, V, VI, VII, VIII,
IX, X, XI, XII, XIII, XIV, XV, XVI, XVII, etc.
I certainly can't expect anyone to track
down every one of Sanny's claims, but
that is quite unnecessary in order to
compute a /rate/ of improvement.
Some jokers assume the rate must be
near-infinite, but I believe the actual
number is more like a doubling in speed
(which Sanny refers to as strength)
about once a week, lately. Anyway, my
most recent game involved a couple of
small errors on my part and I had to win
in the endgame (blush), down two pawns
with two Rooks apiece. Don't try that at
home, kids-- I am a trained professional.
-- help bot
|
| |
Date: 28 Jul 2008 12:17:52
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 6:04=A0am, Sanny <[email protected] > wrote:
> > =A0 But over the course of time, there
> > has been a gradual, but very real
> > improvement.
>
> Yes its now very difficult to improve the game further. This month
> only 30% improvement was done.
>
> > > The game is 30% stronger
You know, there is a new version of Rybka
coming out, and at first blush, the claim was
that it will be about a hundred points stronger
than the current version. But on closer
inspection, this number was qualified, then
downgraded yet again, until it is not entirely
clear whether the new version will show any
substantial improvement against lowly
humans (or ordinary chess engines).
The problem is a conflict of interest, in that
they have to say it is a big improvement, or
risk poor sales. But then they don't want
negative reactions from testers, so they try
to /have it both ways/.
> It will play in same time but with a 30% better move.
Better is a qualitative term, while 30% is
quantitative; you probably mean that the
search has been sped up by 30%.
I had one recent game where I was about
to promote a pawn and the GC program
resigned, apparently foreseeing a forced
checkmate. The thing is, my King was on
the opposite side of the board, so that
would mean the program saw the pawn
Queening, its shift to the other side, an
approach by the Queen, a move to get my
Knight in position, and finally, the mating
move-- not bad. (But note who it was
doing the checkmating here).
> > =A0 The program often falters in being
> > overly materialistic, while neglecting
> > development of its pieces and/or
> > King safety. =A0Not too long ago, I had
>
> Yesterdays improvement have corrected this problem.
Wow. So then, after years of waiting,
we finally now have a GC program which
is eager (but not too eager) to develop
all its pieces and castle? Amazing.
> Lets play a few
> games and see if it plays much stronger moves.
Why bother? If, as you say, all these
"improvements" have fixed the many
weaknesses, I will have no chance
against such a perfect machine. Just
one slip and I'm toast.
> With the new improvement you will have to think longer. I think you
> will have to think 50 sec / Move to win the Beginner Level. And
> Beginner Level will think in your time so most of the moives will be
> made in "0" seconds.
>
> As in 50 seconds getclub can think on your time and play the best
> move. So you are going to see lots of moves made in "0" seconds when
> playing with Beginner Level.
It will presumably, perfectly anticipate
my blunders, having its reply ready?
Incredible.
> Whats the configuration of your Computer? How old it is?
I have an AMD Turion64 with, oh, maybe
1 gigabyte of memory. But my OS is 32-
bit.
> =A0I have a old Pentium 3.2 GHz. It works better for all jobs and I do
> not feel upgrading it anytime. As all program works in just 1 second.
With me, the bottleneck is clearly the
internet connection, but a lot of junk gets
loaded along with Vista. I can still recall
a time when the top chess programs
would bog down somewhere around 7 or
8 plies. Then along came Chess Genius,
which rushed past them all, making it to
about 9 plies before slowing to a crawl.
So, when I see modern chess engines
hitting search ply depths way up in the
twenties, I am not exactly burning with
desire for even more speed. The one
area where a ridiculous amount of speed
would come in handy though is in the
analysis of an entire game; this now
takes quite a while, unless I just want to
check for crude blunders or quickly draw
a graph of the game's progress to get a
very general idea.
-- help bot
|
| |
Date: 28 Jul 2008 05:42:34
From: Sanny
Subject: Interesting game with Rybka
|
An interesting game with Rybka. Rybka took GetClubs Rook or may be it
was a sacrifice?
As game went equal afterwards. But Bishop got pinned by Rook and then
Rybka took the Bishop
Game Played between Rybka and easy at GetClub.com
--------------------------------------------------------------------------------
Rybka: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM23611&game=Chess
--------------------------------------------------------------------------------
White -- Black
(Rybka) -- (easy)
1. e2-e4{50} c7-c5{0}
2. Ng1-f3{14} Nb8-c6{0}
3. d2-d4{12} c5-d4{0}
4. Nf3-d4{12} e7-e6{0}
5. Nb1-c3{16} Bf8-d6{264}
6. Nd4-b5{18} Bd6-e5{244}
7. Nb5-d6{16} Be5-d6{20}
8. Qd1-d6{12} Qd8-e7{60}
9. Bc1-f4{16} e6-e5{36}
10. Qd6-e7{16} Ng8-e7{52}
11. Bf4-e3{24} Nc6-d4{20}
12. Be3-d4{22} e5-d4{30}
13. Nc3-b5{38} Ke8-g8{32}
14. Ke1-c1{16} d7-d5{94}
15. Nb5-c7{20} d5-e4{30}
16. Nc7-a8{20} d4-d3{104}
17. Na8-c7{16} Bc8-f5{40}
18. Nc7-b5{16} Rf8-c8{92}
19. Rd1-d2{22} Rc8-c5{90}
20. Nb5-c3{16} Ne7-d5{28}
21. c2-d3{14} Nd5-c3{32}
22. b2-c3{12} e4-d3{20}
23. Bf1-d3{16} Rc5-c3{24}
24. Kc1-b2{14} Rc3-d3{60}
25. Rd2-d3{16} Bf5-d3{58}
26. Rh1-d1{18} a7-a6{98}
27. Rd1-d3{44} h7-h5{104}
28. Rd3-d8{18} Kg8-h7{0}
29. Rd8-d7{20} b7-b6{24}
30. Rd7-f7{14} Kh7-h6{28}
31. Rf7-a7{16} b6-b5{90}
32. Ra7-a6{14} Kh6-h7{108}
33. Ra6-a5{26} Kh7-g8{82}
34. Ra5-b5{26} g7-g6{44}
35. a2-a4{26} Kg8-g7{24}
36. Rb5-h5{20} Kg7-f6{28}
37. a4-a5{24} g6-g5{30}
38. a5-a6{20} Kf6-e6{50}
39. a6-a7{18} g5-g4{50}
40. Qa7-a8{Q}{22} g4-g3{54}
41. Qa8-c6{30} Ke6-f7{0}
42. Rh5-h7{44} Kf7-g8{0}
--------------------------------------------------------------------------------
Rybka: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM23611&game=Chess
Lots of Ups and down in this game.
Please let me know of wrong moves made by GetClub.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| | |
Date:
From: Martin Brown
Subject: Re: Interesting game with Rybka
|
|
| |
Date: 28 Jul 2008 03:10:06
From: Sanny
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 2:12=A0pm, Peter Osterlund <[email protected] > wrote:
> Sanny <[email protected]> writes:
> > This month only 30% improvement was possible. As all the ideas of
> > improvements have exhausted.
>
> There are many things that can be done to improve the program.
>
> * Read "Thirty Ways to Improve the Performance of Your Java
> =A0 Programs". Available here:
>
> =A0 =A0 =A0 =A0 =A0 =A0 =A0http://www.glenmccl.com/jperf/
>
> =A0 Pay special attention to section 3.2 (making methods final to avoid
> =A0 virtual function calls) and section 4.1 (strings are immutable).
>
> * Improve the algorithm, for example by implementing alpha-beta search
> =A0 with good move ordering. Right now, comparinggetclubwith a strong
> =A0 computerchessprogram is like comparing quick sort and bubble
> =A0 sort. Quick sort is not just a constant factor faster than bubble
> =A0 sort. The speed difference gets bigger the larger the problem gets.
>
> =A0 The same is true forGetClubchessvs strongerchessprograms. =A0For
> =A0 example, assume that glaurung using 30 milliseconds per move has the
> =A0 same strength asGetClubusing 15 seconds per move. In that case, if
> =A0 you give both programs 10 times more thinking time, you would
> =A0 probably find thatGetClubat 150 seconds per move would lose badly
> =A0 against glaurung at 300 milliseconds per move.
>
> =A0 The reason is that glaurung is using more advanced algorithms, like
> =A0 well tuned alpha-beta pruning and late move reductions, so its
> =A0 branching factor is much lower than GetClubs branching factor.
>
> --
> Peter Osterlund - [email protected]://web.telia.com/~u89404340
These all things goes over my head. I think the Java improvement will
work.
Lets see if that increases the speed again by 10-12%
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| |
Date: 28 Jul 2008 03:04:56
From: Sanny
Subject: Re: GetClub improved by 30% today.
|
> > I think that was played yesterday
> > before the improvement?
>
> =A0 No, I "improved" a half-dozen times
> between then and now, so no matter
> which ofGetClub'sinnumerable
> "improvements" you are talking about,
> my own are just as numerous and
> twice as tasty! =A0;>D
Lets see how you fare Beginner & Easy Levels.
> > Play a game today with Beginner Level and tell me how strong it feel.
>
> =A0 It feels about the same every time
> you announce yet another nutter
> butter peanut butter "improvement".
>
> =A0 But over the course of time, there
> has been a gradual, but very real
> improvement.
Yes its now very difficult to improve the game further. This month
only 30% improvement was done.
> > The game is 30% stronger So you have to think a bit longer to win.
>
> =A0 If I think longer, I may not come up
> with anything more than a headache.
It will play in same time but with a 30% better move.
> =A0 As in the game I described above,
> my advantage lies in a better
> understanding ofchess, and the fact
> that theGetClubprogram has yet to
> fully master simple tactics. =A0With no
> substantial book knowledge, we often
> arrive in /terra incognita/ early on, as
> in the game described at top. =A0I then
> rely upon an attempt to calculate
> variations, but also on judgment and
> general principles.
Yes I saw that game. It was 16 depth calculation and Easy level sees
just 10-12 depth ahead. And it lost a Bishop for a pawn. I think with
yesterdays improvement you may find better moves.
> =A0 The program often falters in being
> overly materialistic, while neglecting
> development of its pieces and/or
> King safety. =A0Not too long ago, I had
Yesterdays improvement have corrected this problem. Lets play a few
games and see if it plays much stronger moves.
> =A0 I find these games reasonably
> enjoyable, and it is especially
> refreshing to not have to deal with a
> book-aping opponent who moves
> by-rote without thinking for himself.
With the new improvement you will have to think longer. I think you
will have to think 50 sec / Move to win the Beginner Level. And
Beginner Level will think in your time so most of the moives will be
made in "0" seconds.
As in 50 seconds getclub can think on your time and play the best
move. So you are going to see lots of moves made in "0" seconds when
playing with Beginner Level.
> The fact is, the peculiarities of the
> program force me to think right from
> the get-go, and I enjoy being able to
> actually out-calculate a Pentium
> class machine.
Whats the configuration of your Computer? How old it is?
I have a old Pentium 3.2 GHz. It works better for all jobs and I do
not feel upgrading it anytime. As all program works in just 1 second.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| |
Date: 28 Jul 2008 01:55:03
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 4:19=A0am, Sanny <[email protected] > wrote:
> > =A0The games are getting more and more
> > interesting over time, with the program
> > playing lines that I have to consider as
> > potential refutations to my moves. =A0For
> > instance, in my most recent game I had
> > a nice combination (a4, ...Ba6, ab5,
> > ...Bxb5, Bxc4, ...Bxc4, Qa4) which won
> > on positional grounds; but I overlooked
> > ...Bd5! completely, and had to mix it up
>
> Which game you are talking about.
Well, there is only one game of mine
wherein those moves were played:
http://www.getclub.com/playgame.php?id=3DDM23350&game=3DChess
> I think that was played yesterday
> before the improvement?
No, I "improved" a half-dozen times
between then and now, so no matter
which of GetClub's innumerable
"improvements" you are talking about,
my own are just as numerous and
twice as tasty! ; >D
> Play a game topday with Beginner Level and tell me how strong it feel.
It feels about the same every time
you announce yet another nutter
butter peanut butter "improvement".
But over the course of time, there
has been a gradual, but very real
improvement.
> The game is 30% stronger So you have to think a bit longer to win.
If I think longer, I may not come up
with anything more than a headache.
As in the game I described above,
my advantage lies in a better
understanding of chess, and the fact
that the GetClub program has yet to
fully master simple tactics. With no
substantial book knowledge, we often
arrive in /terra incognita/ early on, as
in the game described at top. I then
rely upon an attempt to calculate
variations, but also on judgment and
general principles.
The program often falters in being
overly materialistic, while neglecting
development of its pieces and/or
King safety. Not too long ago, I had
an over-the-board game in which my
human opponent made the same
kind of mistake as the GC program
often does, only in his case it was
a /very simple/ tactical error which
cost him a piece. I had another
human opponent who kept putting
off castling in favor of something
else, turn after turn, and this
ultimately cost him material. Both
of these players were very roughly
1500ish USCF, which is why I
reject the claims of some critics
that the GC program's rating or
strength must be sub-zero, like in
Antarctica.
I find these games reasonably
enjoyable, and it is especially
refreshing to not have to deal with a
book-aping opponent who moves
by-rote without thinking for himself.
The fact is, the peculiarities of the
program force me to think right from
the get-go, and I enjoy being able to
actually out-calculate a Pentium
class machine.
-- help bot
|
| |
Date: 28 Jul 2008 01:27:49
From: Sanny
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 1:12=A0pm, Tobias Heidelmann <[email protected]
dortmund.de > wrote:
> Sanny schrieb:
>
> > GetClub will play 30% stronger moves. So it will be a lot of challenge
> > to win the Beginner Level.
>
> How so you measure improvement? Why 30%?
Earlier the game speed used to double every week. But this week only
30% improvement was achieved.
The improvement is judged by playing GetClub against other commercial
programs.
By looking at number of moves needed to win a game by stronger
opponent I can judge the improvement.
Here is a case.
Initially Beginner used to loose in 20 moves. After 2 times
improvement Beginner was able to defend itself to 25 moves.
Easy Level which thinks 4 times used to loose in 30 moves.
So when Beginner started loosing in 30 moves It means now beginner is
4 times stronger than before.
A week back Beginner plays till 45 moves. An increase of +15 moves
means Twice improvement.
Now Today Beginner Plays till 50 moves against Jester. +5 moves extra.
Since number of moves increased 5 extra moves comes out to 30%
improvement. As +15 moves means 2 times improvement (1/3 =3D 30%)
Last 2 months the game was improved 5-10 times. But now the rate of
improvement has reduced.
This month only 30% improvement was possible. As all the ideas of
improvements have exhausted.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| | |
Date: 28 Jul 2008 22:39:58
From: Patrick Volk
Subject: Re: GetClub improved by 30% today.
|
On Mon, 28 Jul 2008 01:27:49 -0700 (PDT), Sanny
<[email protected] > wrote:
>On Jul 28, 1:12�pm, Tobias Heidelmann <[email protected]
>dortmund.de> wrote:
>> Sanny schrieb:
>>
>> > GetClub will play 30% stronger moves. So it will be a lot of challenge
>> > to win the Beginner Level.
>>
>> How so you measure improvement? Why 30%?
>
>Earlier the game speed used to double every week. But this week only
>30% improvement was achieved.
>
>The improvement is judged by playing GetClub against other commercial
>programs.
>
>By looking at number of moves needed to win a game by stronger
>opponent I can judge the improvement.
>
Umm, not how it works.
>
>Here is a case.
>
>Initially Beginner used to loose in 20 moves. After 2 times
>improvement Beginner was able to defend itself to 25 moves.
A lot of chess programs are configurable, you know.
>
>Easy Level which thinks 4 times used to loose in 30 moves.
>
>So when Beginner started loosing in 30 moves It means now beginner is
>4 times stronger than before.
You know they generally have Elo ratings for certain levels. The
consensus is that they're inaccurate, but are consistent within a
program (e.g. You'd see something approximating the win rate for an
Elo difference of 100 points).
>
>A week back Beginner plays till 45 moves. An increase of +15 moves
>means Twice improvement.
So, by logic, at the beginning GC would crap the bed after 2 moves?
>
>Now Today Beginner Plays till 50 moves against Jester. +5 moves extra.
>Since number of moves increased 5 extra moves comes out to 30%
>improvement. As +15 moves means 2 times improvement (1/3 = 30%)
50/45 = 1.111... = 11% improvement by your metric.
Must be that new math.
>
>Last 2 months the game was improved 5-10 times. But now the rate of
>improvement has reduced.
>
>This month only 30% improvement was possible. As all the ideas of
>improvements have exhausted.
Considering your numbers, that doesn't surprise me. So if GC plays a
150 move knight chase, we'll see a commercial version?
I know I'm not an expert with chess, but I think I'm familiar... I
honestly did not know that Elo rating were determined by how many
moves... I suppose that's why they give 4 hour clocks (and a bonus
after 40 moves).
>
>Bye
>Sanny
>
>Play Chess at: http://www.GetClub.com/Chess.html
>
>
>
>
|
| | |
Date: 28 Jul 2008 11:12:03
From: Peter Osterlund
Subject: Re: GetClub improved by 30% today.
|
Sanny <[email protected] > writes:
> This month only 30% improvement was possible. As all the ideas of
> improvements have exhausted.
There are many things that can be done to improve the program.
* Read "Thirty Ways to Improve the Performance of Your Java
Programs". Available here:
http://www.glenmccl.com/jperf/
Pay special attention to section 3.2 (making methods final to avoid
virtual function calls) and section 4.1 (strings are immutable).
* Improve the algorithm, for example by implementing alpha-beta search
with good move ordering. Right now, comparing getclub with a strong
computer chess program is like comparing quick sort and bubble
sort. Quick sort is not just a constant factor faster than bubble
sort. The speed difference gets bigger the larger the problem gets.
The same is true for GetClub chess vs stronger chess programs. For
example, assume that glaurung using 30 milliseconds per move has the
same strength as GetClub using 15 seconds per move. In that case, if
you give both programs 10 times more thinking time, you would
probably find that GetClub at 150 seconds per move would lose badly
against glaurung at 300 milliseconds per move.
The reason is that glaurung is using more advanced algorithms, like
well tuned alpha-beta pruning and late move reductions, so its
branching factor is much lower than GetClubs branching factor.
--
Peter Osterlund - [email protected]
http://web.telia.com/~u89404340
|
| |
Date: 28 Jul 2008 01:19:52
From: Sanny
Subject: Re: GetClub improved by 30% today.
|
> =A0The games are getting more and more
> interesting over time, with the program
> playing lines that I have to consider as
> potential refutations to my moves. =A0For
> instance, in my most recent game I had
> a nice combination (a4, ...Ba6, ab5,
> ...Bxb5, Bxc4, ...Bxc4, Qa4) which won
> on positional grounds; but I overlooked
> ...Bd5! completely, and had to mix it up
Which game you are talking about. I think that was played yesterday
before the improvement?
Play a game topday with Beginner Level and tell me how strong it feel.
The game is 30% stronger So you have to think a bit longer to win.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| |
Date: 28 Jul 2008 10:12:14
From: Tobias Heidelmann
Subject: Re: GetClub improved by 30% today.
|
Sanny schrieb:
> GetClub will play 30% stronger moves. So it will be a lot of challenge
> to win the Beginner Level.
How so you measure improvement? Why 30%?
|
| |
Date: 28 Jul 2008 00:56:34
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 28, 3:01=A0am, Kenneth Sloan <[email protected] > wrote:
> > =A0 If there are any advanced mathematicians
> > out there-- can anyone calculate the rate at
> > which the GetClub program is improving?
> PI * i
Um, it looks like you need to turn off your
shift key or something; all those characters
came out as letters or special characters,
instead of /numbers/. ; >D
-- helpless bot
|
| | |
Date: 28 Jul 2008 17:05:46
From: Kenneth Sloan
Subject: Re: GetClub improved by 30% today.
|
help bot wrote:
> On Jul 28, 3:01 am, Kenneth Sloan <[email protected]> wrote:
>
>>> If there are any advanced mathematicians
>>> out there-- can anyone calculate the rate at
>>> which the GetClub program is improving?
>
>> PI * i
>
>
> Um, it looks like you need to turn off your
> shift key or something; all those characters
> came out as letters or special characters,
> instead of /numbers/. ;>D
>
>
> -- helpless bot
>
You do know the difference between "number" and "numeral", don't you?
--
Kenneth Sloan [email protected]
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/
|
| |
Date: 27 Jul 2008 23:24:14
From: help bot
Subject: Re: GetClub improved by 30% today.
|
On Jul 27, 3:33=A0pm, The Historian <[email protected] > wrote:
> What's 30 per cent of the null set?
If there are any advanced mathematicians
out there-- can anyone calculate the rate at
which the GetClub program is improveing?
The reason I need to know is that I need
to keep up a similar rate of improvement, if
I am to have any hope at all of not being
caught from behind, passed, and left eating
Sanny's dust. ; >D
The games are getting more and more
interesting over time, with the program
playing lines that I have to consider as
potential refutations to my moves. For
instance, in my most recent game I had
a nice combination (a4, ...Ba6, ab5,
=2E..Bxb5, Bxc4, ...Bxc4, Qa4) which won
on positional grounds; but I overlooked
=2E..Bd5! completely, and had to mix it up
some more to come out on top. Yet I
cashed in too soon, nearly throwing the
win away.
Personally, I think this is great training
for my games against relatively weak
human opponents, but I must admit that
my higher-rated human opponents will
stick to hackneyed book lines, like fish
to water; I get very little practice in that
area at GetClub, but these games seem
to be more interesting than any book-
theory contests I know of.
-- help bot
|
| | |
Date: 28 Jul 2008 02:01:53
From: Kenneth Sloan
Subject: Re: GetClub improved by 30% today.
|
help bot wrote:
> On Jul 27, 3:33 pm, The Historian <[email protected]> wrote:
>
>> What's 30 per cent of the null set?
>
> If there are any advanced mathematicians
> out there-- can anyone calculate the rate at
> which the GetClub program is improveing?
>
PI * i
--
Kenneth Sloan [email protected]
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/
|
| |
Date: 27 Jul 2008 22:14:39
From: Sanny
Subject: Re: GetClub improved by 30% today.
|
> 1. b3 Nf6 2. Bb2 g6 3. e4 Bg7 4. Bc4 O-O 5. e5 Nh5 6. Qf3 d6 7. Qe4 Nc6 8.
> d4 Nb4 9. exd6 exd6 10. Ne2 d5 11. Bxd5 Nf6 12. Bxb7 Nxe4 13. Bxe4 Re8 14.
> f3 Ba6 15. Nbc3 Nxc2+ 16. Bxc2 Qxd4 17. Be4 Qe3 18. Nb5 Bxb5 19. Bd3 Bxd3
> 20. Kf1 Qxe2+
> {White resigns} 0-1
>
> In this game,GetClubdidn't see deeper than 6 plies (on a 3GHz P4
> computer). It used to claim that it saw deeper in earlier versions, so
> I don't really understand where your "30% stronger moves" comes from.
You are using computer help So that program is 10 times stronger than
GetClub. So you will not be able to see the 30% improvement.
Only those players who play using their mind will be able to see the
difference. It really does not matter much if a program is 10 times
stronger or 8 times stronger. So 30% improvement is not seen when
using computers help.
Bye
Sanny
Play Chess at: http://www.GetClub.com/Chess.html
|
| |
Date: 27 Jul 2008 12:35:03
From: The Historian
Subject: Re: GetClub improved by 30% today.
|
On Jul 27, 8:14 am, SBD <[email protected] > wrote:
> On Jul 27, 6:49 am, Sanny <[email protected]> wrote:
>
> > So play a game with Beginner Level and it will give you a good
> > Challenge.
>
> An infinite number of monkeys at an infinite number of keyboards have
> a better chance of duplicating the output of Alekhine's Parrot....
> wait, that only takes one monkey, never mind.
That reminds me of a noted historian's comment on the Alekhine's
Parrot author's criticisms of Lawson's Morphy book: "A monkey sh*tting
on a marble table."
|
| |
Date: 27 Jul 2008 12:33:00
From: The Historian
Subject: Re: GetClub improved by 30% today.
|
On Jul 27, 6:49 am, Sanny <[email protected] > wrote:
> GetClub will play 30% stronger moves. So it will be a lot of challenge
> to win the Beginner Level.
>
> Now Beginner Level will not be as easy as it used to earlier. A lot of
> tough competition at GetClub.
>
> Heres a game against Jester with GetClub Beginner Level. Game was
> equal till 40 moves but later GetClub lost a Rook for Knight and pawn.
>
> Game Played between sanjay11 and beginner at GetClub.com
>
> --------------------------------------------------------------------------------
> sanjay11: (White)
> beginner: (Black)
> Game Played at:http://www.GetClub.com/Chess.html
> View Recorded Game:http://www.getclub.com/playgame.php?id=DM23521&game=Chess
> --------------------------------------------------------------------------------
>
> White -- Black
> (sanjay11) -- (beginner)
>
> 1. e2-e4{2} d7-d6{0}
> 2. d2-d4{6} Ng8-f6{0}
> 3. Nb1-c3{8} g7-g6{0}
> 4. f2-f3{6} Nb8-c6{8}
> 5. Bf1-b5{10} Bc8-d7{20}
> 6. Bc1-g5{10} Bf8-g7{12}
> 7. Ng1-e2{14} Ke8-g8{14}
> 8. Ke1-g1{16} e7-e5{24}
> 9. d4-d5{18} Nc6-e7{18}
> 10. Bb5-d7{12} Qd8-d7{8}
> 11. Qd1-d2{12} c7-c5{36}
> 12. Bg5-h6{16} Bg7-h6{10}
> 13. Qd2-h6{10} Kg8-h8{6}
> 14. Qh6-h4{14} Nf6-h5{18}
> 15. g2-g4{16} Nh5-g7{10}
> 16. Ne2-g3{14} f7-f5{24}
> 17. g4-f5{12} g6-f5{10}
> 18. Ra1-e1{12} f5-f4{8}
> 19. Ng3-e2{14} Ne7-g6{34}
> 20. Qh4-h6{10} Rf8-f7{16}
> 21. Kg1-h1{12} Kh8-g8{38}
> 22. Ne2-g1{10} Ra8-f8{24}
> 23. Ng1-h3{12} Ng7-e8{32}
> 24. Nh3-g5{10} Rf7-f6{12}
> 25. Ng5-e6{16} Rf8-f7{6}
> 26. Rf1-g1{14} Ne8-c7{22}
> 27. Ne6-c7{10} Qd7-c7{6}
> 28. Rg1-g5{20} Qc7-c8{0}
> 29. Nc3-b5{12} c5-c4{12}
> 30. Re1-f1{24} Qc8-c5{14}
> 31. Nb5-c3{12} Qc5-e3{8}
> 32. Qh6-h3{14} Rf7-c7{6}
> 33. Nc3-b5{24} Rc7-c5{10}
> 34. Qh3-d7{16} Qe3-e2{22}
> 35. Rg5-g1{12} Qe2-c2{50}
> 36. Qd7-d8{12} Rf6-f8{6}
> 37. Qd8-d7{10} Rf8-f7{32}
> 38. Qd7-e6{12} Rc5-b5{10}
> 39. Qe6-e8{16} Kg8-g7{18}
> 40. Qe8-b5{16} Kg7-f8{8}
> 41. Qb5-a5{24} Rf7-d7{14}
> 42. Qa5-a7{58} Kf8-g7{0}
> 43. Rf1-f2{14} Qc2-d3{14}
> 44. Qa7-b8{12} Kg7-h6{8}
> 45. Qb8-c8{12} Qd3-e3{20}
> 46. Rf2-g2{18} Rd7-e7{24}
> 47. Qc8-f5{14} Kh6-g7{6}
> 48. Rg2-g6{10} h7-g6{0}
> --------------------------------------------------------------------------------
> sanjay11: (White)
> beginner: (Black)
> Game Played at:http://www.GetClub.com/Chess.html
> View Recorded Game:http://www.getclub.com/playgame.php?id=DM23521&game=Chess
>
> Are you seeing that GetClub playing much stronger in this game?
>
> So play a game with Beginner Level and it will give you a good
> Challenge.
>
> Help Bot play a few more games and tell me how you find it now?
>
> Bye
> Sanny
>
> Play Chess at:http://www.GetClub.com/Chess.html
What's 30 per cent of the null set?
|
| |
Date: 27 Jul 2008 17:57:43
From: Peter Osterlund
Subject: Re: GetClub improved by 30% today.
|
Sanny <[email protected] > writes:
> GetClub will play 30% stronger moves. So it will be a lot of challenge
> to win the Beginner Level.
...
> So play a game with Beginner Level and it will give you a good
> Challenge.
Here is one game:
[Date "2008.07.27"]
[White "GetClub beginner level"]
[Black "Glaurung 2.1"]
[Result "0-1"]
1. b3 Nf6 2. Bb2 g6 3. e4 Bg7 4. Bc4 O-O 5. e5 Nh5 6. Qf3 d6 7. Qe4 Nc6 8.
d4 Nb4 9. exd6 exd6 10. Ne2 d5 11. Bxd5 Nf6 12. Bxb7 Nxe4 13. Bxe4 Re8 14.
f3 Ba6 15. Nbc3 Nxc2+ 16. Bxc2 Qxd4 17. Be4 Qe3 18. Nb5 Bxb5 19. Bd3 Bxd3
20. Kf1 Qxe2+
{White resigns} 0-1
In this game, GetClub didn't see deeper than 6 plies (on a 3GHz P4
computer). It used to claim that it saw deeper in earlier versions, so
I don't really understand where your "30% stronger moves" comes from.
--
Peter Osterlund - [email protected]
http://web.telia.com/~u89404340
|
| |
Date: 27 Jul 2008 06:14:16
From: SBD
Subject: Re: GetClub improved by 30% today.
|
On Jul 27, 6:49 am, Sanny <[email protected] > wrote:
> So play a game with Beginner Level and it will give you a good
> Challenge.
An infinite number of monkeys at an infinite number of keyboards have
a better chance of duplicating the output of Alekhine's Parrot....
wait, that only takes one monkey, never mind.
|
|
