Main
Date: 16 Aug 2008 04:10:14
From: Sanny
Subject: GetClub Twice Stronger again !!!
Here is Strong game played by GetClub against Rybka after the
improvement done.

Can you spot mistakes in this game by GetClub?

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=DM24853&game=Chess
--------------------------------------------------------------------------------

White -- Black
(Rybka) -- (easy)

1. d2-d4{16} Ng8-f6{0}
2. c2-c4{8} e7-e6{0}
3. Ng1-f3{12} Nf6-e4{20}
4. Qd1-d3{16} d7-d5{22}
5. c4-d5{10} e6-d5{30}
6. Nb1-d2{44} Bf8-b4{62}
7. a2-a3{16} Bb4-d2{40}
8. Nf3-d2{12} f7-f5{24}
9. Nd2-e4{14} f5-e4{20}
10. Qd3-g3{12} Ke8-g8{24}
11. Bc1-f4{14} Nb8-c6{48}
12. Bf4-c7{18} Qd8-f6{42}
13. e2-e3{16} Bc8-g4{96}
14. Bf1-b5{32} Ra8-c8{44}
15. Bc7-d6{12} Rf8-f7{118}
16. Ke1-g1{12} Qf6-f5{46}
17. h2-h3{14} Bg4-h5{24}
18. Ra1-c1{20} Bh5-g6{96}
19. Rc1-c5{22} h7-h6{66}
20. Rf1-c1{14} Bg6-h7{112}
21. Bb5-c4{18} Nc6-d4{24}
22. e3-d4{32} Rc8-c5{22}
23. Bd6-c5{14} Kg8-h8{62}
24. Bc4-e2{36} a7-a6{40}
25. Bc5-d6{16} b7-b6{88}
26. Be2-a6{16} Rf7-f6{296}
27. Rc1-c7{16} Rf6-g6{222}
28. Qg3-e5{16} Qf5-g5{96}
29. Qe5-g5{508} Rg6-g5{26}
30. Bd6-e7{14} Rg5-g6{20}
31. Ba6-e2{28} Bh7-g8{26}
32. Be2-h5{14} Rg6-e6{20}
33. Be7-f8{12} Kh8-h7{42}
34. Rc7-g7{36} Kh7-h8{0}
35. Rg7-c7{16} Bg8-h7{26}
36. Bf8-g7{14} Kh8-g8{0}
37. Bh5-f7{10} Kg8-g7{0}
38. Bf7-e6{12} Kg7-f6{42}
39. Rc7-h7{14} Kf6-e6{28}
40. Rh7-h6{12} Ke6-d7{92}
41. Rh6-b6{16} Kd7-c7{76}
42. a3-a4{44} Kc7-b6{64}
43. b2-b4{14} Kb6-c6{142}
44. h3-h4{12} Kc6-d7{322}
45. h4-h5{14} Kd7-e7{62}
46. a4-a5{12} Ke7-f6{144}
47. Kg1-f1{36} Kf6-f7{20}
48. a5-a6{298} Kf7-e6{0}
49. h5-h6{140} Ke6-d7{94}
50. h6-h7{122} Kd7-c6{106}
51. Kf1-e2{20} Kc6-b6{104}
52. Ke2-e3{26} Kb6-a6{28}
53. Ke3-f4{18} e4-e3{86}
54. f2-e3{186} Ka6-b6{96}
55. Kf4-e5{60} Kb6-c6{146}
56. b4-b5{86} Kc6-b7{130}
57. Ke5-d5{20} Kb7-b6{84}
58. Qh7-h8{Q}{18} Kb6-b5{52}
--------------------------------------------------------------------------------
Rybka: (White)
easy: (Black)
Game Played at: http://www.GetClub.com/Chess.html
View Recorded Game: http://www.getclub.com/playgame.php?id=DM24853&game=Chess

So what do you say how good Easy Level played against Rybka?

Bye
Sanny

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




 
Date: 18 Aug 2008 09:35:28
From: Sanny
Subject: Re: GetClub Twice Stronger again !!!
> And the REALLY funny thing (ok, well, ONE of the MANY really funny
> things) is that I have counted at least 20 "twice stronger" posts from
> Sanny. Assuming that he actually means twice faster, and that it
> really WAS twice faster (which is, of course, impossible), thatGetClubwould now find ALL moves in one second that it used to need
> almost TWO WEEKS to find.

I remember when 2 years Back Taylor Kingston used to play at GetClub
it used to play moves in 1 hour. Now it plays the same move in just 1
second.

Baby level plays each move in just 2 seconds. While those days Master
Level using 1 hour analysis used to play wrong moves.

So really that much improvement has been done.


> Pretty friggin' hilarious.... Reading Sanny's posts is always better
> than reading the comics. :-)

Play Chess at GetClub that will be more of fun. Play 1 game daily at
GetClub and you will enjoy a lot.

Bye
Sanny

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



 
Date: 18 Aug 2008 08:30:13
From:
Subject: Re: GetClub Twice Stronger again !!!
> >Because now it plays the same move in half time. Say it used to see 8
> >depth in 20 secs. Now after improvement it sees the 8 depth in just 10
> >sec. So that is Twice improvement.
>
> That's not twice *stronger*
>
> That would be twice *faster* PROVIDED it's actually faster (ie: as if you
> moved to a faster computer.)
>
> Simply finding the same move in 10 seconds that used to take 20 seconds i=
s
> not necessarily the same as being stronger. =A0It just means you found a
> particular move faster than you used to. =A0That's not twice "stronger".

And the REALLY funny thing (ok, well, ONE of the MANY really funny
things) is that I have counted at least 20 "twice stronger" posts from
Sanny. Assuming that he actually means twice faster, and that it
really WAS twice faster (which is, of course, impossible), that
GetClub would now find ALL moves in one second that it used to need
almost TWO WEEKS to find.

Pretty friggin' hilarious.... Reading Sanny's posts is always better
than reading the comics. :-)

jm


 
Date: 17 Aug 2008 22:11:38
From: Sanny
Subject: Re: GetClub Twice Stronger again !!!
> ?> ELO ratings? Not bloody likely. At that rate, GC would be around 30,00=
0
> ?> elo by now.
>
> >According to Rybka everytime a playes becomes twice stronger his elo
> >increases by +50.
>
> "Twice stronger" in what way?

Because now it plays the same move in half time. Say it used to see 8
depth in 20 secs. Now after improvement it sees the 8 depth in just 10
sec. So that is Twice improvement.


> Do you really not understand just how vague that is??
>
> >So nowGetClubelo is +50 than it was earlier.
>
> Based on *ONE* program's estimated ratings??
>
> That's not valid.

I play against Rybka the worlds best program to test GetClub Chess.

> You need a lot more opponents than that to get a valid rating.
>
> And you need a lot more games to get that rating, too. =A0Hundreds of gam=
es
> are needed to get anything close to a reliable indicator of a change in
> strength like that.

Yes, I also analyze games played by other players to find weaknesses
in the game.

Bye
Sanny

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




  
Date: 18 Aug 2008 09:37:35
From: Guest
Subject: Re: GetClub Twice Stronger again !!!
>"Sanny" <[email protected]> wrote in message
>news:0e3c24ed-793a-4d08-9b0a->[email protected]...
>> ?> ELO ratings? Not bloody likely. At that rate, GC would be around
>> 30,000
>> ?> elo by now.
>>
>> >According to Rybka everytime a playes becomes twice stronger his elo
>> >increases by +50.
>>
>> "Twice stronger" in what way?
>
>Because now it plays the same move in half time. Say it used to see 8
>depth in 20 secs. Now after improvement it sees the 8 depth in just 10
>sec. So that is Twice improvement.

That's not twice *stronger*

That would be twice *faster* PROVIDED it's actually faster (ie: as if you
moved to a faster computer.)

Simply finding the same move in 10 seconds that used to take 20 seconds is
not necessarily the same as being stronger. It just means you found a
particular move faster than you used to. That's not twice "stronger".

(And if you are still playing GC & Rybka at different time levels, then that
introduces an extra set of problems, issues, and caveats... Just suffice to
say it's not valid.....)

Let me give you an example....

You search for 60 seconds. During that time, your program looks at lots of
moves. For a while it thinks move XYZ is best. Then it looks deeper and
thinks ABC is best. Then it thinks a bit more and decides JFK is better.
Chess programs regularly change their minds as they search deeper.

So at the end of 60 seconds (like if you were in a game and had 60 seconds
to find a move), you picked move XYZ.

Your 'twice faster' would find that same XYZ move in 30 seconds. BUT, and
this is important, the game is set at 60 seconds a move. So you still have
30 more seconds to think.... In that 30 seconds, the odds are good it's
going to change its mind. Is it going to be a better move, a worse move, or
just a different move?

That's what timed games are about. It may be a fixed 60 seconds a move, or
it may be 40 moves in two hours and you cut up the time as needed, or it may
be 'game in 30' where you have to play the whole game in 30 minutes.


The 'twice better' (which isn't accurate, but...) meaning is only somewhat
valid for beginner kind of play where there are no time limits of any sort.
That is a valid kind of play for beginners, but you need to qualify any
results from that because that is definetly not how real games are played.
And it makes a difference. Time control really makes a difference!

For beginners, you can't really give ratings because they don't understand
ratings. But you can sort of fake it without too many people complaining.

But by the time you start doing ratings of 1800, 2000, etc. you can't do
that. That's not beginner quality play. For the people who do play at
those ratings for real, its all about timed play and your ratings are a
joke.

Ratings and time control are linked. You can't seperate them. They are
defined as a pair.

Games with no time control can not get ratings. (Caveat... can you get
ratings with postal chess? There are still time controls but they are loose
enough to be considered without controls for this discussion..)



>> Do you really not understand just how vague that is??
>>
>> >So nowGetClubelo is +50 than it was earlier.
>>
>> Based on *ONE* program's estimated ratings??
>>
>> That's not valid.
>
>I play against Rybka the worlds best program to test GetClub Chess.

Well hot shit.

It's still not valid. What part of "ONE" do you not understand?



>> You need a lot more opponents than that to get a valid rating.
>>
>> And you need a lot more games to get that rating, too. Hundreds of games
>> are needed to get anything close to a reliable indicator of a change in
>> strength like that.
>
>Yes, I also analyze games played by other players to find weaknesses
>in the game.

No, you need to start do serious testing and stop this stupid fake rating
stuff.

Why the heck do you think so many many many (ALL) chess programmers use test
positions?

It's a way they can run a few hundred or thousands of standard positions so
they can get a tolerable idea of what their program does understand and what
it doesn't understand.

And as a way to see if there are any regressions. Positions that it used to
solve but now can't.

If you do still want to get accurate ratings, we've told you many times how
to do that..... And it's not by playing Rybka.





Bye
Sanny

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






----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.pronews.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= - Total Privacy via Encryption =---


 
Date: 17 Aug 2008 10:22:26
From:
Subject: Re: GetClub Twice Stronger again !!!
On Aug 16, 10:56=A0pm, Sanny <[email protected] > wrote:
> > > Here is Strong game played byGetClubagainst Rybka after the
> > > improvement done.
>
> > Why do you keep saying thatGetClubis *twice stronger*?
>
> > ELO ratings? =A0Not bloody likely. =A0At that rate, GC would be around =
30,000
> > elo by now.
>
> According to Rybka everytime a playes becomes twice stronger his elo
> increases by +50.
>
> So now GetClub elo is +50 than it was earlier.
>
> Here is recent ratings for GetClub Chess.
>
> Baby: 2000+ (2 sec / move)
> Beginner: 2100+ (15 sec / move)
> Easy: 2200+ (1 min / move)
> Normal: 2300+ (4 min / move)
> Master 2400+ (20 min / move)
>
> Higher levels are not worth playing as they take lot of time. But
> Beginner & Easy give good challenge to all.
>
> Bye
> Sanny
>
> Play Chess at:http://www.GetClub.com/Chess.html

These ratings are probably off by about 500-700 points. Just so you
know....

One other thing that you should understand is that REAL ratings are
also based on a time control. Example: if a player is rated 2000 at
tournament time controls (40 moves in 2 hours), and another player
needs 2 hours to find the same move that the 2000-rated player can
find in 2 minutes, then the slower player is NOT rated 2000, but
significantly lower.

Sanny, this is why your ratings are completely wrong.

jm


jm


 
Date: 16 Aug 2008 22:56:08
From: Sanny
Subject: Re: GetClub Twice Stronger again !!!
> > Here is Strong game played byGetClubagainst Rybka after the
> > improvement done.
>
> Why do you keep saying thatGetClubis *twice stronger*?
>
> ELO ratings? =A0Not bloody likely. =A0At that rate, GC would be around 30=
,000
> elo by now.

According to Rybka everytime a playes becomes twice stronger his elo
increases by +50.

So now GetClub elo is +50 than it was earlier.

Here is recent ratings for GetClub Chess.

Baby: 2000+ (2 sec / move)
Beginner: 2100+ (15 sec / move)
Easy: 2200+ (1 min / move)
Normal: 2300+ (4 min / move)
Master 2400+ (20 min / move)

Higher levels are not worth playing as they take lot of time. But
Beginner & Easy give good challenge to all.

Bye
Sanny

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






  
Date: 17 Aug 2008 13:50:14
From: Guest
Subject: Re: GetClub Twice Stronger again !!!
>"Sanny" <[email protected]> wrote in message
>news:[email protected]...
? > > Here is Strong game played byGetClubagainst Rybka after the
? > > improvement done.
? >
? > Why do you keep saying thatGetClubis *twice stronger*?
? >
? > ELO ratings? Not bloody likely. At that rate, GC would be around 30,000
? > elo by now.

>According to Rybka everytime a playes becomes twice stronger his elo
>increases by +50.

"Twice stronger" in what way?

Do you really not understand just how vague that is??


>So now GetClub elo is +50 than it was earlier.

Based on *ONE* program's estimated ratings??

That's not valid.

You need a lot more opponents than that to get a valid rating.

And you need a lot more games to get that rating, too. Hundreds of games
are needed to get anything close to a reliable indicator of a change in
strength like that.


It is mathematically impossible to get even a semi-accurate rating if you
only do it against one opponent.

It just simply does not work that way.



And what little you are doing, you aren't even doing right. You aren't
doing it time based. You give GC 5 or 10 times longer than Rybka, while
forcing Rybka to run at such absurdly reduced time limit that it can't even
play as well it should.



This is *NOT* about your program. This is about *your* refusal to do
accepted methods to get to a reliable rating estimate for your program.
People have been telling you this dozens or hundreds of times, and you are
still refusing to do things the right way.

And your refusal to even do standard test positions. Which wont give a good
rating estimate, but it would allow people to look at the results and
objectively judge the program's abilities in those types of situations.

I don't know how strong your program is, and I honestly don't care one way
or the other. This is all about your flat out refusal to even attempt to do
any sort of reliable testing to get those ratings you brag about.


It's a joke Sanny. People are laughing at you behind your back. Heck, some
are even doing it to your face.

If you want people to believe your ratings estimates, you must follow
accepted methods to get those ratings.

Until you do, you are the circus clown everybody is laughing at.

The poor country boy on his first trip to the Big City, who's trying to look
fancy and sophisticated while walking out of the lavatory with paper stuck
to his shoe.

The sucker that some pretty girl brought to a party so that everybody else
could laugh at.



I understand that doing reliable ratings tests are hard and time consuming.
Even doing automated tests on some internet chess server could be hard
because of how your GC is programmed. That's why I even gave you a way to
'cheat' and report some standard test positions. You could do those tests
in an evening and then compare those results to what other programs can give
at the *SAME* time controls.

That wouldn't be the same as a rating, but it would allow comparison of your
program to others that do have known strengths etc.

It would be something solid, rather than your ratings that you pick out of a
hat.







----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.pronews.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= - Total Privacy via Encryption =---


 
Date: 16 Aug 2008 09:26:37
From: Guest
Subject: Re: GetClub Twice Stronger again !!!
"Sanny" <[email protected] > wrote in message
news:[email protected]...
> Here is Strong game played by GetClub against Rybka after the
> improvement done.

Why do you keep saying that GetClub is *twice stronger*?

ELO ratings? Not bloody likely. At that rate, GC would be around 30,000
elo by now.

Length of the game.... A couple moves longe rmakes the program 'twice
stronger'? Quite possibly. You have mentioned things like that before.

Let me quote you something that International Master Charles Kalme once said
about a chess program that he used to be involved in....

Quote***

... the machine was never really in the game and had no viable plan beyond
the defense it was forced into. It is easy to make the best moves, in the
sense of losing more slowly, if these entail meeting immediate tactical
threats.

End Quote***

In other words, the program had no idea what to do to win. All it knew how
to do was lose more slowly when forced into it.

Admittedly, the chess program he was involved with was pathetic. Inovative,
but pathetic even by the standards of the time. But the priciple remains.
He is clearly pointing out there is a difference in knowing how to win and
just being able to loose more slowly when forced into it.


If you want people to take you seriously, you better start providing some
testing results to back it up.

Take the old Bratko test and the Win At Chess positions and report GC's
results at 5 seconds, 10 seconds and 30 seconds per position. Those
positions are standard testing positions from human games.

That at least is a reasonaly simple thing you can do to prove to the world
your program is actually improving.

And heck, if your program is actually decent, wouldn't you *WANT* to brag
about how well your program does on standard chess positions?






----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.pronews.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= - Total Privacy via Encryption =---


 
Date: 16 Aug 2008 15:27:11
From: Bjoern
Subject: Re: GetClub Twice Stronger again !!!
Sanny wrote:
> Here is Strong game played by GetClub against Rybka after the
> improvement done.
>
> Can you spot mistakes in this game by GetClub?

Plenty of huge mistakes e.g.
- 11...Nc6 just obviously gives up a pawn for nothing
- 18...Bg6 worsens the position of a piece while also giving up further
material
- 20...Bh7 is just idiotic - that Bh5-g6-h7 maneuver doesn't gain
anything while in the meantime white wins a piece
- 25...b6 just gives away another pawn for no reason.

Why do you call this a strong game?

Rybka is just mopping up after move 16, okay, while Rybka is objectively
winning from then onwards, it is reasonably to try to see whether there
is some hope of swindling a draw. However by move 30 Rybka is so
obviously winning, that I think we see one major lack in the GetClub
program: It does not know when to resign.

If the computer would resign in hopeless positions like this, that would
be a great new feature (but just to clarify, which seems necessary given
your past track record: the rules of chess all GetClub to resign, but it
does not allow the computer to decide that its opponent is resigning...)