an idea for analysis project of WBEC AEGT,...

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 06:01:42:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.
Note that I suggest testing only premier division programs or programs of the AEGT king class and I doubt if there is enough time only for this test(serious testing only of the next premier division will probably take a long time because my idea is to use 3 minutes per move on fast hardware for blunder check and it means that the blunder check may need some months of computer time).
I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.
Uri

[Telmo Escobar]

Geschrieben von:/Posted by: Telmo Escobar at 14 August 2004 07:34:48:
Als Antwort auf:/In reply to: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 06:01:42:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.

Go specific: do you want help finding mistakes in Movei's games? Have you already some tester making this?
Telmo

[Roger Brown]

Geschrieben von:/Posted by: Roger Brown at 14 August 2004 07:52:44:
Als Antwort auf:/In reply to: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 06:01:42:

I think that it is better if testers
spend more time on analyzing games by
blunder check with the commercial
programs and not on playing games.

I think it would be better if those
idle persons spent their machine time
searching for answers to world peace, a
cure for cancer and to write really
funny scripts for television.
That is just my perspective!


There are so many games by testers
that programmers usually have no time
to analyze even only the public games
of their programs.

One can only feel sorry for them. So
much good data and so little time.
Woe!

It may be more productive for
programmers to have a list of positions
that their program blundered so they
can use them to test a future version
of their program.

Uri, speaking for myself only, what the
programmers may find more productive
does not concern me. They may find the
games produced by Dann, Gunther,
Lyapko, Graham, Utzinger, Leo or the
AEGT group productive.
Or they may not.
They may find analysis of games
productive.
Or they may not.
They have largely (with a few
exceptions) been very silent on
everything but bug reports. There are
a few programmers helping out with the
AEGT. Their help is welcome.
Perhaps they are busy with their paying jobs. My respect for them offering up a chess playing engine for free knows no bounds.
I am not going to do stuff on my time
unless there is a payoff for me. Basic
human motivation. I play games because
Heinz got me excited about his idea.
He is like that. For no-one else am I
going to playing games at 80 minutes
for 40 moves!
Unless I am getting paid. What is the
author of an amateur engine going to
offer for the time spent?
Based on the zero response to some
light hearted analysis I posted based
on games of Gunther's tournament not
even an acknowledgement that the game
was analysed!
Why should I spend my time that way
Uri?
I play the games because that is fun.
That is my payoff. That is how I get my joy. The fun of
belonging to a group of persons who
have no other payoff but the fun of
contributing to a wider purpose.
Conducting analysis is going to
require my intervention as machine
analysis has limitations. Did I
horrify someone with that remark?

It may be a good idea if testers
will test the premier division programs
by blunder check of shredder8 and
Fritz8 to find positions when they went
wrong.

I do not have those programs installed.
The thought of installing them to do
this analysis is not terribly exciting
to me.

Note that I suggest testing only
premier division programs or programs
of the AEGT king class

Why? I would think that it would be
the engines of the lower classes that
would need the analysis in order to
improve. Then again, what do I know?

I even doubt if the testing that I
suggest is practically possible because
it is possible that testers are
interested in playing games and not in
other tests to help programmers to
improve their program.

The games do not help the programmers?
Let us test what I am saying. I am
saying that the authors care little
about these tournaments much less
analysis of their engine's games.
I would like any author who cares to do
so to offer an opinion.
Would you like to see analysis of your
engine's games? Why?
Let us see where the rubber meets the
road.
I was thinking of asking Telomo Escobar
- whose talents are certainly well
suited for the task despite his
numerous protestations about his age -
to do some analysis but no longer.
I find your presentation abrasive,
dismissive and not a little arrogant.
The playing of games and keeping it fun
is why the AEGT was formed. You want
to make what we do work. Without pay.
UGH.
You know that I am a Movei fan Uri and
we do exchange the odd e-mail now and
again but I suggest that you check this
single track approach to things.
It could give the wrong impression.
Later.
Ps. That is coming from a guy who has played thousands of games for various authors throughout the years.
Without regret.
To all of those persons who run dozens, hundreds, thousands of games, who do what they do purely for love - THANKS!
The world is a better place because you contributed.
My two cents.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 14 August 2004 08:05:51:
Als Antwort auf:/In reply to: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 06:01:42:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.
Note that I suggest testing only premier division programs or programs of the AEGT king class and I doubt if there is enough time only for this test(serious testing only of the next premier division will probably take a long time because my idea is to use 3 minutes per move on fast hardware for blunder check and it means that the blunder check may need some months of computer time).
I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.
Uri

I think its 'easy' to do by the author or one of his beta testers, in case of WBEC one engine have only 4 games/3days where the total number of played games is 30-36/3days!
I think the authors get enough support with the played games at this time controls, you ask to play the tourneys and analize the games then and when I ask for a simple resign option to save time the answer is that it can cost points in case of a bug of the opponent?
If all authors think that way playing tourneys at those timecontroles are impossible at all!!
Leo.



WBEC Ridderkerk homepage.

[Ross Boyd]

Geschrieben von:/Posted by: Ross Boyd at 14 August 2004 11:24:42:
Als Antwort auf:/In reply to: A response..... geschrieben von:/posted by: Roger Brown at 14 August 2004 07:52:44:

Based on the zero response to some
light hearted analysis I posted based
on games of Gunther's tournament not
even an acknowledgement that the game
was analysed!

Hi Roger,
Just to be sure, your light hearted analyses were very much enjoyed here. I hope I did post some appreciation... if not it was a terrible oversight.
As far as Uri's suggestion is concerned, I think programmers can ask testers for practically anything but if it ain't fun or interesting... then it probably ain't gonna happen.
Most of the amateurs don't own several powerful PC's to test with, so we rely on CC enthusiasts to help us gauge strength. Most of us have jobs, wives, lives, families, pets etc and therefore limited time resources... I'm lucky that my wife's name is Trace so she knows I am thinking about her while implementing TRACE's woeful king safety routines.
Yes, it amazes me that there are people who are willing to play tournaments with the weaker engines like mine. And its quite a buzz to follow those tournaments and experience the heartache, the highs, the lows, the undiscovered bugs and the occasional unexpected victories. To me, that's what its about. Its pure unadulterated fun.
I wish I could do it full time and get paid for it. That would be heavenly.
Anyway forgive the many digressions,
Best regards,
Ross

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 13:20:15:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Telmo Escobar at 14 August 2004 07:34:48:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.

Go specific: do you want help finding mistakes in Movei's games? Have you already some tester making this?
Telmo

I probably have no tester for this purpose(I did not check emails in the last 24 hours so I cannot say it as something certain) and I guess that movei is too weak and testers may be more interested in other engines.
If there are no testers to do this task for better engines then I even do not dare to ask it for Movei.
Uti

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 13:22:11:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 13:20:15:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.

Go specific: do you want help finding mistakes in Movei's games? Have you already some tester making this?
Telmo

I probably have no tester for this purpose(I did not check emails in the last 24 hours so I cannot say it as something certain) and I guess that movei is too weak and testers may be more interested in other engines.
If there are no testers to do this task for better engines then I even do not dare to ask it for Movei.
Uti

t and r are neighbours in the keyboard and I noticed the mistake only after I already click the reply
Uri

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 13:52:05:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Leo Dijksman at 14 August 2004 08:05:51:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.
Note that I suggest testing only premier division programs or programs of the AEGT king class and I doubt if there is enough time only for this test(serious testing only of the next premier division will probably take a long time because my idea is to use 3 minutes per move on fast hardware for blunder check and it means that the blunder check may need some months of computer time).
I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.

I think its 'easy' to do by the author or one of his beta testers, in case of WBEC one engine have only 4 games/3days where the total number of played games is 30-36/3days!
I think the authors get enough support with the played games at this time controls, you ask to play the tourneys and analize the games then and when I ask for a simple resign option to save time the answer is that it can cost points in case of a bug of the opponent?
If all authors think that way playing tourneys at those timecontroles are impossible at all!!
Leo.

You do a good job with WBEC and I did not mean to criticize you about it.
I do not blame somebody who does something for free.
I apologize if people understood that I complain against testers.
I only meant to express the opinion that more analysis may be more productive for the programmers relative to more games with long time control espacially when WBEC is now not the only long time control tournament and there is also AEGT and more tournaments at long time control.
My opinion about what information programmers need more may be different than the opinion of other programmers so it may be interesting to read also the opinion of other programmers.
Movei is also not in the premier division and it probably not going to promote to that division so I talked about stronger programs than Movei.
Based on results in other tournament it seems that movei was very lucky in the first division(I did not see it does so well in other tournaments that are not blitz and in Frank quinsky's tournament(10 minutes/40moves) it even failed to promote to the second division when Nejmet,Danchess and Naum promoted.
It lost against Yace and chess genius7 in chess War 40/40 games and also scored worse than programs like Amy or Amyan or Wildcat in AEGT queen tournament.
Uri
Uri

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 14:03:25:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 13:52:05:

I think that it is better if testers spend more time on analyzing games by blunder check with the commercial programs and not on playing games.
There are so many games by testers that programmers usually have no time to analyze even only the public games of their programs.
It may be more productive for programmers to have a list of positions that their program blundered so they can use them to test a future version of their program.
It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.
Note that I suggest testing only premier division programs or programs of the AEGT king class and I doubt if there is enough time only for this test(serious testing only of the next premier division will probably take a long time because my idea is to use 3 minutes per move on fast hardware for blunder check and it means that the blunder check may need some months of computer time).
I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.

I think its 'easy' to do by the author or one of his beta testers, in case of WBEC one engine have only 4 games/3days where the total number of played games is 30-36/3days!
I think the authors get enough support with the played games at this time controls, you ask to play the tourneys and analize the games then and when I ask for a simple resign option to save time the answer is that it can cost points in case of a bug of the opponent?
If all authors think that way playing tourneys at those timecontroles are impossible at all!!
Leo.

You do a good job with WBEC and I did not mean to criticize you about it.
I do not blame somebody who does something for free.
I apologize if people understood that I complain against testers.
I only meant to express the opinion that more analysis may be more productive for the programmers relative to more games with long time control espacially when WBEC is now not the only long time control tournament and there is also AEGT and more tournaments at long time control.
My opinion about what information programmers need more may be different than the opinion of other programmers so it may be interesting to read also the opinion of other programmers.
Movei is also not in the premier division and it probably not going to promote to that division so I talked about stronger programs than Movei.
Based on results in other tournament it seems that movei was very lucky in the first division(I did not see it does so well in other tournaments that are not blitz and in Frank quinsky's tournament(10 minutes/40moves) it even failed to promote to the second division when Nejmet,Danchess and Naum promoted.
It lost against Yace and chess genius7 in chess War 40/40 games and also scored worse than programs like Amy or Amyan or Wildcat in AEGT queen tournament.
Uri

Correction
I see that it does better than Amy in the queen AEGT but so far it did worse than Slowchess in that tournament.
It also does very poorly in RWBC and the only tournament that it does well so far is the infinite loop but the total picture is that movei is not very strong and I have no reason to think that it is better than the middle of the table of thefirst division of WBEC(I compared with 008_251s or 00_8247s but
the only difference relative to 00_8_247 is that it is slightly faster and has a bigger book except the nunn positions.
Uri

[Dan Honeycutt]

Geschrieben von:/Posted by: Dan Honeycutt at 14 August 2004 16:11:19:
Als Antwort auf:/In reply to: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 06:01:42:

[snip]

I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.
Uri

I belive "tests to help programmers to improve their program" are the responsibility of the author. Testers with some free machine time do offer to help the authors (I had some time and ran some tests for you prior to WBEC 7). In such cases the tests should be a team effort - maybe the author needs some games with blunder check or maybe he has a new king safety eval and needs games to compare old versus new. Testers running blunder check on a select group of engines burn a lot of machine time to produce data that is of interest to very few people. Better they spend this time running games which are fun, of interest to many, and also provide useful information to the authors.
Dan H.

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 14 August 2004 17:10:05:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Dan Honeycutt at 14 August 2004 16:11:19:

[snip]

I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.
Uri

I belive "tests to help programmers to improve their program" are the responsibility of the author. Testers with some free machine time do offer to help the authors (I had some time and ran some tests for you prior to WBEC 7). In such cases the tests should be a team effort - maybe the author needs some games with blunder check or maybe he has a new king safety eval and needs games to compare old versus new. Testers running blunder check on a select group of engines burn a lot of machine time to produce data that is of interest to very few people. Better they spend this time running games which are fun, of interest to many, and also provide useful information to the authors.
Dan H.

I think that test suites can be interesting to many people and blunder check games can give test suites based on practical positions that programs
blundered unlike most test suites that are based on gm games.
Games are also productive and I do not claim that they are less important but I think that the situation today is that most authors have enough games to look at them and it is better to get more of different information and not more games relative to the situation of today.
Different information may be tactical mistakes of the program or may be comments about the games of the program that can help the programmer to concentrate on the important parts of the games without spending time to look for the important parts.
Uri

[Telmo Escobar]

Geschrieben von:/Posted by: Telmo Escobar at 14 August 2004 17:30:06:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 13:20:15:

It may be a good idea if testers will test the premier division programs by blunder check of shredder8 and Fritz8 to find positions when they went wrong.

Go specific: do you want help finding mistakes in Movei's games? Have you already some tester making this?
Telmo

I probably have no tester for this purpose(I did not check emails in the last 24 hours so I cannot say it as something certain) and I guess that movei is too weak and testers may be more interested in other engines.
If there are no testers to do this task for better engines then I even do not dare to ask it for Movei.
Uri

Why do you say that your engine is "too weak"? It performs quite well in Leo's 1st division, near to the level of promotion. As there is a bit of randomness in any tournament, one has to think Movei is not significantly weaker than Premier division engines.
Also significant is that Movei plays very agressive and enterprising chess, I guess that style is what chess aficionados like most.

Telmo

[Dan Honeycutt]

Geschrieben von:/Posted by: Dan Honeycutt at 14 August 2004 18:35:28:
Als Antwort auf:/In reply to: Re: an idea for analysis project of WBEC AEGT,... geschrieben von:/posted by: Uri Blass at 14 August 2004 17:10:05:

[snip]

I even doubt if the testing that I suggest is practically possible because it is possible that testers are interested in playing games and not in other tests to help programmers to improve their program.
Uri

I belive "tests to help programmers to improve their program" are the responsibility of the author. Testers with some free machine time do offer to help the authors (I had some time and ran some tests for you prior to WBEC 7). In such cases the tests should be a team effort - maybe the author needs some games with blunder check or maybe he has a new king safety eval and needs games to compare old versus new. Testers running blunder check on a select group of engines burn a lot of machine time to produce data that is of interest to very few people. Better they spend this time running games which are fun, of interest to many, and also provide useful information to the authors.
Dan H.

I think that test suites can be interesting to many people and blunder check games can give test suites based on practical positions that programs
blundered unlike most test suites that are based on gm games.
Games are also productive and I do not claim that they are less important but I think that the situation today is that most authors have enough games to look at them and it is better to get more of different information and not more games relative to the situation of today.
Different information may be tactical mistakes of the program or may be comments about the games of the program that can help the programmer to concentrate on the important parts of the games without spending time to look for the important parts.
Uri

Agree that blunder check games produce good test positions but good test positions arise naturally as people interested in a game look it over and discover the crucial juncture.
I don't think you can ever have too many games. I don't look at all of them but I keep a crude database of version-opponent-time control-result which helps me to see if I've made progress.
I still think your blunder check suggestion has a low ratio of reward to time spent. But maybe I'm wrong and it would produce more useful information than I envision. I still don't feel we authors have any business telling the testers what to test. It's their time and their machine and they can do whatever they please. We use what they produce to the extent we can. Whatever else we need for development is our problem.
Dan H.