buggnubg - Bugs: bug #38138, "take or pass" cube...

 
 

bug #38138: "take or pass" cube evaluation functionality is buggy or simply unimplemented in gnubg-cli.exe

Submitter:  sorin <sorinpop>
Submitted:  Mon 21 Jan 2013 07:26:35 PM UTC
   
 
Originator Name:  sorin pop Item Group:  * Other bug
Category:  * Evaluations Release:  * 0.90-devel
Privacy:  Public Operating System:  * Windows XP
Severity:  3 - Normal Open/Closed:  Closed
Assigned to:  None Priority:  5 - Normal
Status:  None
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Sat 07 Dec 2013 08:29:54 PM UTC, comment #7: 

I've run the latest version through a large number of games now and I am not having the doubling issues anymore.  Thank you very much for your help.

Vivek Mittal <vmittal>
Mon 25 Nov 2013 11:33:06 PM UTC, comment #6: 

Okay. There are a few things going on here. First the position that is posted in this bug report by the original poster is actually an obvious take (unless I am reading it wrong). Player O has all of their pieces off and player X is on move. So O definitely has a take. Unless I am missing something I think the position posted is not a good example.

Secondly, there was a bug in the external interface regarding cubing that was fixed in early 2012 that would make it appear like most cubes are takes. Your 2011 version pre-dates that fix. The bug in question we dealt with (and closed) here: https://savannah.gnu.org/bugs/?36485

So, what I'd like you to try is to get a newer version of GNUBG. There are a couple version you might consider. There is a newer version 1.02.000 from July of this year that uses the new neural network (It is stronger than the versions earlier than 2013). It is recommended. That can be downloaded here: http://files.gnubg.org/media/windows/gnubg-release-1_02_000-20130728-setup.exe

If you want to use the older neural network but have the bug fixes that relate to the socket (external) interface then you can download:

http://files.gnubg.org/media/windows/gnubg-MAIN-20121023-setup.exe

Can you try one of them out and see if you still have issues?

Michael Petch <mdpetch>
Group administrator
Mon 25 Nov 2013 07:25:34 PM UTC, comment #5: 

I'm using version 0.90-mingw 20111003

I'm generating the FibsBoard position myself and then sending it to GNUBG.

Vivek Mittal <vmittal>
Mon 25 Nov 2013 07:08:30 PM UTC, comment #4: 

What is generating your FibsBoard positions. Are you building a bot that plays on the Fibs Server and taking the output from Fibs and sending it to GNUBG (via the socket) or are you generating a FibsBoard position yourself and sending it to GNUBG?

Second what version of GNUGB are you using. The original author was using an older one, but don't know what you are using now.

Michael Petch <mdpetch>
Group administrator
Mon 25 Nov 2013 07:55:39 AM UTC, comment #3: 

I tried the board string provided and my gnubg-cli gave back the response drop. 

However, I understand what the original submission was talking about as I have the same feeling - that the cli almost always gives back a take response.  I've even had a case where my winning percentage was 0.1 % and the chance of losing by a gammon was 48% and the cli reported a take decision on a double.

I'll try and post some examples when I next come across this issue.

Vivek Mittal <vmittal>
Fri 07 Jun 2013 07:56:37 AM UTC, comment #2: 

I believe this a matter of the documentation being incorrect and that the players are reversed.

Michael Petch <mdpetch>
Group administrator
Thu 28 Feb 2013 04:04:03 AM UTC, comment #1: 

For some reason I didn't receive an email notification of this bug report. Were you able to resolve the issue? My first question would be what version are you using? And is it possible you have the players reversed?

I don't believe there is a bug at present that would prevent a cube action from being reported wrong.

Michael Petch <mdpetch>
Group administrator
Mon 21 Jan 2013 07:26:35 PM UTC, original submission:  

Hi, gnubg-cli.exe seems to be working wrong when I am sending it a FIBS board which needs a take or pass cube decision evaluation. I build the FIBS board string so that the "was doubled" value is set to one, that's how gnubg-cli.exe knows that it has to make a take/pass decision. And it seems to recognize that, it's only that it seems that it is always returning the reply "take", even for obvious "pass" board states. I never saw it return one "pass", and it never seems to "think" on it, even if it uses the analisys settings and those are set to a high standard (at least 3 plies, pruning, etc.), it just always returns a "take" right away. So my guess is that it is either a bug, or this feature is just simply not implemented in gnubg-cli.exe, it just doesn't call GNU backgammon's appropriate cube evaluation functions. Any feedback would be very welcome, even if just a confirmation of my guess. Of course, if this could be also fixed, would be just great!

Here is an example of an obviously "pass" board state, you can test it with this, if you send this string to gnubg-cli.exe with some simple socket tester program:

board:dianepop:opponent:7:0:0:0:2:2:1:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:1:1:0:-1:0:0:0:0:1:1:1:1:1:-1:0:25:0:0:0:0:2:0:0:0

sorin <sorinpop>

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

No files currently attached

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by vmittal (Posted a comment)
  • -email is unavailable- added by mdpetch (Posted a comment)
  • -email is unavailable- added by sorinpop (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only logged-in users can vote.

     

    Follows 1 latest change.

    Date Changed by Updated Field Previous Value => Replaced by
    2013-06-07 mdpetch Open/ClosedOpen Closed

    Back to the top

    Powered by Savane 3.13-d3ae.
    Corresponding source code