Computer-Backgammon
Torsten Schoop
with Tony Lezard's program Dueller I am playing a round robin
tournament with a series of 1000 25-point matches each pairing.
Participants: Snowie 4.1, GNUBG (from 2 May 2003) and JellyFish 3.5.
All programs at highest strength (GNUBG at Supremo (checker play and
double decisions).
Standings this week:
GNU - JellyFish 3.5 = 306 - 227 -> 57.4 %
Snowie 4.1 - GNU = 242 - 227 -> 51.6 %
Snowie 4.1 - JellyFish 3.5 = 93 - 80 -> 53.7 %
Ciao
Torsten
Albert Silver
> Hi all,
>
> with Tony Lezard's program Dueller I am playing a round robin
> tournament with a series of 1000 25-point matches each pairing.
> Participants: Snowie 4.1, GNUBG (from 2 May 2003) and JellyFish 3.5.
>
> All programs at highest strength (GNUBG at Supremo (checker play and
> double decisions).
What MET is GNUBG using?
Albert Silver
Torsten Schoop
> What MET is GNUBG using?
>
> Albert Silver
>
I do not know the commands Dueller send to the command line interface
version of GNU (default?).
Torsten
Torsten Schoop
with Tony Lezard's program Dueller I am playing a round robin
tournament with a series of 1000 25-point matches each pairing.
Participants: Snowie 4.1, GNUBG (from 2 May 2003) and JellyFish 3.5.
All programs at highest strength (GNUBG at Supremo (checker play and
double decisions).
Standings this week:
GNU - JellyFish 3.5 = 330 - 248 -> 57.1 %
Snowie 4.1 - GNU = 262 - 249 -> 51.3 %
Snowie 4.1 - JellyFish 3.5 = 112 - 92 -> 54.9 %
Ciao
Torsten
Albert Silver
The MET must be set manually by the user. You need to set the MET, and
then save the settings (this can be done from the graphical user
interface BTW). As you made no changes, you are most likely using the
worst one (IMHO) the Zadeh MET. I'd suggest changing this to the
Snowie MET. Not because I necessarily think it is better than the
Jacobs/Trice, but because that way both GNU and Snowie would be using
the same MET, and any difference of opinion on a doubling decision (or
even checker play) would be due to a different evaluation and not a
different MET.
Albert Silver
Back4U2 BBL
>
> GNU - JellyFish 3.5 = 330 - 248 -> 57.1 %
> Snowie 4.1 - GNU = 262 - 249 -> 51.3 %
> Snowie 4.1 - JellyFish 3.5 = 112 - 92 -> 54.9 %
gnubg - JellyFish: 578 matches
Snowie 4.1 - gnubg: 511 matches
Snowie 4.1 - JellyFish: 204 matches
The reason why snowie-jfish only played 211 matches so far?
Nardy
Torsten Schoop
>
> gnubg - JellyFish: 578 matches
> Snowie 4.1 - gnubg: 511 matches
> Snowie 4.1 - JellyFish: 204 matches
>
> The reason why snowie-jfish only played 211 matches so far?
>
It's simple. I am playing on 3 machines. gnubg - JellyFish and Snowie
4.1 - gnubg since May but Snowie 4.1 - JellyFish since July
Ciao
Torsten
Torsten Schoop
It's the programmers part to set the MET they want to default.
Torsten
Albert Silver
> > > >
> > > > Albert Silver
> > > >
> > >
> > > I do not know the commands Dueller send to the command line interface
> > > version of GNU (default?).
> > >
> > > Torsten
> >
> > The MET must be set manually by the user. You need to set the MET, and
> > then save the settings (this can be done from the graphical user
> > interface BTW). As you made no changes, you are most likely using the
> > worst one (IMHO) the Zadeh MET. I'd suggest changing this to the
> > Snowie MET. Not because I necessarily think it is better than the
> > Jacobs/Trice, but because that way both GNU and Snowie would be using
> > the same MET, and any difference of opinion on a doubling decision (or
> > even checker play) would be due to a different evaluation and not a
> > different MET.
> >
> > Albert Silver
>
> It's the programmers part to set the MET they want to default.
>
> Torsten
The default playing strength is 0-ply.
Albert Silver
Torsten Schoop
You are kidding.
Torsten
Albert Silver
No, absolutely not. Make a new install in an empty directory and see for yourself.
Albert Silver
Torsten Schoop
You are kidding again and I am a PhD and not a kindergartner. I will
ignore your further statements.
Torsten
Albert Silver
> > > > >
> > > > > Torsten
> > > >
> > > > The default playing strength is 0-ply.
> > > >
> > >
> > > You are kidding.
> > >
> > > Torsten
> >
> > No, absolutely not. Make a new install in an empty directory and see for yourself.
> >
>
> You are kidding again and I am a PhD and not a kindergartner. I will
> ignore your further statements.
>
> Torsten
Torsten, I have no idea what you are talking about, much less the
relevance of your being a PhD. Look, when you first install GNU
Backgammon it has a series of default settings. These default settings
are far from ideal. Although I know them well, I checked and
re-installed in an *empty* directory (otherwise it will use your
previous settings) just to confirm my memory:
- The default board is some dreadful one with semi-transparent pieces
- The default MET is the Zadeh
- The default Analysis settings are 0-ply for checker play and 0-ply
for cube decisions
- The default Player settings (for playing against GNUBG) are 0-ply
for checker play and 0-ply for cube decisions
This was once again confirmed using the next-to-last Installation
Archive dated July 1st, 2003.
Naturally, all these settings can be changed by the user from with the
GUI, but they remain the defaults. My personal preferred settings are:
- Board: Easy Going (a design I made - though not available in that
older build)
- MET: Jacobs/Trice (though I will now use the new Mec26 released
today and that outscores the Woolsey MET by 1.2%)
- Analysis Settings: Supremo for checker decisions and World Class for
cube decisions (Supremo and WC are the same for cube)
- Player Settings: Supremo for checker decisions and World Class for
cube decisions
Dueller overrides the Player defaults when running the command-line
version of GNUBG, but does not affect the MET default.
Albert Silver
George
> > ignore your further statements.
> >
> > Torsten
>
> Torsten, I have no idea what you are talking about, much less the
> relevance of your being a PhD. Look, when you first install GNU
> Backgammon it has a series of default settings. These default settings
> are far from ideal. Although I know them well, I checked and
> re-installed in an *empty* directory (otherwise it will use your
> previous settings) just to confirm my memory:
>
> - The default board is some dreadful one with semi-transparent pieces
> - The default MET is the Zadeh
> - The default Analysis settings are 0-ply for checker play and 0-ply
> for cube decisions
> - The default Player settings (for playing against GNUBG) are 0-ply
> for checker play and 0-ply for cube decisions
>
> This was once again confirmed using the next-to-last Installation
> Archive dated July 1st, 2003.
>
> Naturally, all these settings can be changed by the user from with the
> GUI, but they remain the defaults. My personal preferred settings are:
>
> - Board: Easy Going (a design I made - though not available in that
> older build)
> - MET: Jacobs/Trice (though I will now use the new Mec26 released
> today and that outscores the Woolsey MET by 1.2%)
I installed the today build-030828, but there is no new Mec26 MET.
I mean in the MET directory there isn't any Mec26.xml file.
Am i missing something?
Back4U2 BBL
> > GUI, but they remain the defaults. My personal preferred settings are:
> >
> > - Board: Easy Going (a design I made - though not available in that
> > older build)
> > - MET: Jacobs/Trice (though I will now use the new Mec26 released
> > today and that outscores the Woolsey MET by 1.2%)
>
> I installed the today build-030828, but there is no new Mec26 MET.
> I mean in the MET directory there isn't any Mec26.xml file.
> Am i missing something?
http://users.skynet.be/bk228456/GNUBgW.htm
1a. Updated or new files.
2nd par: New MET
There is no need to download the (still buggy) latest build to use mec26
Nardy