[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Unexpected results Gnu vs Gnu

From: Peter Carlson
Subject: Re: [Bug-gnubg] Unexpected results Gnu vs Gnu
Date: Sun, 03 Sep 2006 15:24:13 -0600
User-agent: Thunderbird (Windows/20060719)


Thanks for the reply. However maybe I didn't pose my original question correctly. I dont think the problem is cli vs gui. I think the problem is in how we use the cli.

The gui plays the whole game as such has complete game history.
The cli only has a snapshot (we paste in the positionid) and ask for a hint.

Is it possible that the fact the cli doesn't have the full game history somehow put it at a disadvantage?

PS. We are now up to about 150 games and are seeing the same lopsided results. I realize that we need to play many more games to get a "true" statistical pool, however 150 *should* start to see an equalizing of the statistics.

Ian Shaw wrote:

Peter Carlson wrote:
Am up to 20 games with the same results.

It sounds like a laborious process. You would need more like 2000 games to get a meaningful result with your method.

If you want to compare the two, try this.

Open a gui and cli copy.
Set them both to use the same settings, which they should be when you open the program.
Make gnubg the player for all four players, at the same skill level.
Set both to use manual dice.
Enter the rolls and see what each plays.

Note the positions where the cli play is different from the gui play.

Even faster.
Let the gui play itself for a session using automatic dice.
Save the session
Open the session with the cli.
Analyse the session with the cli.
Note the positions where the cli disagrees with the gui.

You will have to check that the analysis settings are the same as the evaluation settings.

-- Ian

This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]