Re: [Bug-gnubg] bias in the number of gammons won every 100 games

[Michael Petch]

On 2014-04-28 4:29 PM, Mrs Bibi e Bibo wrote:
> Hi,
> I did a simple experiment in which I let the machine play against itself
> 1800 games in $ mode (unlimited). Actually I played the first 20 games
> before letting  the machine to play against itself using the "end game"

Since I know GNUBG doesn't cheat or manipulate the dice in someones
favor under normal circumstances there has to be (from my perspective)
some other explanation and the ones that comes to mind is that one of
the players is a stronger opponent than the other. A weaker opp is more
likely to get gammoned more.

Just out of curiosity, when you set up th GNUBG player what skill level
did you have it play at (expert, beginner, world class etc)?

One thing that crossed my mind is that if you set the bot (GNUBG player)
to a skill level higher than Expert it will play the opening roll (if it
wins the opening roll of a game) at the higher skill level. When you hit
"End Game" it then plays BOTH players using Expert Level (this is done
for speed).

You could always archive up your SGF file and send a copy to
address@hidden . When I have a chance I can look at the match data.

-- 
Michael Petch
GNU Backgammon Maintainer / Developer
OpenPGP FingerPrint=D81C 6A0D 987E 7DA5 3219 6715 466A 2ACE 5CAE 3304