|
BGonline.org Forums
Bot Comparison S4 & Gnubg Snowie ERs
Posted By: Michael Depreli In Response To: Bot Comparison S4 & Gnubg Snowie ERs (Michael Depreli)
Date: Tuesday, 27 April 2010, at 12:33 p.m.
Tim:
If I understand correctly, these "error rates" don't include plays where all the bots agreed on the same incorrect play. Right? Michael, do you have any sense for how much difference this might make to the results?
I had a quick look through the rolled out sessions to get a rough idea how many extra moves / cube actions I would roll out. I set a threshold of 0.03 and there were far fewer than I expected. Many plays obviously don't make it passed 0-ply as they are trivial and given the number of bots and plies and their differences of opinion many of the positions with close moves / cube decisions have already been rolled out.
So I'm guessing there wouldn't be a huge difference in the numbers. The more interesting question might be how many positions are all the bots really messing up so much so that the best move appears way down the ranking and wouldn't have been rolled out.
Michael
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.