Posted By: Michael Petch In Response To: PR xg-gnu-snowie-bgblitz (Daniel Murphy)
Date: Wednesday, 29 January 2014, at 7:52 p.m.
In Response To: PR xg-gnu-snowie-bgblitz (Daniel Murphy)
The only way I could see a 1.4 is if somehow those 300 matches for some unknown reason had an inordinately high number of cube decisions per game (but too high for me to believe that the number is right). I don't think Snowie counted cube decisions, just moves. GNUBG and XG count both moves and close cube decisions (cube decision or close ones).
The only other thing I could think of is that at some point someone didn't see a distinction between any cube decision and a close decision. It wasn't until 2002 where GNUBG was given that notion. Prior to that any potential cube action (player holding the cube or cube centred) may have been considered a potential cube. In 2002 effectively any position that was < 0.25 (threshold) from the optimal cube decision was considered close (It was originally 0.10 but that was for a day before being changed to 0.25). In 2006 it was changed to 0.16 where it has remained ever since.
In a situation where all cube actions (no thresholds) are considered that would dramatically increase the overall cube decision and drop gnubg's error rate per decision putting it closer to Snowie.
The person who amended the documentation was Joern Thyssen (he might know where the numbers came from or who supplied them). If I have a chance I can drop him a line. I think it should be at least be removed from the docs to avoid confusion.
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.