| |
BGonline.org Forums
XG weaknesses?
Posted By: Timothy Chow In Response To: Guy returns to backgammon after not playing for four years; guy asks questions about eXtreme Gammon and XG Roller double/beaver position (Flaccus)
Date: Sunday, 7 November 2010, at 6:17 p.m.
I think it's not too meaningful to do comparisons on freak positions. Some freak positions will be messed up by Bot A and not Bot B, and vice versa. What I think you want to know is whether XG has some systematic weaknesses in "normal" positions that are peculiar to XG. (All the bots still have trouble with tasks like rolling a prime all the way around the board, or more generally playing positions with a gazillion checkers back.)
This is a very interesting question and I don't think anyone has done a systematic study of it. In principle one could go through Michael Depreli's bot-comparison positions manually and try to arrange the mistakes into categories, but this is a difficult and arduous task that nobody has attempted, to my knowledge.
I would be interested to hear if anyone has accumulated any anecdotal evidence about the weaknesses of the various bots. For example, here's my impression of GNU 2-ply's weaknesses:
— It tends to overestimate the defender's chances in holding-game cubes.
— It tends to overestimate the defender's chances in late blitz cubes.
— It has a slight bias in favor of slotting in the opening.
— It sometimes badly misjudges middle-game doublets of 33, 44, or 55 (particularly 44) where one has to decide whether to advance the back anchor or build the front position.
I'm curious whether others agree with this list and whether there's a similar list for XG 3-ply/4-ply/XGR+.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.