|
BGonline.org Forums
XG 3 Ply Error #1 - All Plies
Posted By: Timothy Chow In Response To: XG 3 Ply Error #1 - All Plies (Keene)
Date: Tuesday, 16 December 2014, at 1:21 a.m.
Keene wrote:
Yes, of course I understand about wastage and EPC, however, those (I believe) are just calculated numbers that it shows you, however, it relies on raw numbers to decide what play is best, it doesnt go looking at EPC and wastage, then think to itself "hmm". I am sure someone can confirm.
When it's not a straight race, the bot doesn't do either of these things, if by "raw numbers" you mean the pip count. It does some complex neural-net black magic.
I accept the answers from XG, I derive my own way of thinking about the 'whys' of them, and then I derive play options.
This is a perfectly reasonable approach in general. The only real disagreement I have is that I believe it is a bad habit to treat XG's verdict as gospel. XG is making no sense here. 7/2 is much better than 7/3 6/5? Come on. Extraordinary claims demand extraordinary evidence. If this is correct then something is radically wrong with our collective understanding of backgammon. Now, I agree that it's fine to come up with theories to try to explain why the bot move is correct. But in this case, those theories need to be tested rigorously because accepting them would require a serious overhaul of our understanding of this type of mutual holding game. The bot is not a god whose mysterious ways it is our humble privilege to defend at all costs. It should be cross-examined. I claim that I've cross-examined it in similar positions (e.g., this one) and its testimony has contradicted itself and fallen apart. This is not a position where its verdict should be trusted blindly. I'm sure that what's causing it to rate 7/2 over 7/3 6/5 is that Blue's blotty board is causing XG to get confused about whether White should volunteer a shot or Blue should hit a shot in the subsequent play.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.