|
BGonline.org Forums
Bias in how most do XG analysis?
Posted By: Bob Koca
Date: Thursday, 31 July 2014, at 5:25 p.m.
Suppose you use world class default setting for XG analysis.
There is a first pass using 4 - ply. Then a second pass with as best as I can tell (the columns are not labeled but I assume first is for checker play and next is for cubes) with roller + if the cube is at all different and roller + if checker play is an error or blunder or outplay.
Suppose that 4 - ply thinks your cube action was correct. There is a chance that roller + would overrule it resulting in an error for you but that is not checked. Only checking the errors results in bias towards lower PR. Similar comments apply for the other presets.
For a custom setting it would be good to be able to instruct a deeper look at all plays that are within some level of being close. It doesn't seem possible to do that though (except by going through by hand).
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.