| |
BGonline.org Forums
GNU error rating vs. XG PR
Posted By: Timothy Chow
Date: Tuesday, 29 December 2009, at 12:43 a.m.
I'd like to make sure I understand exactly how GNU and XG are computing ER/PR.
In both cases, total lost EMG equity is divided by "number of decisions" and multiplied by a constant. The constant is 1000 in the case of GNU and 500 in the case of XG. The crux of the matter is the denominator.
I believe GNU counts all checker plays that are unforced, meaning that there is more than one way to play it. For cube decisions, the documentation says that GNU includes no-double decisions only if they are close, where "GNU Backgammon considers a cube decision close if the relevant equities are within 0.25 from each other or if the position is too good." Presumably, a no-double decision is always counted if the player gets it wrong, though the documentation doesn't seem to state this explicitly. The documentation also suggests that the GNU error rating will be about 1.4 times the Snowie ER. Do people find this to be about right in practice?
The XG webpage says that XG excludes forced moves and unimportant moves. What is the exact definition of an unimportant move? And what does XG do about cube decisions? One would expect GNU error ratings to be slightly less than 2x the XG PR; is this about right in practice?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.