On Fri, Jan 02, 2026 at 07:59:59PM +0900, TAKAHASHI Kaoru wrote: > By the way, I think there's also demand for emulation of eXtreme Gammon2 > Performance Rating.
XG PR should be close to half of GNUbg error rate. Its definition is this, plus a few refinements. Has anyone tried to calculate such a PR from GNUbg analyses after the fact, (processing a text export of match analyses and computing the PR from individual errors, for instance) and comparing it with XG PR? Then some statistical test could compare the influence of the exact PR formula vs. the difference between GNUbg and XG analyses. If the latter is dominant, computing a XG-style PR may not matter much.
