| |
BGonline.org Forums
Disagree with Stick's Book Review
Posted By: Daniel Murphy In Response To: Disagree with Stick's Book Review (Rich Munitz)
Date: Friday, 10 December 2010, at 7:22 p.m.
"Universally accepted" better applies, I think, to Snowie error rates. By the time GNU Backgammon was created, Snowie ownership was widespread, and Snowie error rates were widely bandied about as meaningful measures of ability. They still are. Gnubg uses a different, and to many minds better, method of calculating error rate. As Ian Shaw wrote in November 2009, "To most backgammon players, the gnubg/XG method is more intuitively correct, but they are used to the Snowie numbers." But Gnubg catered to the "universally accepted" Snowie error rates by also reporting a "Snowie equivalent" error rate. That explains why the Gnubg scale is not universally accepted, and why the Snowie scale still is.
As a long-time Gnubg user, I have to admit that I've never put the Gnubg error rate scale to any use. For purposes of self-rating (like seeing how you improve over time as you add matches to your database), it hardly matters since all I need to know is "the lower the better." And when I share error rates reported in Gnubg-analyzed matches, I state the Snowie equivalent error rate, on the assumption -- well-founded, I think -- that it is meaningful to readers but the Gnubg error rate is not.
I'm perplexed at Tim's assertion that "PR" is a "new system" and that Xavier "was hardly anxious to make sure his product lined up nicely with Snowie's." As far I have been able to see, XG's method of calculating "PR" is nearly identical to Gnubg's method of calculating "ER." The methods differ in minor details. I don't know of any one place that lists all the ways they differ, and don't know of any place at all where it's demonstrated that the XG formula is superior to the Gnubg formula. The main difference, as far as I can see, is not in the formula but in the reporting. Gnubg says in effect here's the Gnubg rating, using a better formula than Snowie's, and here also is the Snowie equivalent rating, for those of you more used to Snowie rating. The Gnubg scale never caught on. Perhaps it would have, if Gnubg hadn't bothered to report Snowie equivalent ratings. Whereas I hear XG saying here's a "new" PR formula, differing in small and possibly insignificant and possibly not even better ways from the Gnubg formula. But instead of also reporting a Snowie equivalent rating, as Gnubg does, XG certainly and intentionally does make PR "line up nicely" with Snowie ER, by dividing total error by 500 instead of 1000 -- points being (1) the only reason to divide by 500 is in fact to get XG PR to "line up" with Snowie ER, and (2) dividing by 500 is a kludgy way to do that. I'd agree that XG wants PR to be the new "universal" standard, but it's ironic that it does so by making PR not too different from Snowie ER, thus perpetuating the Snowie scale.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.