Originally Posted by ian2
Great discussion! My thoughts are very much in-line with what dizzlmcwizzl's was saying, and I agree with the above as well. But to me, "not perfect" is the key here: I feel that NTRP's attempt to "equate levels across local areas" is based on a faulty premise. A set of faulty premises actually: the algorithm makes statistical determinations based on a sample size that is too small to support such determinations; on top of that, the sample itself is likely to be NOT representative of the "average" level of play in a given area. Then the problem is compounded by giving this questionable "denominator" too much weight.
I'm actually not sure that this works better than having no equalizing in place, even for the stated purpose of equating NTRP levels across local areas. As for the role this plays in calculating individual year-end DNTRP: while benchmark recalculation produces a ripple effect for all players, it almost certainly affects players who happened to play directly against benchmarks in a non-linear way, and can result in some odd NTRP outcomes for such players. It appears that one blow-out loss (or win... it goes both ways) against a benchmark player might outweigh a player's entire record during the year.
If you didn't at least TRY to normalize the ratings across areas, then you would definitely get the ratings "drift" that people constantly claim on here in stronger tennis areas but really doesn't actually exist and a bias towards larger areas at nationals. The normalization isn't perfect for sure, but it does do as good of a job as it can of at least eliminating the creation of hot spots whose teams win nationals every year.