Originally Posted by
sagegrouse
This is like regression analysis with 100 variables and 26 observations (1985-2010). If you try all the potential combinations of variables you are guaranteed a really good fit. Is this science? Uh, no! It's curve-fitting. And his approach is even worse, since so many of his criteria are essentially the same.
Here's a rather Grouse-y analysis of his method:
1. Win at least 19 games the previous year. Actually, he looked at every single number of wins and came up with 19 as including all but one team. Curve fitting. Hunh?
2. He was unimpressed with the record of one-win leagues. Another throwaway.
3. Be a regular participant (i.e. the prior year in the NCAAs). This means being one of the top 40 or so teams the previous year. This is another three-inch hurdle, but of course, both Kentucky and UNC have tripped over it. Not terribly meaningful.
4 Be at least a fifth seed the previous year. This is a lesser included subset of the previous category. Anyone have an Occam's Razor?
5. Have won one game in the NCAA-T the previous year. Ditto. How many 5th seeds or better fail to win at least one tournament game? So we have three of his six criteria being very closely -- uh -- related. Moreover, pretty much all #5 seeds have won 19 games the previous year. Moreover and furthermore, how many teams from one-bid leagues get a #5 seed or win an NCAA tournament game? I would have been roasted by my colleagues doing statistical analysis with this kind of logic.
6. Return two of the team #1's in scoring, assists, and rebounding. You may disagree, but I say this is hokum. What about % of points returned? Percent of rebounds? Or minutes, denoting experience? Or, two of the top three in each of these categories. The fact that he sorted through 26 data points and found a fit with two of the team leaders in three categories doesn't impress me. And, of course, Duke doesn't qualify for 2011 because Scheyer very slightly out-scored Kyle and Nolan.
Anyway, you asked. Pardon the dyspeptic response.
sagegrouse