Quote:
Originally Posted by Fighting Banana Slug
As an advanced stat fence sitter, this was pretty informative, but ultimately a pretty poor prediction. Not so sure it is more accurate than any old school guy's prediction going into the season. One obvious problem, is that the point totals seem like they total more than the points available over the season. I mean, how can the worst team be in the 80+ range? I realize the 3 point game makes it more difficult, but I would have thought the average should be around 82. That seems like a flaw that could be easily fixed. I would also be interested in seeing maybe an average of this years Fenwick, to see if the prediction on Fenwick for this year is valid.
Honest question for stat guys, how much above zero does the R2 value need to be to assert a "strong" correlation? I understand anything more than zero is better than a coin flip, but I think I could predict games at better than 50% without any advanced stats.
|
Part of the issue with his analysis is he attempted to deal with changing rosters by predicting lines and time on ice for various players. His solution is a bit of a house of cards with advanced stats being just part of the input. To me the overall methodology is flawed enough that it's hard to say to what extent the advanced stats were predictive or not.