First, sorry I walked away. Work had me tied up.
I disagreed with this and stated so below. For many reasons I value statistics. Others do to as it is the source of a billion dollar industry in gambling.
I also value statistics. I was crunching numbers on advanced baseball stats before there was WAR.
You echo his sentiment and do not address my point.
I echo the his sentiment insofar as he questions a portion of your data set. Your posting of a handful of definitions regarding FEI tells us nothing about how they determined the scope or weighting of that data set.
Maybe I misunderstood this statement? I don't see any objective, empirically derived data except that which I provided.
I was getting at this...
Number of players from Pittsburgh who participated in games 7 years ago and will participate in 2013: 0.0
Number of players from Notre Dame who participated in games 7 years ago and will participate in 2013: 0.0
There is an infinite universe of statistics to use. We must make judgments on which are useful. Empiricism helps inform us of what portions of objective reality affect the outcomes we are trying to measure. He showed a genuine concern about the usefulness of the data you presented which is based on objective reality. You called it an "eyeball test" and, frankly, I have no idea where you got that characterization.
Statistics do have a wonderful virtue of having the ability to be applied in a number of ways to predict behavior. So it is contradictory in that what rikkitikki said did not provide any evidence for his opinion and was dismissive of the historical trends plus the current trends as provided by professionals who make their living off of analyzing data. It is an eyeball test in the absence of evidence.
You didn't provide evidence either. Does 5 year old data improve our ability to predict the outcomes of various college football games today? I'm not talking about the r coefficient in comparing 2008 team quality with 2013. I'm talking about whether adding that to our data set improves our ability to forecast over a model which ignores outcomes that old. It's a pertinent question.
The stats I provided are adjusted and weighted accordingly and present a well rounded picture of each teams current performance on the field, as well as historically.
This plus a cursory analysis of the historical stats from each team show that the teams are performing similar to the last 5 years for whatever reason.
The question isn't merely correlation between 2008 and 2013. It's whether the 2008 results improve forecasting.
Anyway, my issue with your statements comes down to this: he raised an issue and I thought your answer was non-responsive. If the best counterpoint you can make is that it's somebody else's model and you don't know why it incorporates that information then say so. It's not an attack on statistical modeling in general to question Football Outsiders' methodology. And it's no answer to such concerns to say that team success from five years ago correlates to team success today. That's the realm of Pancake Flops (it's the first hit on google if you're not familiar)
