[CTcard]

Systems like Sagarin and SP+ give us objective, independent, meaningful ways to gauge competitiveness so we don’t have to rely on subjective, difficult to define criteria.

These systems are certainly objective and independent, whether they are meaningful requires some consideration - perhaps with the tone of meaningful-for-what. 

On S&P. 
The system is meant to be predictive, but after the seventh week of the season it incorporates only backward looking data. in itself, that's fine, as the backward looking data is far more accurate and likely the most useful for predictions. 
Preseason predictions are much rougher as there is no current-season data to use. That is why Connelly eventually drops his preseason data from the algorithm - when he does so is to balance too-small data sets versus unreliable data.

Sagarin is another matter. He doesn't write as much about exactly what his algorithm does and considers it a trade secret. It appears to depend far more on results than these newer analytics. 

Another issue which appears inherent in the data set (i.e. in schedules across the NCAA) is that the interconnectedness is poor - though I have never seen a direct analysis of this. Teams tend to play a large percentage of their games against their own conference with most non-conference games against local teams. Thus the data linking, say, the SEC to the Pac 12 is sparse. Or ACC to Pac 12. 
Now SP seems to be reasonably sophisticated in handling this but brilliant algorithms cannot create information that is absent in the data. 

As an aside, I have always had an intuitive preference for FEI but haven't bothered looking at any analytical comparisons of FEI vs SP+ or others. 

Thanks to very good algorithms and models I could rank for you every team in my lifetime and feel confident that the distinctions drawn are pretty meaningful. You can’t do that as well with wins but for something like Sagarin or SP+ this isn’t that mysterious. 

Well, Sagarin is kind of mysterious because he doesn't publish just what he does - though it's clearly mostly related to wins and points so it's similar to "doing that" with wins on a strength of schedule adjusted basis. 

The more detailed analytics are pretty good for that as long as one considers what they really measure. 
It tells you something about a team's control of factors that typically lead to winning against the field they play (with an SOS adjustment). 
It doesn't really tell you about how good a team is one year versus another independently of the quality of the entire field in a given year. 
Of course that is mostly fine because it could be that is all one can meaningfully analyze anyway, but it's worth keeping in mind. 

------------------------------

None of this is to take any issue with the predictions listed above. I fully expect Stanford to be among the very worst football teams in the ACC next fall.

[CTcard]

Systems like Sagarin and SP+ give us objective, independent, meaningful ways to gauge competitiveness so we don’t have to rely on subjective, difficult to define criteria.

These systems are certainly objective and independent, whether they are meaningful requires some consideration - perhaps with the tone of meaningful-for-what. 

On S&P. 
The system is meant to be predictive, but after the seventh week of the season it incorporates only backward looking data. in itself, that's fine, as the backward looking data is far more accurate and likely the most useful for predictions. 
Preseason predictions are much rougher as there is no current-season data to use. That is why Connelly eventually drops his preseason data from the algorithm - when he does so is to balance too-small data sets versus unreliable data.

Sagarin is another matter. He doesn't write as much about exactly what his algorithm does and considers it a trade secret. It appears to depend far more on results than these newer analytics. 

Another issue which appears inherent in the data set (i.e. in schedules across the NCAA) is that the interconnectedness is poor - though I have never seen a direct analysis of this. Teams tend to play a large percentage of their games against their own conference with most non-conference games against local teams. Thus the data linking, say, the SEC to the Pac 12 is sparse. Or ACC to Pac 12. 
Now SP seems to be reasonably sophisticated in handling this but brilliant algorithms cannot create information that is absent in the data. 

As an aside, I have always had an intuitive preference for FEI but haven't bothered looking at any analytical comparisons of FEI vs SP+ or others. 

Thanks to very good algorithms and models I could rank for you every team in my lifetime and feel confident that the distinctions drawn are pretty meaningful. You can’t do that as well with wins but for something like Sagarin or SP+ this isn’t that mysterious. 

Well, Sagarin is kind of mysterious because he doesn't publish just what he does - though it's clearly mostly related to wins and points so it's similar to "doing that" with wins on a strength of schedule adjusted basis. 

The more detailed analytics are pretty good for that as long as one considers what they really measure. 
It tells you something about a team's control of factors that typically lead to winning against the field they play (with an SOS adjustment). 
It doesn't really tell you about how good a team is one year versus another independently of the quality of the entire field in a given year. 
Of course that is mostly fine because it could be that is all one can meaningfully analyze anyway, but it's worth keeping in mind. 

------------------------------

None of this is to take any issue with the predictions listed above. I fully expect Stanford to be among the very worst football teams in the ACC next fall.