[stupac2]

[quote author=CTcard link=topic=18293.msg217290#msg217290 date=1512581488]

I'll have to think a bit about what I think these numbers really mean.

Not sure if that is significant or not.

I'm not sure either.  What is significant is: who is good?  Say top-4 good.  Clemson, Georgia, and Oklahoma all have good offenses when I watch them.  Alabama less so, but let's pretend.  So their rankings are respectively 102, 88, 116 and 92.  Which end (Temple or Illinois) is better?  Least variance? (Illinois).  Illinois? Good teams with a good season are all over the map, but the absolute best is clustered between 88 and 116.  The median value of this is 102. That's the absolute best! (Clemson).  But on their shoulder are Nebraska (sucks), Charlotte (basketball school), SMU (kinda okay), and Coastal Carolina (basketball school with killer mascot).  So what are we supposed to do with this information?

Appreciate Supac's effort, but not sure these numbers tell us much. Context is everything.

Feels like hashing through rainfall data, knowing an outcome I want it to support; and italicizing something like absolute best with an exclamation point; when that just randomly fell out of the data without correlating to anything meaningful.
[/quote]

The best offenses can have off days. Also, because these are opponent-adjusted a strong-but-not-insane game against a bad opponent can have weird effects. But you expect something sorted by variance to be kinda scattershot, and you definitely expect both the very best and very worst to be sorted to the "low variance" side. In fact it's probably biased to the bottom, it's easy to be consistently terrible, consistently excellent is harder.

Anyway, yes, there are lots of problems with this exercise, it was just an easy "data I have available to me" check on our expectations. If anyone has a better suggestion for some way beyond the eye test to see if Stanford's variance is atypical, I'm all ears.

[stupac2]

[quote author=CTcard link=topic=18293.msg217290#msg217290 date=1512581488]

I'll have to think a bit about what I think these numbers really mean.

Not sure if that is significant or not.

I'm not sure either.  What is significant is: who is good?  Say top-4 good.  Clemson, Georgia, and Oklahoma all have good offenses when I watch them.  Alabama less so, but let's pretend.  So their rankings are respectively 102, 88, 116 and 92.  Which end (Temple or Illinois) is better?  Least variance? (Illinois).  Illinois? Good teams with a good season are all over the map, but the absolute best is clustered between 88 and 116.  The median value of this is 102. That's the absolute best! (Clemson).  But on their shoulder are Nebraska (sucks), Charlotte (basketball school), SMU (kinda okay), and Coastal Carolina (basketball school with killer mascot).  So what are we supposed to do with this information?

Appreciate Supac's effort, but not sure these numbers tell us much. Context is everything.

Feels like hashing through rainfall data, knowing an outcome I want it to support; and italicizing something like absolute best with an exclamation point; when that just randomly fell out of the data without correlating to anything meaningful.
[/quote]

The best offenses can have off days. Also, because these are opponent-adjusted a strong-but-not-insane game against a bad opponent can have weird effects. But you expect something sorted by variance to be kinda scattershot, and you definitely expect both the very best and very worst to be sorted to the "low variance" side. In fact it's probably biased to the bottom, it's easy to be consistently terrible, consistently excellent is harder.

Anyway, yes, there are lots of problems with this exercise, it was just an easy "data I have available to me" check on our expectations. If anyone has a better suggestion for some way beyond the eye test to see if Stanford's variance is atypical, I'm all ears.