Quote:
Originally Posted by Jay Random
The difficulty about averaging stats is that not all stats deserve to be weighted equally. So how do you weight them, and for that matter, how do you get rid of redundant information due to dependent variables?
At that point you're into hocus-pocus modelling, in which sheer opinion is hidden under a veneer of applied math. It's been said that with seven variables, you can make any data set fit any curve you choose if you fiddle enough with the coefficients.
|
Totally agree ... yet I think the average of the for and against still helps in quieting the noise from one stat without eliminating it all together.
If you only look at CF% for example the Flames are rock stars. If you bring in the high quality stuff their tough start reflects they are far from a perfect team.
Calgary's numbers ...
For
CF 5th
SCF 5th
HDCF 6th
Average 5.33
Rank of this average 5th
Against
CA 2nd
SCA 11th
HDCA 25th
Average 12.6
Rank of this average 14th
Overall ranking 6th
In this example I think this does a good job of placing the Flames. But you're right if I weight high danger more aggressively then they fall.