power ranking stat calculation... any thoughts on how to do this?
So after the various power rankings threads, I got to thinking about how I would calculate power rankings if I were designing my own. And I came up with a concept that I'm calling Least Anomaly Calculation.
Here's how this power ranking theoretically works (I used the CFL in my initial trials because it has a very small set of teams):
1. Create a list of all possible rankings.
2. For each ranking, compare against all actual game results and find all anomalies. (For example, if Montreal is ranked ahead of Winnipeg in a particular ranking and Winnipeg has defeated Montreal, this is an anomaly).
3. Each anomaly historically graded, so that a more recent anomaly carries more weight than an anomaly at the beginning of the season.
4. Each ranking is assigned its anomaly score.
5. The ranking with the lowest anomaly score is used as the correct anomaly.
So then I started do build something to calculate this. I decided on using Excel. And promptly realized that just calculating all of the possible rankings, even for something as simple as an eight team league, would require 40 thousand lines within Excel. Extrapolating it up to a 30 team league like the NHL would be well beyond the scope - 21220228784975284690904678400000 lines.
So my question to all the stats enthusiasts out there, is how would you go about calculating it? Excel or some similar program? Pen and paper and come up with a better formula for arriving at this result? Write your own program?
|
