CFL.ca Staff

TORONTO -- There is no doubt that Canadian Football League fans are passionate about OUR game. What has become equally apparent is that fans are passionate about Power Rankings.

Since we released our third weekly CFL Power Rankings 24 hours ago, we have seen the story quickly become one of our most read. Facebook and Twitter have lit up with comments and over at Riderfans.com some questioned whether we were indulging in some sort of recreational substances. The answer to that question is no!

When we set out to provide our fans with the CFL Power Rankings we decided to differentiate ourselves from other power rankings. While most power rankings are subjective (based upon the author’s opinion) we decided to create power rankings that are objective (based upon a mathematical model).

Our goal was to provide fans with a new, unique way of looking at Canadian football. And based on the level of discussion that has taken place we feel that we have achieved our goal.

Many fans feel that our model has made us look “foolish”, but we're quite confident in our model and hope that, as time passes, fans will see that it has merit, as well.

Some people feel that the model ignores winning. If the model were only concerned with winning, why have power rankings? If winning was all that mattered then we would simply refer fans to our standings page and by extension, all power rankings – subjective or objective – would be pointless.

The CFL Power Rankings do not "ignore" winning. It's just that the statistical model that you will have read about assumes that winning is a function, or end result, of all those individual events that occur during the course of a game.

To give you a little more background on how this idea developed, we determined, based on a thorough review of game results last year, that the most important positive predictors of team success were how well the quarterback played and how well the team rushed the ball. That is, these facets of the game contributed to a team scoring more points.

Conversely, on the negative side of things, we determined that the number of sacks given up and the number of field goals missed contributed to a team scoring fewer points.

We ran through hundreds of different variables that impact the score of the game and it was these four variables that impacted the score the most.

What's really cool about the method we used (called regression analysis) is that it not only tells us how "predictive" each of these four variables are but how much the score is impacted by the occurrence of these variables.

Is it 100 per cent accurate? No, of course not.

There are thousands – if not millions - of variables the impact who wins and loses each week in the CFL. We have only tried to quantify the end result of the game by using four of these variables. Even if we were to add more variables such as fumble recoveries or time of possession, our preparatory work indicates that these variables do not add any measure of accuracy to the model. And we also acknowledge that there are simply some elements of a football game that just cannot be measured using math. Surely you’ve heard about the “Argo bounce”!

The model that we have presented based on our work has 75 per cent predictive accuracy, which is pretty good if you really think about it. But it will work even better once more time has passed.

It is our hope that the CFL Power Rankings have provided a new angle to look at Canadian football as you enjoy the games this weekend.