Cullen: NHL Power Rankings Methodology

Scott Cullen
10/14/2009 11:33:41 AM
Decrease Text SizeIncrease Text Size
Text Size

I have always been fascinated by numbers and how statistics apply to sports.  As a kid, I memorized stats and was nearly as enamoured with the stats generated by sports as I was by playing the sports in the first place.

This fascination continued as I went to university, where I studied Economics and Business Management at Wilfrid Laurier.  At the time, I had begun following the Sagarin Ratings in USA Today and I found them very interesting, yet something stood out to me.

It was a time when the Colorado Avalanche were one of the dominant teams in the NHL, led by Joe Sakic and Peter Forsberg, yet in the 1996-1997 season, Sakic and Forsberg both missed significant time and I recalled looking at the Sagarin ratings and thinking that the flaw of any stastistical model that uses overall results is that it doesn't take into account who is actually playing and it matters who is playing, or else we're just evaluating the laundry.

After all, the Avalanche minus Sakic and Forsberg were not the same team and shouldn't have been held to the same high standard.

In any case, it was this realization that prompted me to try and generate my own method for power rankings and it involved establishing ratings for every player.

The individual player ratings are generated using a weighted formula that includes the following statistics:

Goals per game, Assists per game, Plus-Minus, Power Play Goals, Shorthanded Goals, Game Winning Goals, Shots on Goal, Blocked Shots, Hits, Giveaways, Takeaways and Faceoffs.

Goaltenders are measured using goals against average, save percentage, won-loss differential and shutouts.

As more and more statistics have become readily available over the years, and continue to evolve, the ratings have been tweaked to reflect new and additional information.

Naturally, goals are most important (since there is a 100% correlation between scoring more goals and winning the game) and the scale decreases so that the more peripheral stats are virtually tie-breakers.  That is, if you have a pair of 30-goal scorers, you would prefer the one that hits more frequently to the one that doesn't.  It wouldn't likely provide a decisive advantage but, all other things being equal, would indicate a preference for one player over another.

Arriving at a single rating number for each player simply helps to synthesize an individual's contribution to the collective effort.  Player ratings generally fall between 55 and 100, give or take a few points, with the very best players in the league pushing, or occasionally surpassing, 100, while fringe NHLers will be in the mid-50s.

So, now that each player has a rating based on his statistical production, the next step is to generate team rankings and I do that by weighting each player's contribution based on their role on the team.  Thus, first line forwards count much more significantly than fourth-line forwards and top-pair defencemen are more impactful than third-pair defencemen and so on.

What this methodology allows is for players to be moved in and out of the lineup based on injuries, suspensions, trades -- whatever transaction will affect the lineup, so that a value can be established for the current team.   I readily acknowledge that it is an estimated value for any team, but I readily acknowledge it because there is not a flawless, guaranteed accurate measure that exists, and I've found this model to be representative, even if the results aren't always popular.

Additionally, by using statistics alone to generate the rankings, there should be no illusions of bias or favouritism affecting the rankings. 

That means the team that is going on the ice that night, not the one that was playing weeks or months ago or will be playing when an injured star returns because, when the injured star returns to action, that will also be reflected in the rankings.

Consider the rankings to be a snapshot (think photography, not Ovechkin coming down the wing) of a team's value at a given time.    For the purposes of, that will most often be on Mondays, when the rankings are updated each week.  That means the rankings are always subject to change based on any number of factors, whether it's injuries, trades or simply improved results.

While the end result of the weighted individual grades provides an approximate value for each team, over the years I've incorporated small weightings for factors like divisional strength, power play and penalty killing results to help smooth statistics that are compiled against different levels of competition.  These aren't huge factors in the overall rating but certainly if teams are of similar value otherwise, the team in the stronger division will most often prevail in the rankings.

The premise, then, of my NHL Power Rankings is to determine the team that would be favoured to win a seven-game series on neutral ice.  The theoretical premise is based on neutral ice because home-ice advantage is a circumstance of play, not a measure of a team's actual quality. 

Given that any team could beat any other in a single game, the likelihood swings towards a more statistically-relevant result when it comes to a seven-game series.

A couple observations that I've found over the years:

There is not a uniform difference from one spot to the next, so don't automatically assume that a team that moved up three spots in the rankings necessarily improved by more than a team that moved up by one spot in the rankings.  The movement between places has as much to do with the performance of other teams in a similar place in the rankings.

The key injuries that are listed for each team are for top six forwards, top four defencemen and starting goaltenders.  We have a page on the site that lists all injuries, but I'm including the most impactful ones in the rankings simply for a quick reference. 

The rankings will tend to favour teams with strong goaltending.  I have no problem with this since, at the very least, solid goaltending seems to be needed to win a series of any significance.

Teams that win by a larger margin of victory are rewarded.  Again, I fully support this result, since it's preposterous to me to reward a 2-1 win the same as a 10-1 win.  The standings, which give two points for every victory, make this kind of evaluation and it's why I am just as concerned, in the big picture, with goal differential as I am with points in the standings.

Hopefully, this helps address some of the questions you may have regarding the NHL Power Rankings but you are welcome to e-mail me at if you would like additional clarification.

Cabbie on

New York-bred hip hop artist Action Bronson discusses his friendship with Kevin Love, his jumpshot, Mike Napoli's beard, obscure sports references and Derek Jeter's Brand Jordan commercial. More...

He has speed in his DNA, learn more about Olympic champion Donovan Bailey's nephew, Jaden and his success on the gridiron in the latest Powerade 24. More...

© 2015
All rights reserved.
Bell Media Television