The trade deadline has officially passed, which means this is a good time to check in with our power rankings. With just a little over a month remaining in the NHL season, most teams have less than 20 games left to play. We’ve entered the final quarter of the season and the playoff picture is going to solidify over the next five weeks or so.
If you’ve been following along this season, the power rankings are based on a Bradley-Terry model, which is a pairwise ranking system. We feed individual game results into the model and the outputs are measures of how good each team is based on their individual game results. For more on the model, refer to this article. In addition to the margin of victory, the model also accounts for whether a team is home or away. All data is via Natural Stat Trick.
Rankings and other charts
This first plot shows the ranking of the teams. I’ve made some adjustments to the methodology based on feedback from the last time we published the outputs. Previously, we looked at expected goal and goal results. But a reader pointed out that we should make an effort to blend the two and I’ve done that here. The approach is a bit crude but still defensible I think.
I’ve looked at how well expected goal share and goal share predict future goal share. On average throughout the season, expected goal share predicts future goal share about 2.75 times better than current goal share. So, to blend the models, I’ve weighted the expected goal version of the model 2.75 times the goal version. This has the effect of rewarding teams who have outperformed their expected goals and punishing those who have underperformed them to a degree that seems reasonably appropriate based on the predictive validity of each statistic.
On the plot, the blue dot is expected goals, the red dot is goals, and the purple dot is the blended model. Because as we all learned in elementary school, when we mix blue and red, we get purple.
The Golden Knights are still the best team in our model but the gap is closed between them and the second place Tampa Bay Lightning. This is a good example of the impact of using this blended approach. Vegas has underperformed their expected goals by a wide margin this year and so instead of grading them purely on their expected goals, we’ve pulled them back closer to the pack as penalty for their goal results. This adjustment also moves the Bruins and Avalanche ahead of the Penguins giving us a cluster of five teams at the top of the rankings.
At the bottom, no matter what method we use, the Detroit Red Wings continue to be the worst team in the NHL by a comfortable margin. They are tanking with clinical efficiency in Steve Yzerman’s first season as General Manager. Even Tim Murray is probably tipping his hat.
The next plot shows each team’s distribution of results both in terms of expected goals and actual goals.
The Avalanche are interesting here. They have a noticeably sharp peak just on the positive side of 0 in terms of expected goal differential. Put another way, they’ve played lots of close games. They’ve won lots of those games and that’s the reason their goal results look a little better than their expected goal differentials.
The Golden Knights are interesting in the opposite way. They have a slight dip in their curve right at the 0 mark meaning they’ve tended to play more one sided games in either direction. And given they’re the top team in our power rankings, they’ve been on the right side of those games more times than the wrong side even if the results haven’t always followed.
The next plot that we normally look at is strength of schedule and I’ve included it just out of habit but at this point in the season, it’s mostly meaningless.
With teams having played well over 60 games, the difference in quality of competition has smoothed to the point of being almost unnoticeable. What we’re actually looking at in this plot is mostly the distribution of the quality of NHL teams and that’s why we see a similar shape for all teams with slight differences depending on the division.
And finally, we can look at how teams have performed against differing competition. A team with a fit line that slopes down and to the right tends to play worse against tougher competition and/or better against easier competition (which is expected), while a team with a flatter line tends to perform similarly against all levels of competition.
The Lightning have been one of the teams with a flatter line through most of the season but some recent poor performances against good teams have them looking a little bit more like a team that struggles as the competition increases. However, the struggles don’t persist to the extent of a team like the Blue Jackets or the Sharks.
This weekend brings February to a close meaning we’re officially into the stretch run of the NHL season. By now, we know what most teams are. We might see some slight movement over the next few weeks as they incorporate new players acquired at the trade deadline but for the most part, teams in March are who they are with injuries being the only complicating factor.
As of today, the Lightning in the group of five teams at the top of the power rankings along with the Golden Knights, Bruins, Avalanche, and Penguins. All of those teams made at least small additions at the deadline with the Lightning being the most aggressive in trading both of their first round picks and their top prospect in an attempt to upgrade their forward group.
Over the next five weeks, we’ll find out of those moves put the Lightning in any better position come the playoffs.