Ken Holland, long-standing General Manager of the Detroit Red Wings, once said, “I always look at the standings on U.S. Thanksgiving, right at the end of November. From there, you move in a pack. You’ve sort of shown the winning percentage you’re going to move at.”
Holland is the kind of man from whom you safely can take some hockey advice.
So now that Turkey Day’s arrived, what can be said about the Tampa Bay Lightning and their winning percentage?
One simple method to apply Holland’s Thanksgiving logic is to project forward a team’s point percentage to date.
After 20 games the Bolts have accumulated 20 points, which gives them a point percentage of .500. Assuming this holds, some simple math shows that they will finish the season with 82 points. Considering that since the league returned to play after the lockout, the point cutoff for the playoffs in the Eastern Conference has been in the mid-90s, this means that the Bolts better pick up the point pace, lest they be watching from the proverbial sidelines come the post-season.
This type of projection is often bandied about the water cooler and in the media because it’s an easy and straight-forward way of making a prediction. But it’s not a sophisticated method, and while Holland talks about moving in a pack, there are more than enough surprises over the course of a season to make a fan crave more insight and certainty.
Hockey fans are hardly the only enthusiasts to be interested in how their teams’ season might end. In particular, baseball fans have worked hard to create tools that offer more useful results than mere point projections.
The [Pythagorean Expectation] formula is so accurate, in fact, that analysts have come to explain deviations from the formula as ‘luck’.” Darcy Norman, author of Hockeynomics
One of the most famous and useful of these tools combines fifth-grader math with scoring patterns to make eerily accurate winning percentage predictions.
First popularized by Bill James for baseball, the Pythagorean Expectation formula uses runs scored and runs allowed to calculate a winning percentage. James’ formula has since been adapted to hockey and can be written as:
Winning Percentage = Goals For2/(Goals For2 + Goals Against2)
For such a seemingly simple formula, the Pythagorean Expectation method gives surprisingly good results. As explained in Darcy Norman’s Hockeynomics, hockey-statistic superstar Allan Ryder compared the formula’s predicted results with the actual results for the NHL seasons between 1946 and 2003 and found that over 90% of teams’ winning percentages were explained by the formula. Norman explains that “the formula is so accurate, in fact, that analysts have come to explain deviations from the formula as ‘luck’.”
Based on their results to date of 20 games played, 55 goals for, and 67 goals against, the Bolts should have won eight of their games (in fact they’ve won nine). At this rate, the formula predicts that they will have won 33.5 of their games once all 82 have been played. So based on their scoring tendencies so far, the Pythagorean Expectation formula predicts a bleak outcome for the Bolts’ playoff hopes.
And as an aside, based on the Tampa Bay Rays’ runs scored of 707 and their runs allowed of 647 in the 2011 season, the James’ formula predicts that they should have won 92 games. They won 91.
The Problem with Hockey’s Single Point
But the astute reader may have noticed that the formula predicts winning percentage, not point percentage. So Pythagorean Expectation doesn’t tell the whole hockey story.
For example, the Bolts’ winning percentage is .408 (based on nine wins in 20 games), but their point percentage is .500 – and even the newest puckheads know that it is points that matter.
As a baseball tool, the original Pythagorean formula doesn’t account for single points, but with playoff berths often decided by a narrowest of margins analysts have experimented in attempts to find a formula that is as elegant as the original James’ method, but also takes into account points accrued through ties or overtime losses. Some of this work, particularly that of Ryder is very illuminating, if beginning to move beyond most fans’ tolerance for mathematics.
A simple (if a cheaters’) work-around is to combine the Holland method with the James’ method. The Pythagorean Expectation can be used to calculate predicted wins, while a straight projection can be used to top up the point total with single points earned through overtime losses.
Using this cheats’ model, the Bolts are predicted to earn 34 wins (rounding up from 33.5 determined above), which is good for 68 points. Then, based on a single point earning percentage of .100 (derived from the two single points earned by the Bolts in their first 20 games), a straight projection finds that the Lightning will have earned eight single points by season’s end, bringing the predicted total to 76 points.
This suggests that if the Lightning do not alter their scoring rates, they will end up with even fewer points than the already disappointing tally of 82 points that a straight projection predicts.
While the Pythagorean Expectation model isn’t perfect, it highlights that the Lightning desperately need to reverse their scoring ratio if they want to make the playoffs next spring.
Happy Thanksgiving weekend, all.