After reading Michael Lewis’ Moneyball, I was inspired to do some research to see if there were similar metrics in hockey that provided insight as to which players are the best value. To answer this question, it is important to know which metric correlates best with teams’ final rankings for a season.
A highly correlated metric would show the rankings matching almost exactly to the final standings–that said, the overall order of teams from best to worst should be the same order as the metric from best to worst.
After some research, I found the metric that correlates the most with final ranking is goal differential, which may be obvious, as the way you win games is by scoring more goals than the other team. That being said, judging players by the number of goals they score and the goals they allow (for goalies) may seem to be the best way to value players, but in reality, it is nonsensical, as players who receive more ice time have a huge advantage in this respect. This leaves undervalued players who haven’t received much ice time and therefore haven’t scored many goals to be constantly overlooked.
This leads us to the next best metric following goal differential, which is a team’s PDO value, or their total shot percentage plus their total save percentage. Contrary to popular belief, I found that PDO value may actually be useful in predicting a team’s final ranking, rather than how lucky they are.
PDO ranking has a total correlation of .817 with final ranking, which is second to goal differential. It also has a .814 correlation value with goal differential, indicating that teams ranking high in PDO tend to also be ranked high in goal differential.
Below is a reference chart including the numbers I just explained.
We can conclude that possibly the best way to predict a player’s worth is by looking at their S% (and SV% for goalies) as it will help the team’s PDO value go up. If the average shot percentage is 9%, and the average save percentage is 91%, we are looking for players with stats above these averages, as they would help bring the team average up above the average PDO value of 99.99.
Using regression analysis I found that for every .074 points above the average PDO value of 99.99 a team finishes with, they will score 1 additional point that season above the average, which is 92 (I have used this to “predict” teams total points from prior years, and it works fairly accurately, only missing by an average of 7 points). Using this, you can put values on players based on their S% and SV%.
Using the formula [[[[171+S%/20]+91]-99.99]/.074 will show us how many extra points adding said player to your team will earn (assuming you have a team full of players who shoot 9% and a goalie who with a save percentage of .910).
We can also determine, that since every team has the same cap value of 81.5 million, each team should be paying $885,869.57 per point. You can multiply that value by the value you get from the equation above to predict a player’s worth in salary.
Putting this into perspective, the 2018-19 Tampa Bay Lightning, one of the most talented teams hockey has seen in years, recorded a team shot percentage of 12.2%, one of the highest ever recorded. The team was carried mainly by expensive players like Brayden Point, Steven Stamkos, and Nikita Kucherov, who shot 21.5%, 19.2%, and 16.7%, while many of the other players shot with an average or even below-average percentage. If a team can grab a number of low budget players who all shoot somewhere around 12%, they may see a similar caliber team to the 2018-19 Tampa Bay Lightning.
Below are the top 10 most undervalued players in the NHL from the past 3 years (and some others who I thought deserved a mention). The choices are mainly players who have high shooting percentages and a high number of shots, but play fewer games than the average player.
In the chart below, players highlighted in blue are set to be RFA’s next year, and players highlighted in red are set to be UFA’s next year.
Let me explain some of my picks.
Eric Haula–Haula has been a mass producer for the past three years, especially for only seeing the ice 64% of the time he’s available. Shooting close to 300 shots, he has been a consistent shooter, scoring 15% of his shots. Not only is he under-played, but he is underpaid as well.
Paul Byron–While Byron may seem like a key player for the Canadians, he is actually only seeing the ice 58% of the time he’s available. With a 16% shooting percentage, there is no reason Byron shouldn’t be seeing more ice time.
Sven Bartschi–Bartschi has only hit the ice in 85 games in the past 3 years, but has shot 17% and produced more than many players who have seen double the ice time that he has received.
Boyle, Garland, and Sanford–While these players are hitting the ice most of the time they’re available, they are clearly underpaid based on their S% and production rates.
Marleau and Thornton–Of course these are both star players, but their current age has caused many to see them as washed up or players way past their prime. In my opinion, this has caused them both to be currently undervalued, as they still producing at incredible rates; In fact, Thornton is shooting his best since the 2000-01 season.
Alexander Kerfoot–While many say Kerfoot needs to work on his shooting, his stats say otherwise. Kerfoot has shot 15% on nearly 300 shots in the past 2 years (only been in the league for that long) and while he sees proper ice time, he should be seeing even more respect from the league and possibly better line positions on the Leafs.
By: Sam DiSorbo