Determining the value of youth in an NHL lineup
Over the summer, the NHL made waves with its reinvented World Cup of Hockey tournament.
There was plenty of news before the rosters were even picked as the eight-team field included traditional powers in Canada, the United States, Russia, Sweden, the Czech Republic and Finland, with the added curveball of a Team Europe, as well as a squad made up of the best Canadian and American players aged 23 and under in Team North America.
To the surprise of numerous pundits, Team North America served up the most exciting hockey of the tournament and kept pace with the world’s top grizzled veterans.
None of this came as any surprise to the hockey analytics community, which has consistently said that young players are wildly undervalued.
But is that assertion true?
If there’s a bias against youth in the NHL, we would expect to see teams passing over more talented young players for less capable older ones. This sounds obvious enough, but proving discrimination is more complicated than many pretend.
Consider the following analysis…
We took every forward who played in the NHL since 2005-06 (a total of 11 seasons) and separated them into three age groups: 18-22 years old, 23-26 and 27 years and older.
How age, shooting percentage and ‘puck luck’ factor into future performance
We chose these groupings because they roughly mirror the NHL’s salary cap structure, with most 18-22 year olds falling under the entry level salary cap, 23-26 year olds under the restricted free agency system, and 27 is the latest age at which free agency can begin.
Because the NHL has complicated free agency rules it’s not a perfect mapping against salary, but it’s close.
We also set a cutoff of 41 games played in a single season so that players with a short sample of games wouldn’t skew our results.
We began by looking at three measures of offensive performance: (i) points per game; (ii) individual scoring chances per game; and (iii) individual high danger scoring chances per game. As the charts show, both at the median and at the bottom 5% of the distribution, 18-22 year olds outperform their more experienced peers in every measure of offensive performance.
Points per Game
Age | Top 5% | Median | Bottom 5% |
18-22 | 1.02 | 0.48 | 0.11 |
23-26 | 1.00 | 0.44 | 0.11 |
27+ | 0.97 | 0.45 | 0.10 |
Individual Scoring Chances per Game
Age | Top 5% | Median | Bottom 5% |
18-22 | 3.59 | 1.94 | 0.96 |
23-26 | 3.59 | 1.84 | 0.72 |
27+ | 3.29 | 1.78 | 0.63 |
Individual High Danger Scoring Chances Per Game
Age | Top 5% | Median | Bottom 5% |
18-22 | 1.69 | 0.91 | 0.43 |
23-26 | 1.72 | 0.85 | 0.32 |
27+ | 1.59 | 0.82 | 0.27 |
To be fair to hockey’s gatekeepers, they will often concede that younger players are able to compete offensively though struggle with defensive play.
Defensive ability isn’t as easily measured as offense; however, where a player is performing poorly, one would expect his on-ice scoring chances against per game to be higher.
Here the data becomes a bit more ambiguous. The top 5% of younger players do worse than their more experienced peers in terms of defensive play (i.e. they give up more scoring chances per game), as do the bottom 5%. But the median player is still better.
On-Ice Scoring Chances Against Per Game
Age | Top 5% | Median | Bottom 5% |
18-22 | 3.81 | 6.31 | 9.24 |
23-26 | 3.36 | 6.37 | 9.12 |
27+ | 3.49 | 6.65 | 9.17 |
Overall, the data above seem to validate analysts’ claims that younger players are more capable than veterans and should be given more of a chance.
In other words, younger players are victims of discrimination.
Unfortunately, this conclusion is wrong, at least based on the data presented so far, for two reasons.
To begin with, when looking for signs of discrimination, the top end or even median of a distribution don’t tell us a lot.
Anyone who is capable of generating close to 40 points over the course of an 82-game season and gives up less than 7 scoring chances per game belongs in the NHL. The fact that younger players do better as a group has more to say about the younger players in our specific data set than anything to do with player selection generally.
You may have views on whether Nikita Kucherov (who was 22 at the start of last season) is better than Jonathan Toews (27 at the time), but there’s no 30-team NHL in which both aren’t playing first line roles.
To raise a suspicion of discrimination, one needs to look at the bottom end of the distribution. Only when observing the guys teams pick for their final roster spots do we truly get a sense of the potentially irrelevant factors that drive teams’ decisions.
At this point, however, we don’t have a good sense of who the “worst” players are.
What analytics can tell us about the role of fighting in hockey
The reason for this is the aggregate numbers tell us nothing about the specific individuals involved.
For example, let’s imagine the worst 18-22 year old was in the bottom 5% of both offensive and defensive play, meaning he generated less than 0.16 points per game and gave up more than 9.24 scoring chances against.
Meanwhile, in the 27+ sample the worst offensive and defensive performances were provided by different people. In that case, it would be possible to find a player who provided bottom 5% offence of 0.10 points per game but top 5% defence of 3.49 scoring chances against.
It’s perfectly reasonable for a team to pass on a player who gets five additional points over the course of a season in favour of one who can suppress 470 additional scoring chances. Indeed this is what teams claim they’re doing by giving defensive roles to veterans.
To figure out if there’s any weight to this claim, we looked at time on ice in order to determine what a “bubble” player in the NHL has to offer. The thinking here is that if you’re trying to identify a bottom tier player, you need to look at what each group of players does with bottom tier minutes.
As it turns out, the bottom 5% of all NHL forwards in our sample logs 8.52 minutes per game. We set an ice time cutoff at this amount for each group of players, and because we were working with a much smaller sample, we looked only at a median number in each age group.
The results of that analysis are in the table below.
Median Performance at ≤ 8.52 Minutes Per Game
Age | Points | Ind. Scoring Chances | Ind. High-Danger Scoring Chances | On-ice Scoring Chances Against |
18-22 | 0.14 | 0.68 | 0.35 | 3.29 |
23-26 | 0.10 | 0.59 | 0.27 | 2.98 |
27+ | 0.09 | 0.55 | 0.26 | 2.89 |
As you can see, there does appear to be something to the claim that younger players provide an offensive bump but suffer from defensive lapses.
They’re not that much worse, but the difference is enough that, all things being equal, a team might be wise to forego a bottom-tier younger player’s modest increase in offence for a veteran’s equally modest improvement in defence.
However, because the NHL’s Collective Bargaining Agreement heavily favors veteran players, all things are not equal.
The table below shows the cap hit incurred by NHL teams during the 2015-16 season for the bottom 50 forwards in terms of ice time (out of 377 total) who played a minimum of 41 games.
A couple of things become clear. First, more than half were 27 and older; whereas, only 4 among that group were 18-22 years old.
Hockey’s great shot volume debate
In addition, while all of the younger players earned above the league minimum ($575,000), the fact that they were subject to the NHL’s entry level salary cap meant that their downside risk was far lower, meaning a maximum base salary of $925,000, which was significantly less than the salary of the highest paid 23-26 year old in the group ($1,600,000) and the highest paid 27+ year old ($3,333,300).
Salaries of Bottom 50 NHL Forwards
Age | No. of players | Average | Median | Min. | Max. |
18-22 | 4 | 847,080 | 881,670 | 700,000 | 925,000 |
23-26 | 20 | 760,780 | 658,750 | 562,500 | 1.6M |
27+ | 26 | 1,447,400 | 1.2M | 575,000 | 3,333,330 |
These numbers don’t tell the full story because younger players are often on two way contracts, meaning if they’re sent down to the minors they get paid a much lower salary. In addition, they don’t have no movement clauses. As a result a younger player who isn’t working out at the NHL level offers far greater roster flexibility than a veteran who is underperforming on a longer-term deal.
So while the average salary numbers don’t look that much different, filling marginal roles with 18-22 year olds gives a team both option value and mitigation of downside risk.
When viewed in this light it becomes clear that NHL teams are overpaying for veteran fourth liners who provide a small improvement in defensive play and worse offensive play.
Although performance alone doesn’t support the discrimination claims implied by many analysts, when you add in the business realities of the game it does appear that teams should tend toward more younger players.
While there may be valid reasons for not rushing along a younger player’s development and allowing him to dominate at the minor league level before bringing him up to play in the NHL, it’s important for teams to recognize the costs of that exercise.
The Department of Hockey Analytics employs advanced statistical methods and innovative approaches to better understand the game of hockey. Its three founders are Ian Cooper, a lawyer, former player agent and Wharton Business School graduate; Dr. Phil Curry, a professor of economics at the University of Waterloo; and IJay Palansky, a partner at the law firm of Armstrong Teasdale, former high-stakes professional poker player, and Harvard Law School graduate. Please visit us online at www.depthockeyanalytics.com