The Detroit Red Wings made headlines recently when Ken Holland signed 28 year-old forward, Justin Abdelkader to a 7 year contract worth $4.25M per season. There was a fairly visceral and predictable reaction from the hockey stats community, noting the high shooting percentages he has been enjoying over the past year and the decline in performance we’ve historically seen from forwards aged over the age horizon of Abdelkader’s contract. However, the piece of the discussion that really struck home for me was comments around the wear and tear to Abdelkader’s body given his physical style of play.
As a community, we’ve done a fair amount of work on how players age over time, coming to the conclusion that typically player production peaks in the 23-26 window, then typically decline steeply as players enter their 30s. However, to my knowledge there hasn’t been much work performed on the differences in player aging between more physical and less physical players.
Under the current CBA, GM’s enjoy a relatively cost controlled structure for players up until the age 27, so when it comes to player aging and skill degradation the real contract risk applies to players 27 and over. With that in mind, I used waronice.com to look at players active in the NHL in both their 25 and 26 year-old seasons as a baseline to eliminate a large number of fringe NHLers who would be unlikely to be awarded large deals at age +27. The time frame used was the 10 year period of the 2005-06 to 2014-15 regular season. I took this population (n=535) and split it into three groups: players with a high number of hits taken and received per game , or Total Hits, in their 25 and 26 year old seasons (>2.6 Total Hits per game), players with a medium number of Total Hits per game (1.75-2.6 Total Hits per game), and players with a low number of Total Hits per game (<1.75 Total Hits per game).
Running a regression, I found that for the entire population, Total Hits per game for the 25 and 26 year-old season was a statistically significant variable when predicting games played from ages 27 on wards (r2 0.07, p-value 2.6-10).
I then looked at how many games, on average, players from each cohort played from their age 27 season on wards:
As we can see, the players with a low number of Total Hits played almost as many games, on average, as the players from the medium and high Total Hit groups combined. On reflection, since this population includes a large number of current NHLers, I thought it would be worthwhile to look at just the players who were no longer active in the current 2015-16 NHL season to see if there was any difference in the distribution:
The distribution is virtually identical to the larger population when we exclude current NHLers.
Next I wanted to look at differences in how players dropped out of the NHL as they age, by low, medium, and high Total Hit cohorts. The first graph including current NHLers:
..And again, this time excluding current NHLers:
As we can see, players from the low total hit cohort consistently enjoyed longer NHL careers than their more physical peers.
Finally, when discussing this project with Garret Hohl, he brought up a really good point. Does aging of physical players differ if you look at forwards vs defensemen and high minute player vs low minute players? More pointedly, we know that teams often fill the lower end of their roster with replacement level “energy” players, who hit with a high frequency, so maybe it is this end of the roster than is turning over more rapidly.
To dig into this further, I combined the medium and high Total Hit cohorts, then split each group in half based on high ice-time and low ice-time based on the median of the average 5-on-5 ice-time from the 25 and 26 year-old season (Forward median TOI: 11.2, Defense median TOI: 15.13). I included a fifth column for the low Total HIT cohort:
As we can see, the forwards and defensemen with high ice-time last longer than their peers with low ice-time, but none of the four groups lasted as long as the low total hit cohort. It appears that just because you’re top 6/4 guy at age 25-26 doesn’t mean you’re immune to the impacts your physical game will have on your body.
Closing Thoughts
None of this should be considered particularly earth shattering. It’s pretty intuitive that the toll heavy hitters take on their body can result in them breaking down faster than their less physical players.
However, this doesn’t always seem to factor into the decision making of General Managers, who often sign big deals for guys that can make their opponent pay on the boards and the score board, despite the lack of data showing a strong correlation between hitting and winning. It isn’t a new phenomenon to see GM’s award large contracts, with lots of term to players like Justin Abdelkader, who have shown an ability to score goals (career high 23 goals in 2014-15) and provide physical play (4.0 total hits per game over course of his career). At age 28, Nick Foligno (3.9 total hits per game over course of his career) cashed in on his first over 20 goal season, with an extension for 6 years at $5.5M per season.
We will see whether these deals will turn into wins for the GM’s who signed them, or whether they will be remembered in the same light as the David Clarkson contract (career 4.8 Total Hits per game), who was famously awarded a $36.7M contract by the Leafs after his 15 goal season in 2012-13. Since signing that deal, Clarkson has scored only 15 goals in the two plus seasons since that deal was signed, and has missed a predictably significant number of games to injury.
At some point we may start to see general managers apply more caution when signing long-term contracts players in their late 20s known for their physical play, but that day is clearly not today.
Hockey Graphs 
I’d be curious to see what happens if you split by points/60 instead of ice time. I am just curious how a hitter compares to an equivalently offensively talented non-hitter. My hypothesis is that the non-hitters generally have more offensive talent than the hitters and offensive talent is a more rare talent than hitting so less likely to get replaced by a cheaper younger player.
I tend to agree with that hypothesis, David. I looked at this very briefly in a slightly different way, focusing on P60 aging trends for the same low, medium, high buckets. As you may expect, the low hitting group consistently had the highest p/60 as the players age, but I’ll need to dig deeper in future work.
In that same vein, it would also be interesting to see if your final comments are accurate. If players who throw a lot of hits get bigger contracts than their p60 equivalents that don’t. I’d suspect they are, but it’d be interesting to know for sure.