Physicality in Hockey: Is Hitting Valuable?


How important is physicality for an NHL team? To generalize, this is a question that divides the advanced stats community from a lot of traditional ‘purist’ thinking. On one hand, it’s common to hear purists talk about the importance of grinding forwards who finish checks and intimidating defensemen who level their opponents at any opportunity. In contrast, analytics types point to hits that hurt possession rather than help it when the hitter takes themselves out of the play without gaining the puck.

In this post, I try to assess the value of hitting based on the data currently available. First, I’ll take a brief look at the theoretical direct impact of a hit on the play in which it occurs. Second, the majority of the post will examine whether hitting effectively ‘wears down’ the opposition in a way that helps the hitting team as the game progresses. Both of these sections suggest that hitting has limited value for winning hockey games. I’ll end with a discussion of this data’s implications, its limitations, and possible next steps.

Immediate impact of hitting

The first effect of a hit is obvious: it separates an opposing player from the puck. If done skillfully, this can help a team gain possession of the puck and create offense. However, theory suggests that while this could be useful, it is not necessarily a positive move. Since the hitter is rarely in a position to gain control of the puck himself after a hit, the best he can hope for is to create a 50/50 loose puck in which one of his teammates gains possession. In cases where the hitter is ‘finishing his check’ after an opponent makes a pass or shot, the likelihood that his team will gain possession is even smaller; the hit may pressure the opponent to make a mistake, but the opponent has still already made a play and has probably sent the puck to a teammate as planned. In contrast, the hit, especially if it misses, can take the hitter out of position and give the opponent a better scoring opportunity.

Overall, hitting is better than letting the opponent keep possession, but it carries risks and isn’t obviously more beneficial than playing positionally and forcing a turnover in another way. This is a situation that could be studied more accurately by tracking who gains possession after a hit, but I do not have the time to conduct that tracking nor do I know of a case where it has been done and the data made public

Game-long impact of hitting

Advocates of hitting would claim that in addition to the immediate effect of a hit, it can also affect future play. Specifically, it can diminish the ability of the opponent. This is ‘wearing them down’, and weakening the opponent through pure physical pain and fatigue. This impact can also be psychological: an opponent may make worse plays or avoid the ‘dirty areas’ to avoid being hit again.

This type of effect is more interesting to me because it is based in concepts that are rarely explicitly stated in sports. We use terms like ‘grind them down’ or ‘tough to play against’ as euphemisms for hurting the other team so that they’re less likely to win. It’s easy to see professional athletes as machines with set, constant abilities, but this simply isn’t the case. Players are humans, not video game characters with hard-coded stats. They’re just as prone to fatigue or psychological effects as anyone else. Indeed, one of my favorite parts of Justin Bourne’s writing is when he discusses the reality that players don’t like getting hit and will sometimes try to avoid it, even if it means not making the best possible play for the team.

These ‘long-term’ effects of hitting can’t be measured in a single statistic, but they can still be studied. They are beneficial to a hockey team if and only if they help the team score goals and win games. If hitting doesn’t lead to goals then it may be very entertaining but is not a strategy that a team should pursue.


In this post, I examine whether teams get ‘worn down’ or otherwise have their ability diminished over the course of a game by being hit. The data to study this is somewhat limited but not totally unavailable. First, I looked at the basic count of hits taken by a team in any given game (this and all other data in this post come from War on Ice).

The most immediate data to compare it to would be 3rd-period performance since that is when we would see the effects of being worn down by physicality. However, this won’t work because taking hits means having the puck, so hits taken correlates with possession metrics

Instead, I calculated the change in Corsi For % for this team from the first period to the third period. This is obviously an imperfect measurement, but I believe it works to test the possible influence of hitting. Because this is by definition an effect that takes place over the course of a game, it should be measured by its effect between the beginning and the end. In the first period, the team has not been hit and is playing to its full healthy ability, regardless of the hitting tendencies of their opponent. The physical team’s strategy, if effective, would come to fruition later in the game after they have had the opportunity to hit their opponent. If a team is worn down by hitting, we would see their play worsen over the course of the game.


So, what does the data say? In the below graph, I plot at every game in the second half of the 2014-2015 season. The x-variable is the raw ‘hits taken’ by a team over the course of the game. The y-variable is the difference between Corsi For % in the 3rd period and Corsi For % in the 1st, i.e. CF%­3rd – CF%­1st. (All of my data and analysis are available in the link at the bottom of this post)Hits Taken and CF% Change

As you can see, there’s no clear relationship. The R2 value is just 0.0017, which means that being hit explains basically none of the variation in how a team performs between the first and third periods. This suggests that hitting does not have its desired long-term physical effects. A team’s third-period performance has nothing to do with how much they’ve been hit during a game.

What about the effect of physicality on shot quality? Maybe a player who has been hit a lot will shoot from the perimeter but be less willing to go into the ‘dirty areas’ out of a desire to not be hit. To test this, I looked at the change in Scoring Chances For per 60:

Hits Taken and Scoring Chances Rate Change

Again, we see no impact. This holds true regardless of the sample or the exact statistic measured. While the above graphs looked at only half a season each to avoid (or at least limit) overplotting, I also looked at the combined data from the past 5 seasons. In addition, I expanded the number of performance metrics I looked at to see if any of them correlated with hits taken. Finally, I repeated the procedure while only looking at the last half of each season (excluding the lockout-shortened season). I did this to test whether hitting had a larger affect later in the season as teams get more tired and are more likely to be dealing with injuries. The results are as follows:

Y-variable G+/- CF% CP60 C+/- HSCF% HSC+/- HSCF60
Full Sample
-0.039 -0.027 -0.004 -0.027 -0.020 -0.025 -0.027
Game 42+ Only R-Squared -0.040 -0.030 -0.014 -0.030 -0.015 -0.015 -0.023
Y-variable HSCA60 HSCP60 SCF% SC+/- SCF60 SCA60 SCP60
Full Sample
0.008 -0.014 -0.034 -0.033 -0.044 0.005 -0.029
Game 42+ Only R-Squared -0.002 -0.018 -0.039 -0.041 -0.056 0.002 -0.040

(Definitions of all stats shown here are listed at the bottom of the post)

Across the board, we see no indication that being hit leads to a decline in play over the course of a game. I’m honestly surprised that all of these values were so low, as I’d have expected one or two to have a closer correlation just out of chance. Instead, it seems that playing against a physical team does not explain any improvement or decline in play over the course of a game


To be clear, this is far from perfect analysis. First of all, the data is not ideal. Hits taken is not a particularly reliable stat and is known to vary between arenas. Even if every team’s scorekeeper had the same objective standard for what constitutes a ‘hit’, hits taken would not distinguish between a soft tap along the boards and a bone-jarring collision. Essentially, ‘hit quality’ is ignored. Second, while I believe that calculating the difference between first and third period play is a valid course of analysis, it’s not perfect and could likely be improved upon. I tried to limit the influence of other factors, such as the general correlation between hits taken and positive possession, but it may still be flawed. Nothing here is absolute proof that hitting never matters, and it shouldn’t be construed as such.

That being said, this data does point in a single direction: hitting is not a particularly valuable strategy, and does not have an effect beyond the play in which it occurs. The conventional wisdom that you need to ‘set the tone’ or ‘grind them down’ just doesn’t hold up in any way that’s visible in the actual results of the game.

This can potentially guide decisions for both coaches and GMs. Coaches may want to teach players to prioritize playing smart positional defense rather than finishing a check and removing themselves from the play. GMs may benefit from constructing a team with skill on all four lines rather than depth players known for physical play. While traditionally teams have defensively-minded physical line, a team is more likely to win hockey games with a fourth line focused on possession and contributing on both ends of the ice. A fourth line of, say, Brad Boyes, Sean Bergenheim, and [insert your favorite AHL prospect] is probably more valuable than a line that racks up hits but gives up goals. This is especially valuable right now, since most other GMs in the league current do value players that hit. The mismatch in player evaluation creates an opportunity for trades that improve a team by replacing physical players with skilled ones. (Every post in this blog is actually going to be a veiled case for the Islanders trading Matt Martin)

There are a few ways this work could be advanced. First, more could be done to study the immediate impact of a hit, and how likely the hitter’s team is to gain possession of the puck. I tried to look at this by studying the next event after a hit in the NHL play-by-play files, but this wasn’t valuable because it excluding turnovers that happened after a hit but before the next shot. Ideally, this would be studied by a full tracking project. Second, this analysis focuses on the team level, and it’s possible that a study of individual players would reveal skaters who are more likely to experience a drop in play after being hit. Finally, I’m curious about the results of studying the ‘cumulative hits’ that a team takes over the course of a season rather than in a specific game, but I did not dive into this area because even if this relationship did exist, it would not suggest that any specific team benefits from having its players do the hitting.

Questions for you (yes, you!)

As you’ve probably figured out, I am fairly new at hockey analytics. These posts are meant as a learning experience for me, not ‘my revelations of the fundamental truths of the universe’. As such, here are some of the questions that I’ve had during this work that I would love to discuss with people more experienced than me:

  • Overall, is any of this convincing? How could it be improved?
  • Just how unreliable is hits taken as a statistic, and is there any better measure of physicality?
  • Is there a way to improve the y-variables I studied, either by selecting different metrics or by calculating the change over the game in a way other than my ‘3rd period – 1st period’ method?
  • How would you further fix the overplotting in the above graphs? I read some things online and made some improvements, but not as much as I’d have liked


Download the full dataset and analysis here

Appendix: Statistic definitions from War-On-Ice

  •  G+/-: Goals differential (goals for – goals against)
  • CF%: The percentage of on-ice shot attempts (on goal, missed, or blocked) taken by the player’s team
  • CP60: The rate of on-ice shot attempts (on goal, missed, or blocked) taken by the either team per 60 minutes of play
  • C+/-: plus-minus for on-ice shot attempts
  • HSCF%: High-danger scoring chances for, percentage of total
  • HSC+/- Scoring chances differential
  • HSCF60/HSCA60/HSCP60: High-danger scoring chances for/against/total per 60 minutes
  • SCF%: Scoring chances for, percentage of total
  • SC+/-: Scoring chances differential
  • SCF60/SCA60/SCA60: Scoring chances for/against/total per 60 minutes



Passing Project Analysis of UFA Defensemen


Ryan Stimson recently shared the results of his passing project, a great piece of work that’s collected some really interesting data. Spencer Mann followed up by creating some visualizations to summarize the key findings for each player. This data deserves to be used more, so I wanted to try it out on a fairly small question: what does it tell us about the currently available free agent defensemen? As Ryan lays out in his posts, passing is a significant aspect of offensive production, so this data may offer more information than shooting metrics alone. The data does not cover every game last season, so there are some limitations. Anton Volchenkov and David Schlemko are both excluded because they both have less than 100 min of tracked 5v5 time. (Want to help fix that for next season? Ryan is currently looking for more volunteer trackers, which will be a huge part of how much data is collected moving forward) Regardless, the data as it currently stands has some interesting points to show.

Glossary of Terms

First, a quick summary of what each bar in the chart represents:

Glossary from Spencer Mann

Top Four Defensement

Visnovsky Zidlicky

Lubomir Visnovsky is a great offensive contributor with elite zone entry numbers. He’s also excellent on the power play and would provide significant offensive output for any team that signs him. At 38 years old, he is likely looking for a one-year deal, so a team with cap space should be able to acquire him without a long-term commitment. The primary concern, and likely the reason he has not yet been signed, is injuries. Visnovsky has missed significant time due to injuries in the past few years and did not finish the Islanders first round playoff series due to a concussion. He’s not a shut-down guy and would likely not play a full 82 games, but he would definitely provide value for a team that expects to make the playoffs and is looking for an offensive boost.

Marek Zidlicky’s statistics are very impressive, and frankly, much higher than what I expected. His graph shows an inverse from Visnovsky in terms of which categories he is elite in compared to those in which he is just very good. In particular, Zidlicky is involved in a ton of scoring chances, both by pure volume and by his involvement compared to teammates. If there’s an area measured here where he struggles, it’s his limited involvement in controlled zone entries. Like Visnovsky, Zidlicky is 38, but the statistics here show serious offensive output

Ehrhoff Franson

Christian Ehrhoff has had less time tracked than the previous two UFAs, but he did very well in the games that were tracked. Many of his stats are quite similar to Zidlicky but a bit lower overall, especially SC SAG/60. On the other hand, he was much more active in created controlled zone entries. Overall, this backs up a general assessment of Ehrhoff as a very good but not spectacular defenseman.

Franson, like Ehrhoff, has limited tracked time, and his numbers are very interesting. Franson has the greatest disparity in his rankings between different statistics: the middle four are great while the right two columns are awful. At 27, Franson is unlikely to significantly improve in the future but should remain productive for several seasons.

Possible Depth Defensemen

Gleason - Canes Gleason - Caps

The talent pool for UFA defensemen drops off considerably after the top 4 options. Tim Gleason’s stats across two teams suggest a third pairing guy. The distribution of the stats is similar to Franson’s, but Gleason’s stats are worse across the board.’s article on remaining free agents called him a “cheap, veteran 3rd pair”, which seems about right.

Meszaros Hejda

Andrej Meszaros has almost identical stats as Gleason though he is a bit stronger on controlled zone entries. Jan Hejda’s passing summary is flat out bad. At 29 years old, it does not seem like he would add much.


While there’s limited passing ability from defensemen still available on the free-agent market, there is definitely still some talent to be had. That is particularly true for teams looking for a veteran player for a year or two rather than a younger player to commit to long-term. One interesting finding from these summaries is that the two older players – Visnovsky and Zidlicky – did better than would be expected by their general reputation among fans. Given the focus on passing in this data set, I’d guess that these vets rely on passing as a way to produce offense while limiting skating, and thus conserving energy. SportLogiIQ just had a post on Duncan Keith’s playoff performance and suggested that he conserved energy by passing rather than rushing the puck. It would be interesting to examine the data further to see if the same holds true across the regular season for Visnovsky and Zidlicky.