Size and Age in the NHL, Pt. 2
How do teams use their players at different strength states?
So, so, so, I’ve looked at which teams were the oldest, biggest, etc., but what does that all mean? Originally, I planned to use this post to examine what impact if any age, the various aspects of size, and draft position had on team success. However, as I started to work through the data, I thought it made sense to also break out the impact by strength state. Before doing that though, I would want to establish that teams used their players differently based on the biographical details I’ve looked at so far. I’m excluding nationality from this and future posts in this series as it’s more complex and worth doing a deep dive into on its own. In this post, I’ll simply break down the weighted averages of each team season and examine the difference between the all strengths average and Even Strength (EV), Power Play (PP), and Penalty Kill (PK) weighted averages. This will be a short post so I can move onto how much this matters more quickly.
I followed a similar methodology as in my first post. Evolving Hockey helpfully includes Time On Ice breakdowns by strength state in their GAR tables. I used that to calculated the weighted average for each team by strength state, then subtracted each from the overall weighted average and plotted the density of the differences against each other. As before, my a link to my Github code will be included at the end.
Days On Earth
One pattern that is apparent here and repeated in the other metrics I look at in this article is that there is much more variance in the difference between PP and PK weighted averages than EV. This makes sense, I think and generally matches the impressions one might get by just watching a lot of hockey. Teams tend of have set core power players and penalty killers so both units are drawing from a smaller populations than the team at EV. Furthermore, coaches have a reputation for trusting veterans with defensive assignments over rookies. Veteran players also seem less likely to play on the power play as they begin to decline and lose offensive effectiveness.
EV PP PK Min. | -107.010 | -866.18 | -474.04 1st Qu.| -40.432 | -185.37 | 57.88 Median | -21.152 | -20.10 | 223.40 Mean | -22.547 | -17.73 | 255.57 3rd Qu.| -3.506 | 150.61 | 430.13 Max. | 66.411 | 724.24 | 1231.26With all that in mind, it’s clear if we look at a summary of differences that teams definitely skew older on the PK, a bit younger on the PP, though actually the distribution is only slightly shifted younger, and, surprisingly, a bit younger at EV. I would account for this by guessing that older players are predominantly bottom six forwards who get limited minutes at even strength and get most of their ice time on the penalty kill.
*Note: Substack apparently doesn’t support embedding tables, my solution in this post is to try and format text like a table. In the future I may just include screenshots of a table instead.
Height
As I noted in the previous post, NHL teams just don’t vary much size-wise and that holds true for the differences between. Notice that although the is a clear difference is ice time distribution between taller and shorter players with taller players getting more PK time and shorter players getting more PP time, the weighted difference from all situations for the vast majority of team seasons is less than an inch.
EV PP PK Min. | -0.0967 | -1.6073 | -0.6814 1st Qu.| -0.0106 | -0.4245 | 0.0276 Median | 0.0076 | -0.2437 | 0.2224 Mean | 0.0071 | -0.2418 | 0.2371 3rd Qu.| 0.0267 | -0.0433 | 0.4287 Max. | 0.0948 | 0.5726 | 1.6749Although reach gets mentioned fairly often as an asset defensively, especially when talking about players such as Zdeno Chara, it doesn’t appear that coaches leverage this significantly in their PK lineups. The difference in height between the average PP unit and the average PK unit is less than half an inch or roughly 0.7% of the average height. In fact, I suspect that the difference in height is mainly due to defensemen being taller than forwards on average and being weighted more heavily on the PK since they make up 50% or more of the players on the ice compared to 40% at even strength. I suspect the inverse is true for the power play since more teams have shifted to four forward one defenseman setups. I plan to look into the differences between forwards and defensemen in a future post, but this is something to keep in mind when looking at weight as well.
Weight
Here we see a similar pattern as appeared with weight, although this time it is somewhat more significant with the average difference between a PP unit and a PK unit being roughly 3 pounds (about 1.5% of the average weight). Again though, I suspect the difference can be accounted for by the positions of the players who play PP/PK rather than coaches selecting the largest and smallest players for these units (although it is important to clear the net front).
EV PP PK Min. | -1.1050 | -8.8458 | -6.3106 1st Qu.| -0.0969 | -2.9897 | -0.3035 Median | 0.0720 | -1.6092 | 1.4899 Mean | 0.0644 | -1.5765 | 1.3848 3rd Qu.| 0.2296 | -0.1338 | 2.9767 Max. | 1.0200 | 4.9725 | 9.7468It is very interesting to notice the symmetry of the differences between the PP and PK from all situations for height and weight. The distributions are almost perfect mirrors of each other, which, may not be significant, but it is nice to look at on a chart.
BMI
There is not much of a difference between the BMI of players on the power play and PK and all situation, just a wider distribution. This makes sense given that height and weight were both skewed in the same direction for both strength states.
EV PP PK Min. | -0.0691 | -0.6422 | -0.7228 1st Qu.| -0.0123 | -0.1830 | -0.1604 Median | 0.0024 | -0.0438 | 0.0204 Mean | 0.0033 | -0.0334 | 0.0116 3rd Qu.| 0.0177 | 0.1127 | 0.1822 Max. | 0.0897 | 0.6350 | 0.8180There’s not much to analyze here, except to notice that the distribution seems to be roughly equivalent for each strength state with the difference being mostly variance. PK has a wider variance than PP, but that’s because there is less cumulative ice time for PKers than PPers since there are fewer of them on the ice.
Draft Position
Lastly, let’s take a look at draft position. Here we see a very clear difference between each strength state with PP being tilted towards higher draft picks and PK being dominated by lower draft picks. This would seem to be driven by two things, teams selecting offensive players earlier in the draft due to offense being easier to project, and teams giving their higher draft picks more opportunities on the power play, possibly to justify their draft position.
EV PP PK Min. | -2.0473 | -52.284 | -31.515 1st Qu.| 0.2939 | -22.831 | -1.557 Median | 1.0846 | -14.318 | 6.474 Mean | 1.1171 | -14.417 | 7.241 3rd Qu.| 1.9181 | -6.334 | 14.861 Max. | 5.2662 | 19.307 | 49.625One thing that stands out here, but may not be especially significant is that the EV average is for slightly later than all situations. I would attribute this mainly to teams giving their high draft picks a better chance of making the roster earlier in their careers, but sheltering or limiting their minutes at even strength and giving them power play opportunities.
Trends, Conclusions, and Next Steps
It’s interesting to notice two patterns in the trends since the 2007-08 season. In four of the metrics I’ve looked at, there appears to be increasing specialization between special teams units with the difference between PP units and PK units generally growing over time. This is true for age, height, weight, and draft position, though 2020-21 bucks the trend for a few of these. As mentioned previously, I think 4 forward power play units account for a lot of the trend on the PP side, but it’s worth investigating more later. On the PK side, it will be interesting to see if this is due to penalty killers getting larger overall or just defensemen.
BMI stands out for the way the trends for PP and PK have moved in tandem and contrary to the trend at EV. I’m not sure how to explain why denser players would be getting larger shares of special teams time as seasons go by, but this is a curious development.
Next time I’ll dive into what each of these metrics predict for regular season success, broken down by overall results and each strength state.
Fin.
A huge thank you goes to Evolving-Hockey for providing team, player, and even play-by-play data in a clean, easy-to-use format. When I did my last major project I was combing through the NHL play-by-play data up till 2017-18 and it was quite messy. Fortunately, Evolving-Hockey absolutely has the best quality of any of the major hockey stats sites I’ve looked at and it makes hobbyist work like this so much easier. For that reason, although I’ll use their data and make some of my data available, I won’t include anything from their site that can’t also be found on NHL.com or other free sites.