Doogie2K
Jul 21st, 2009
5:14PM UTC

A Correlational Analysis of the Relationship Between Hitting and Standings Points

Introduction

Recently, the prolific Jonathan Willis posted a study suggesting that hitting actually had a negative correlation with winning, a result that seems counter-intuitive, to say the least. However, his analysis looked only at teams on the margins, the five highest- and lowest-hitting teams in each of the last nine seasons, without accounting for the twenty in the middle. The purpose of this study is to complete this analysis by examining the relationship between the hit statistic and points for all 30 teams in each of the last eight seasons. Additionally, I will look at the cumulative relationship between hits and points in the four seasons before and after the lockout, keeping the two periods segregated due to changes in play style and point awarding after the 2004-05 lockout.

My hypothesis, before performing the analysis, is that I do not expect there to be a strong correlation between the two variables in any one given season, though it is plausible that there will be a weak correlation in one direction or the other in the multi-year analysis. My reasoning is that last year alone, there were good teams that hit (Boston), bad teams that hit (Tampa), good teams that didn’t hit (Detroit), and bad teams that didn’t hit (Colorado), just showing up in the margins, and that doesn’t account for all the teams in the middle that are likely to run the gamut in terms of the hit statistic. While hits are useful in separating players from the puck, Big Hits are also just as likely to put you wildly out of position, and when you miss that Big Hit, then you’re not only out of position, but the guy with the puck is 20 feet behind you before you even start to turn around. Furthermore, while hitting is a good way to gain puck possession, you still need to put the thing in the net, and there’s a whole pile of important factors affecting that which have nothing to do with hitting.

Methods

The hit and point data will be culled from the NHL.com statistics portal, then accumulated and analyzed with SPSS v16.0 using a Pearson linear regression analysis with a significance threshold of p < 0.05.

Limitations

The question Willis seemed to be asking was whether hitting and physical play had an effect on winning. I do not believe the data available can tell us this for two reasons. One, the hit statistic is subject to recorder bias, due to both home rink effects and a lack of a uniform definition of what a “hit” is, precisely: scorekeepers tend to use the “I know it when I see it” definition, which isn’t very scientific. Until repeatability can be established, there’s a limit to what one can say about physicality using the hit statistic, though I would submit that for the purpose of this study, it is a sufficient proxy. Two, there are a large number of confounding factors that affect the outcome of any individual game, many of which cannot be consistently and repeatably quantified, despite the efforts of hundreds of video-game developers over the last twenty years. Without being able to account for everything else, it is all but impossible to establish causality. The most we can say with the data on hand is whether there’s a possible relationship or trend between recorded hits and earned standings points.

Results

Table 1 shows the pre-lockout data for the 30-team era, spanning 1999-2000 until 2003-04, where r is the Pearson correlation coefficient and p is the significance.

Season r p
2000-01 0.206 0.276
2001-02 0.284 0.128
2002-03 0.032 0.865
2003-04 0.138 0.468
2000-04 0.042 0.649

Table 2 shows the post-lockout data, from 2005-06 to 2008-09.

Season r p
2005-06 0.057 0.765
2006-07 -0.062 0.746
2007-08 0.112 0.557
2008-09 -0.063 0.741
2005-09 0.001 0.994

While performing the analysis, I noticed an interesting trend: as I went further back in time, the League average number of hits recorded steadily decreased, then hit rock bottom in 2002-03, before abruptly climbing back up again. The data for this appears in Table 3 below.

Season Hits Pts %
2000-01 1982 0.525
2001-02 1871 0.525
2002-03 602 0.532
2003-04 1294 0.530
2005-06 1310 0.557
2006-07 1467 0.557
2007-08 1559 0.555
2008-09 1719 0.557

While this would not affect any individual-season correlations, it could potentially alter the correlation of the cumulative data. Because of this, I reran those correlations using a percentage of league average hits, with the corrected values appearing in Table 4, below. (Note that because the pre- and post-lockout point-percentages were internally consistent, they were not adjusted.)

Season r p
2000-04 0.151 0.426
2005-09 0.004 0.961

Conclusions

It seems abundantly clear from this study that, rather than there being a negative relationship between hitting and winning, indeed there’s actually none at all, a result that contradicts the prior work by Willis and call into question the way conventional hockey wisdom is often cited. I choose that phrase because conventional hockey wisdom has never been “just hit someone,” despite what a certain Conn Smythe quote might suggest. I think most coaches and analysts would tell you that the key to the game is less about knocking guys about, and more about picking your spots and making the most of the opportunities created. Obviously, a bad hockey team that hits a lot is still a bad hockey team. Maybe they don’t convert hits to turnovers, maybe they don’t transition well, maybe they just can’t put the puck in the net; at the end of the game, they still lose more often than not. On the other hand, a good team that hits a lot does all of these things pretty well, and so on. (As an aside, it’s also worth reiterating how problematic the hit stat is, in light of the results uncovered in Table 3. Say what you will about the fluctuating incidence of physicality in hockey, I have a very difficult time believing that the entire NHL decided, en masse, to lay 1,000 fewer hits than average in 2002-03. Certainly, I would advise caution in interpreting any results obtained using hits as a factor.)

I think the important thing to pull out of all this is that changing to a more or less physical brand of hockey likely won’t dramatically affect the outcome of your season. If a team wants to truly get better, then they need to succeed more at the fundamentals of the game — skating, passing, and shooting — and become more adept at reading the play and knowing when to take which action. Being better at those things will naturally put you in better positions to lay hits, if that’s your game (and it needn’t necessarily be), and allows you to take better advantage of the opportunities said hits create. I would suggest both to those who believe that physicality should be a top priority, and those who feel it’s unnecessary, that the truth of the matter is that what you really want are good hockey players, and everything else will follow from that.

Be Sociable, Share!

Tags: , , ,

You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.

Comments...

One Response to “A Correlational Analysis of the Relationship Between Hitting and Standings Points”


  1. That’s good work Doogie, and much closer to what I would have expected to see.

    Thanks for posting it.

Leave a Reply