Tags

, , , , ,

Let’s take a step back from some of the Columbus-specific discussion and look into some of the perceived positives about team statistics.

Image from The Guardian // Ray Stubblebine/Reuters

Post-game hockey analysis can be a strange thing. We’re given information about the actual scoring (somewhat useful) and the shot totals (good in the long-term, not particularly essential from game-to-game). But oftentimes there’s a discussion of how physical a game was or the sacrifices that players make. That kind of grit is lauded and viewed as selfless or otherwise beneficial to the team effort.

Particularly post-playoffs, there’s still a ringing in my ear about the Rangers’ blocked shot totals, or the notion of toughness being an important element of wings. Dave Tippett commented “I’m not sure if [shot-blocking] is good for the game, but it’s good for winning.” USA Today praised their gritty brand of tough hockey. The NY Post even insinuated that lower block totals were the reason for a Rangers loss. Even the Caps were praised for their switch to a more playoff-friendly brand of hockey (both mentally and physically).

But are these traits really necessary? If your team is consistently getting high hit totals or high shot-block totals, is that in any way beneficial to victories? Let’s take a closer look at those two categories and see if we can find anything worthwhile.

First some rationale about the methods and the data involved. We’re going to use road hits only in any correlation study. Previous efforts have identified the hit total markup that occurs in some arenas, including this one from Jonathan Willis at NHL Numbers. Road hits remove some of that bias, but probably isn’t a perfect method. If we were entirely thorough, we’d look into a weighting system and figure out a normalized hits per game. As it is, the road numbers should suffice for general trends.

Next, we’re also going to use road shot blocks. To my knowledge there isn’t any noteworthy bias arena established in this category, but this should help randomize the effects of the official scorers in each location. Much like in hits, we could go into a full normalization study (and it’s entirely possible that Western and Eastern locations could be radically different) but we’ll go with this.

As per the data and its analysis: we’ve collected all the road hit and road blocked shot totals from all 30 teams for every year post-lockout. We’re going to compare them to some selected team-total results for the year using the Analysis of Variance (ANOVA) method in the statistical software package JMP. This gives a bit more insight than a typical r-squared test in Excel and we’ll use 95% significance as our benchmark in a fit-model platform. Now onto the results.

ROAD HITTING RESULTS We’ll take a look at information for ANOVA tests on points, goals against, goals for, and goal differential as a function of only road hits.

Points Earned: There is no correlation between total road hits and game points for a team. The regression plot can be seen below, but it’s not that useful.

Points and Road Hits are not related

This isn’t a shocking result, but it’s something to start with (the p-value is 0.9534, greater than our cutoff of 0.05 from the 95% significance). And since ANOVA doesn’t reveal a significant impact of road hits on points, the regression line is also insignificant.

Goals against per game: This is the first interesting result we run into. The relationship between goals against per game and road hits is significant (p-value is 0.0019, less than the cutoff 0.05, indicating significance). We also find that the regression line is significant as seen below:

Goals against per game and Road Hits have a significant negative correlation

The linear slope of the regression fit is -0.00527. That means that for an increase in road hits there is a corresponding slight decrease in goals against per game. This is seemingly a positive argument for a more physical game, at least in terms of defense. However, the next finding leads to another conclusion

Goals per game: There is a significant correlation between goals scored per game and road hits (p-value is 0.0002). The regression line is also significant as seen below:

Goals per game and Road Hits have a significant negative correlation

The linear slope is -0.000571, meaning there is a slight decrease in goals for per game as road hits increase. When combined with the previous result, we can conclude that heavy-hitting teams tend to have a lesser offense and a better defense. But as seen in ANOVA from points, that doesn’t tend to result in a better or worse record. In fact, one final observation makes hits even more irrelevant.

Goal differential per game: There is no significant relation between road hits and goal differential per game (p-value equals 0.8489). This isn’t really shocking, but I’ve included the regression output from JMP just for completion.

There is no relation between Road Hits and Goal Differential

Road Hitting Conclusions: High team hits is neither a positive nor a negative indication of their overall success. When we hear TV (or print) media drool over the supposed high-quality play from a very physical game and their conclusions come from hit totals, their assertions aren’t valid. Hitting doesn’t lose games, but it doesn’t win them either.

ROAD BLOCKED SHOTS RESULTS Let’s look at those same categories but for road blocked shot totals. These end up providing more interesting (and actually intuitive) conclusions, but more on that later.

Points: Points and road blocked shots have a significant relationship (p-value 0.0008) and the regression plot can be seen below.

Points and Road Blocked Shots have a significant negative correlation

The slope parameter of the linear fit is -0.04607, indicating that as blocked shots increase, the overall points gained decreases. This wouldn’t seem to make sense given all the focus on New York shot blocking success narratives, but we’ll get to the rationale a bit later on.

Goals for per game: This doesn’t have a significant correlation. Even though the plot below gives us the best linear fit for the data, the p-value for the model is 0.0852, so we don’t conclude that there is an relationship. Thus, the line is invalid by ANOVA, but the plot is included anyway.

Goals For and Road Blocked Shots are not related

Goals against per game: Here we do find a significant correlation between blocked shots and goals against through ANOVA (p-value 0.0022). The regression plot is below.

Goals Against and Road Blocked Shots are significantly positively correlated

The linear slope is 0.001047, indicating that more blocked shots typically accompanies more goals against.

Goal differential per game: This ANOVA analysis also results in a significant correlation (p-value 0.0006) and the last regression plot can be seen below.

Goal Differential and Road Blocked Shots are significantly and negatively correlated

The linear slope of this fit is -0.001595, indicating that more road shot blocks accompanies a worse goal differential.

Road Shot Blocking Conclusions: At first, these results don’t make sense, especially given narratives and ideology supporting a team-first mentality. But let’s consider the importance of shot differentials. As has been established in studies of the Corsi number, winning depends very heavily on shot differentials, scoring chance differentials, and net puck possession. An increase in overall shot blocking would suggest an increase in net shots coming against a team (either on net or otherwise). To have that kind of action means more scoring chances against, lowered puck possession, and, ultimately, losing. Of course, accurate confirmation of this would need to be accompanied by more study of team Corsi data which may be something worth exploring in the future.

It is critical to note, however, that these items (like most hockey conclusions) happen in the long run. Goals are inherently luck-driven events and happen far less often than shots, so shot-blocking might be successful in some degree of goal prevention a single playoff series. But over the course of a season, if your team is blocking large numbers of shots, that’s seems to be a sign of a squad that will have a lowered finish in the standings.

CONCLUSIONS: Hitting does not make a team good or bad, but does tend to occur with lowered goal totals (both for and against). Excessive shot blocking is statistically accompanied by more goals against and fewer points in the standings. So at least in these two categories, the amount of grit isn’t really a positive (and may actually be symptomatic of a bigger problem). This certainly doesn’t consider getting to “gritty” areas or having mental toughness, but I think it’s time to stop painting hits or shot blocking as a net positive.

Advertisements