clock menu more-arrow no yes

Filed under:

Do the Bruins use zone finish data to analyze players?

New, comments

NHL teams have all sorts of proprietary methods for statistically analyzing players. Are zone finish stats ones that the Bruins use?

Mark L. Baer-USA TODAY Sports

All numbers from behind www.behindthenet.ca

While most analysis of players focuses on how they help their team when they're on the ice, that's not where a player's story ends. Taking a penalty, for example, can lead to a goal occurring while you're not on the ice. Players like Dominic Moore, Paul Gaustad, and Marcus Kruger absorbing tough minutes helps their team's offensive stars shine with easier deployments. And, finishing your shift in the offensive zone makes whoever gets the next shift's life easier. When it comes to that last one, Bruins players have been surprisingly good. So good, in fact, that it kind of raises questions.

To start this off, I first graphed the % of offensive zone finishes against the % of offensive zone starts for all NHL players last season who played over 30 games.

The r^2 for the relationship was 54%. That's pretty good, and it suggests that there's a definite relationship between your percentage of ozone finishes and your percentage of ozone starts. The error on the coefficients for the linear relationship is also pretty low, meaning we've got a pretty reliable way of predicting the percentage of offensive zone finishes for a player. I'll call this expected percentage xfinishes for the rest of the article. The delta between the observed and expected zone starts is where the real fun begins. For the rest of this article, I'll refer to the difference between the expected and observed zone finishes as dfinishes.

The first thing we need to establish with this new stat, dfinishes, is that it's normally distributed enough for us to be able to use the normal model to draw useful conclusions. The difference between the median and mean is very small, and the graph of the distribution is roughly unimodal and roughly symmetrical. Here is said graph

If anyone's interested, here's a google doc containing all the numbers I played with.

Now, for the Bruins related part. I looked at all of the Bruins last year, plus all the players that were acquired during the offseason, and noticed something. Let's see if you can notice it too.

NAME Off Zone Start % Off Zone Finish % xfinish delta z score percentile
JIMMYHAYES 56.4 56.3 51.768164 4.531836 2.2291372356 0.9870973258
ZACRINALDO 49 54 49.70349 4.29651 2.1133841613 0.9827157631
ZDENOCHARA 45.5 52.3 48.726955 3.573045 1.7575233645 0.9605853919
BRETTCONNOLLY 53.4 54.3 50.931134 3.368866 1.6570909985 0.951249174
BRADMARCHAND 50.1 53.2 50.010401 3.189599 1.5689124447 0.9416655425
PATRICEBERGERON 42.9 51.1 48.001529 3.098471 1.5240880472 0.9362563543
MAXIMETALBOT 46.3 51.8 48.950163 2.849837 1.4017889818 0.9195105785
GREGORYCAMPBELL 39.7 49.9 47.108697 2.791303 1.3729970487 0.9151230813
DAVIDKREJCI 52.2 53.3 50.596322 2.703678 1.3298957206 0.9082233975
DOUGIEHAMILTON 47.5 51.7 49.284975 2.415025 1.1879119528 0.8825656633
MATTBELESKY 52.6 52.9 50.707926 2.192074 1.078245942 0.8595377061
DANIELPAILLE 47.9 51.1 49.396579 1.703421 0.837885391 0.7989521762
KEVANMILLER 51.7 51.9 50.456817 1.443183 0.7098785047 0.7611099728
LOUIERIKSSON 49.1 51 49.731391 1.268609 0.624008362 0.7336886666
ADAMMCQUAID 49.2 51 49.759292 1.240708 0.6102843089 0.7291629687
DENNISSEIDENBERG 43.8 49.3 48.252638 1.047362 0.5151805214 0.6967862738
MILANLUCIC 54.2 52.2 51.154342 1.045658 0.5143423512 0.696493385
MATTHEWBARTKOWSKI 49.1 50.7 49.731391 0.968609 0.4764431874 0.6831203768
DAVIDPASTRNAK 69.4 56.3 55.395294 0.904706 0.4450103296 0.6718435687
CARLSODERBERG 52.1 50.9 50.568421 0.331579 0.1630983768 0.5647792244
RILEYSMITH 52.4 50.5 50.652124 -0.152124 -0.0748273487 0.4701757542
CHRISKELLY 46 48.5 48.86646 -0.36646 -0.1802557796 0.4284755982
MATTHEWIRWIN 53 50.4 50.81953 -0.41953 -0.206360059 0.4182545472
TOREYKRUG 60.3 52.4 52.856303 -0.456303 -0.2244481062 0.4112040343
SETHGRIFFITH 52 46.1 50.54052 -4.44052 -2.184220364 0.014472743

Did you also notice that the Bruins were really, really good in terms of dfinishes last year? I did too. Even the best player in the league last season in terms of dfinishes (Cunningham) was a former Bruin (R.I.P. in peace). Excluding the outlier Seth Griffith, none of the Bruins were in the lower 40th percentile of the league in dfinishes. Only 4 players were negative in terms of dfinishes. The average dfinish of these players was 1.81 without Griffith, and 1.56 with. On average, these players were better than over 75% of the league.

How does this stack up to the rest of the teams in the league? Well, you could do what I did for the Bruins and calculate the average dfinish of players for all 30 teams in the league, then use the standard deviation of that to find your pval. But I'm lazy and know math, so I'm not going to do that. Instead, for my analysis, I'm going to treat each team as a sample, and the average dfinish of each player on the team as a parameter of a sampling distribution. The population of the sampling distribution will be all players in the NHL that played over 30 games. Thanks to the central limit theorem, I know the sampling distribution of the average dfinish for players of all teams will be normal, and therefore all I need to find the pval for the Bruins is the standard deviation of the sampling distribution.

Because we know the standard deviation of the population (2.03), the size of the population (630), and we can estimate the size of each sample as 630/30 or 21, then we can calculate what the standard deviation of the sampling distribution will be. We get .430. That means that the Bruins with Seth Griffith have a z score of 3.64, and without, a z score of 4.23. So... yeah. If you're not familiar with stats, that's pretty darn good. In plain English, this means that there's a 0.0134% chance that the Bruins with Seth Griffith were this good due to random chance. Without including Seth Griffith, that turns into a 0.000000121% chance. It would seem, in all likelihood, that the dominance in this particular stat by these Bruins isn't an accident.

This could provide a reason for some of this offseason's moves. Offseason acquisitions Jimmy Hayes and Zac Rinaldo are both in the top 2% of the league. Let me put that into context. This means that not only have we found something that Zac Rinaldo is good at, it means WE HAVE FOUND SOMETHING ZAC RINALDO IS ELITE AT. Former Bruins fan's punching bag Gregory Campbell along with future Bruins fan's punching bag Max Talbot are also both, apparently, good at ending their shifts in the right zone. A lot of the players the Bruins either let go or traded this offseason, like Lucic, Soderberg, Smith, and Bartowski were all not as good as other players on the team.

Like every NHL team, the Bruins have entirely proprietary means for statistically analyzing players. They've got a whole lot more numbers than us and a whole lot more ways to break em down. So, here's the question. Did we just stumble upon one of their methods? You guys can discuss in the comments below.