Sports Statistics for Shad@MAC 2013

As a long time sports fan as well as along time statistics fan, I have developed a keen interest in sport analytics. Originally coined "moneyball" in Major League Baseball (Wikipedia entry on the original Moneyball), managers in many of North America's profesional sports leagues have turned to similar advanced statistical analysis to try and get the most value out of players and create efficient teams. One sport which has, in my opinion, not received a sufficient treatment of advanced statistics is box lacrosse (hereafter: Lacrosse).



I have spent the last two National Lacrosse League (NLL) seasons tracking some basic team stats, made available through the league's website (www.nll.com) and creating new parameters of analysis.

Before we get into the body of analysis, an 8 team league creates a woefully small sample size, therefore the results as presented are not statistically significant; however I think they show enough promise that they could be researched further using information from past seasons to create a more appropriate number of data points.

First we look at Possessions per game vs Wins

I have defined a possession as winning a face-off, picking up a lose ball, or having the opposing team commit a turnover. This creates some robustness as there are many possession events in a given game. Here we get fairly scattered results, with a low R-squared value that is caused by two outliers.

When we remove those outliers, we see the following:

Here, the R-squared value shows a very strong relationship, which is also very intuitive. If you get the ball a lot, you will win a lot. It is my hypothesis (which I don't consider thoroughly tested) though that possessions are the most important metric determining wins.

To give some comparison, we look at the other type of event which it seems most obvious would lead to wins: goals. While there is some merit to just tracking goals scored, and it does give some interesting results, it does not account for goals against, nor does goals per game change the way the results are recorded. Winning 18-12 counts the same as winning 2-1. For purposes of giving an overview, I have chosen to look at each teams's mean goal differential. Negative should indicate that a team is being outscored often, while a positive should mean a team is outscoring their opponent often.

We see strong results right off the bat this time, having a strong correlation with no outliers. I would still suggest that possessions would be the more important metric as they come with a greater R-squared value and the elimination of outliers become less significant as samples grow larger.

The above is just a small sample of the work I have done and am doing on an ongoing basis. It is not meant to give any meaningful conclusions, but only to show the marriage of two things which I am passionate about and pique our collective curiosities with the possibility of further study being fruitful.

Thank you.