Wins Above Replacement (WAR) is a baseball statistic that seeks to estimate a player’s value to his team by comparing him to a mediocre “replacement level” player that can be brought up from the minor leagues.

FanGraphs, an advanced baseball metrics website, defines WAR as:

An attempt by the sabermetric baseball community to summarize a player’s total contributions to their team in one statistic… WAR is not meant to be a perfectly precise indicator of a player’s contribution, but rather an estimate of their value to date. Given the imperfections of some of the available data … WAR works best as an approximation.

Baseball’s WAR gave me inspiration to come up with a way to measure the effect of randomness on New Orleans gun violence, a topic I introduced in the Fatal Shooting Percentage post.

Baseball sabermetrician Voros McCracken defined randomness for me in a December 2014 email, saying there are a “vast amount of factors that could affect the outcomes you’re studying and that while you can control for some of these, most are so hidden or hard to know, their effects on the outcome seem random. Random chance and ‘luck’ isn’t about things which have a random number generator affecting them, it’s about our own lack of knowledge on the factors determining the outcome.”

With that definition in mind, murder is a bad statistic for measuring gun violence because murder numbers are hugely impacted by randomness. There are three components making up the murder statistic: fatally shot homicide victims, victims of non-gun homicides and victims of homicides that are identified as non-murders. To calculate the number of murders in a given year, one adds fatal shooting victims with non-gun homicide victims and subtracts non-murder homicides.

Measuring variations in these components from year to year creates an estimation of how much randomness has positively or negatively impacted a year’s murder total.

Gun violence since 2011 shows relatively consistent patterns in terms of the number of fatally shot victims produced by an average shooting as well as the number of non-gun homicides and non-murder homicides on an average day. The available data since 2011 shows an average shooting between 2011 and July 2015 resulted in roughly 0.39 fatal victims. In addition, the average day between 2011 and July saw roughly 0.05 non-gun homicides and 0.02 non-murder homicides.

Putting a number on how many murders one would expect in a year based on the number of shootings, is relatively simple. It involves multiplying the number of shootings for that year by the average fatal victims per shooting from 2011 to 2015, adding the average non-gun homicides for a year, and subtracting the average non-murder homicides.

This methodology is intended to produce an estimate of the role of randomness, much in the same way that WAR estimates a ballplayer’s value. The methodology’s accuracy will improve as more data is collected over time.

The results for 2011 through July 2015 are shown in the table below. (“Average murders” is the number of murders one would have expected based on the number of shootings.)

Taking it a step further, it is possible to track changes in the effect of randomness for each day since 2011. This is provided in the graph below.

The reason: New Orleans was lucky in 2014, and it has been unlucky this year.