I create and explore a metric called Scrap Meter. Who are the best and worst, and how does it relate to Defensive Win Shares?

As fans, we often classify players into self-made categories. Sometimes it’s lazy stereotyping, but other times it can truly define an NBA player’s role. One of the most common categories is the scrappy defender, the pest, the blue-collar worker, or the borderline dirty.

There’s an infinite number of names for this player, but you know who I’m talking about. In most cases, it doesn’t take more than visuals to identify the scrappy player. Nevertheless, wouldn’t it be cool if there were some way to quantify scrappiness?

### Introducing The Scrap Meter

Luckily, Stats.NBA.com provides player tracking stats that include hustle stats and run speed stats. Using these, I’ve created a new metric called Scrap Meter. The purpose is to measure a player’s defensive scrappiness. The formula for Scrap Meter is:

Scrap Meter = (Deflections * 0.49) + Steals + Loose Balls Recovered + Charges Drawn + Average Run Speed on Defense

The formula is really simple; it’s just adding the stats together. However, there are a few nuances to it (such as the 0.49) that I will explain in the following section. It may get technical, so feel free to skip to the results.

### The Data

I collected steal data, hustle data, and speed data for all 454 NBA players in the 2016-17 season. These data are in season totals, not per game or per minute. I then filtered for players with greater than 12 minutes played per game and greater than or equal to 15 games played, leaving 328 NBA players.

The variables I chose to use, of course, were Deflections, Steals, Loose Balls Recovered, Charges Drawn, and Average Run Speed. If I simply added these stats together for each player, the high magnitude stats, such as Deflections and Steals, would dominate the low magnitude stats, such as Charges Drawn and Average Speed. So I standardized or normalized each stat. In other words, for each data point, I subtracted the mean of that variable and divided by the standard deviation of that variable. This allows for each variable to be on the same scale, centered around zero.

Additionally, I felt that Deflections needed to be weighted down, because it is not as important as the other stats. Steals, Loose Balls Recovered, and Charges Drawn result in a change of possession. Therefore, they are arguably more important than a deflection, which does not guarantee a turnover. To find a proper weight, I ran a simple linear regression of Steals on Deflections. This resulted in a statistically significant coefficient of 0.49. Meaning, an increase in one Deflection yields an average increase in Steals of 0.49. I believe this is a sufficient weight, as one Deflection is essentially worth 0.49 Steals. To truly know the value of a Deflection would require much more dense and unavailable data. I did not conduct a similar procedure for Average Speed Defense, because its value, in terms of the other variables, is impossible to know.

After standardizing the variables and multiplying 0.49 by the standardized Deflections, the stats were added up for each player to find the Scrap Meter.

### Results

Below are the top 20 players in Scrap Meter for the 2016-17 season.

A lot of the names in the table are to be expected. However, Stephen Curry is ranked third in Scrap Meter. Curry is normally considered an average defender at best, particularly because he’s labeled as soft and un-athletic. However, he’s fifth in the league in Steals, seventh in Deflections, and first in Loose Balls Recovered.

Another surprise is the big-men on the list - DeMarcus Cousins and Greg Monroe. Monroe in particular has been considered a poor defender throughout his career. This year, he’s excelled in his role off the bench and has focused more on defense. It could be argued that rebounds are an element of scrappiness. I would agree, but I thought it would oversaturate the top of the leaderboard with big-men. DeMarcus Cousins easily makes up for it with his Deflections and Charges Drawn.

In case you were curious, here are the bottom ten players in Scrap Meter for the 2016-17 season.

I think it’s fair to say that these players are not engaged on the defensive end. They’re not moving terribly slow, but they just don’t have a knack for finding the ball and forcing turnovers. I honestly expected J.J. Barea to be higher, but again, that could be lazy stereotyping.

To go further with Scrap Meter, I introduced more context by dividing it by total minutes played to form Scrap per minute. Because raw Scrap Meter is based on season totals, it may be biased in favor of those who’ve played more minutes. Scrap per minute combats this to show more Scrap efficiency. Below are the top 20 players in Scrap per minute for the 2016-17 season.

The results are nearly the same, except for a few additions. Most notably DeAndre Liggins at number two, who only recently found time in the Cavaliers’ rotation with the injury to J.R. Smith. The additions of Tony Allen and TJ McConnell are expected, as they are notorious for being pesky defenders. Greg Monroe’s jump to rank three provides further evidence of his new defense-first mentality.

### Comparing Scrap Meter to Defensive Win Shares

Finally, lets compare Scrap Meter to an all-encompassing defensive metric, like Defensive Win Shares. Ideally, the two statistics will be similar, but not identical. To test their relationship, I conducted another simple linear regression of Defensive Win Shares on Scrap Meter. The results show that 15% of the variation in Defensive Win Shares is explained by Scrap Meter. This number is not particularly high, but not incredibly low either. Additionally, Scrap Meter is a statistically significant variable for predicting Defensive Win Shares. To get a visual, the graph of DEF WS vs. Scrap Meter is below, with the red line as the line of best fit. While the points are quite scattered, there’s a discernible, positive relationship between Scrap Meter and DEF WS.

### Conclusion

Scrap Meter obviously isn’t perfect. At the least, Scrap Meter is a random sum of arbitrary numbers. At its best, it’s a cool, and hopefully unique, way to look at defense. Even more, it provides a glimpse into what statistics can do to balance our traditional ways of thinking that stem from visual judgements. Honestly, I came to this idea by browsing the pages of Stats.NBA.com to see what I hadn’t explored before. Maybe this proves that stats are either incredible easy and powerful, or they’re a load of crap. I can’t decide.

*Stats taken from Stats.NBA.com*

Edited by Brian Kang, Peyten Maki.

- John Wall
- Kawhi Leonard
- Chris Paul
- Stephen Curry

## What do you think?

Please log in or register to comment!