Loading scores...
The Numbers Game

Advanced Stats: Correlations

by Ryan Knaus
Updated On: October 4, 2018, 4:04 pm ET

Today's column looks at a wide variety of stats and how they correlate to fantasy values. Is usage rate a useful metric for explaining or predicting fantasy production in 8-cat and 9-cat leagues? How important is scoring in Daily Fantasy Sports, and should points-league owners shy away from players who commit lots of turnovers?


In addition to stats like rebounds, assists, steals and blocks, I'll consider more 'advanced statistics' like Player Impact Estimate, Usage Rates, and Offensive Efficiency. If you're unfamiliar with these terms, it may be helpful to read my advanced-stat definitions from last week before diving into the rest of this column.


Since the focus is on how these stats correlate to fantasy values, let's first explain that term -- correlations "measure the extent of interdependence for variable quantities." Basically, a positive correlation suggests that as one variable rises (e.g. minutes) so does the other variable (fantasy value), and vice versa. A perfect positive correlation is 1.0, and a perfect negative correlation is -1.0.


We'll begin by looking at the NBA's official scoring system, which is the default for FanDuel, Yahoo and other sites. After that we'll look at 9-cat and 8-cat.


Here are the NBA Official scoring settings:


Points: 1

Rebounds: 1.2

Assists: 1.5

Steals: 3

Blocks: 3

Turnovers: -1 


You'll notice that this system doesn't penalize percentages or specifically reward 3-pointers -- a player who hits four 3-pointers gets 12.0 fantasy points, the same as someone who shoots 6-of-16 from inside the arc. Among qualifying players (at least 10 games played and 15.0 minutes per game), here are the correlations for basic stats:



As you can see, scoring as a category emerges as the clearest bellwether of DFS/NBA Official value. You can find players who produce lots of value without high-volume scoring -- Draymond Green, for instance -- but they are the exceptions, not the rule.


Turnovers are a fitting example of the famous warning that 'correlation does not imply causation'. When turnovers increase, we see a strong positive correlation (0.805) with DFS/NBA Official values. Since turnovers are themselves a negative scoring category, and in fact the only way to lose points, this appears counter-intuitive. Turnovers are not the cause of increased value, of course, but they tend to accompany high-volume counting stats. Look no further than the league-leaders in turnovers -- DeMarcus Cousins, Russell Westbrook, James Harden and LeBron James, all of whom rank in the top-five for NBA Official value (the elite exception is Giannis Antetokounmpo, who averages a modest 2.8 turnovers). Same as last season, my takeaway here is that you can essentially ignore turnovers in this scoring system (and many similar points-based leagues).


Editor’s Note: Nobody hit FanDuel’s Golden 5 jackpot this season, meaning the prize has grown to a massive $340,000. And this week, regardless of whether or not someone hits a perfect lineup, they’re giving away that $340K to whoever wins this week’s final contest. Play FREE for your shot at $340,000!


Although the other four stats all show 'moderate positive' correlations to overall value, rebounds and assists are still substantially higher than steals/blocks. The numbers are closer than they were last season, however, since 2016-17 had a slightly different scoring system -- instead of steals and blocks each being worth 3.0 fantasy points, they were worth 2.0 points. For comparison's sake, here's what that looked like:


2016-17 Correlations for DFS/FanDuel


Now that we've defined our terms and looked at the correlations for familiar statistics, let's branch out with more advanced metrics (and minutes!).



As we'll soon see, minutes are universally strong indicators of fantasy value. If a player's minutes rise, their fantasy value is likely to follow suit (and vice versa). This confirms the obvious, while underscoring how important it is to monitor roles and rotations throughout the season. The Rotoworld Season Pass has a terrific "Playing Time Report" that tracks trends throughout the season, and you can check out NBA team's rotations here and here.


Player Impact Estimate proves to have a strong positive correlation, as does usage rate. After that, things get murky. Offensive Rating is moderate/weak, but look at Defensive Rating -- the correlation is essentially non-existent. Fantasy hoops don't care for your defensive efficiency...sorry Aron Baynes, Andre Roberson and Luc Mbah a Moute.


So far, we've been talking exclusively about points-leagues, DFS and the NBA Official scoring. Let's look at the results when we switch to 9-cat and 8-cat values.


Here are the results for 9-cat:



And here they are for 8-cat:



Usage Rate proves to have a solid, positive correlation with fantasy values in all formats. It's substantially higher in 8-cat and DFS/NBA Official, however, simply because those don't penalize turnovers much, if at all. When we look at 9-cat leagues, the correlation for usage drops into the moderate range (0.566), as opposed to being strong (0.732 in DFS) or strong/moderate (0.656 in 8-cat).


A player's individual pace doesn't strike me as particularly meaningful for fantasy leagues, as I discussed last week, and this analysis drives home the point -- for 8-cat and 9-cat it has a very weak correlation (it's basically irrelevant). And if you're looking for an advanced-FG% metric to use in fantasy, True Shooting (TS%) looks like your best bet. It includes the impact of 3-pointers, similar to Effective Field Goal % (eFG%), but also captures the impact of free throws, and has a stronger correlation to overall values in each of the scoring systems discussed today.


At the end of the analysis, I was pleased to find that the numbers supported common-sense thinking (e.g. minutes and usage being of primary importance), highlighted the meaningful correlation of some stats (offensive rating, TS%) and meaninglessness of others (defensive rating, pace), and also showed the differences between various formats (usage has weaker correlation in 9-cat, net rating is weaker in PTS leagues, etc.). Viewed in context as tools to assess fantasy values and production, such metrics can be extremely useful. As usual, send me an email or a message on Twitter with any insights or questions! Good luck this week.

Ryan Knaus

Despite residing in Portland, Maine, Ryan Knaus remains a heartbroken Sonics fan who longs for the days of Shawn Kemp and Xavier McDaniel. He has written for NBC Sports Edge since 2007. You can follow him on Twitter.