Defense Versus Position (Dvp)

Defense versus Position (DvP) data is incredibly popular among NBA DFS players. As said before, basketball is such a daily sport – as compared to NFL where you have to wait a week between slates, in the NBA it’s less than 24 hours for months. Because of this, matchups become really important. DFS sites like DraftKings and FanDuel cannot wildly swing player salaries on a night-to-night basis, which means that players aren’t priced for matchups in NBA as much as other sports, making it a large factor that DFS users need to focus on.

However, there is one big issue with current, raw DvP data: it isn’t adjusted for site salaries.

Let me explain. Let’s say the Charlotte Hornets allow a PG – say, – to score 30 DraftKings fantasy points. Let’s say that same night the Lakers allow a PG – say, – to score 28 DK fantasy points. By raw DvP measures, the Lakers are “better” at defending (or maybe better put, at allowing less fantasy points) point guards than the Hornets.

You can see the obvious problem here – allowing Curry to score only 30 DK points is much more impressive than allowing to score 28, even if it’s a higher number.

Fortunately, we can account for this by adjusting DvP for player salary. If use a formula to calculate what a player “should” score based on his current salary (you can do this by using historic data or a formula like 5x value if you want to be more rudimentary), and then see how much he actually scores compared to that number. Let’s continue our example:

Curry’s expected fantasy points: 45; Curry’s actually fantasy points: 30; Hornet’s “True DvP” in terms of Plus/Minus allowed: -15

Sloans’s expected fantasy points: 23; Sloan’s actually fantasy points: 28; Laker’s “True DvP” in terms of Plus/Minus allowed: +5

This is a much better way to judge how a team defends a position, as it accounts for a player’s salary, which is really just meaning that we’re accounting for opponent and how good their competition is.

The great thing about doing it this way is that, like “True Usage,” no one is really utilizing the data this way. Over at FantasyLabs, the “True DvP” data is incorporated into the Player Models, so you can get an accurate representation of how easy or hard a certain matchup is for that night. Especially early in the season, this is very important.

In general, this course is about taking common NBA DFS data and telling 1) why it can be misleading, and 2) a better way to use it, or more aptly put, how to make it more predictive. And that is the key – if your current DvP data isn’t accurately predicting matchups, it isn’t useful to you, even if the data isn’t “wrong.” Data can be right and factual and still incorrect. Raw DvP is a great example of this. True DvP is a great example of data that is both right and predictive – that’s the middle ground you want for DFS.