A Quick Introduction to Projected +-
"Giannis is projected for the most points tonight, I'd be crazy not to pick him." - A direct quote from a friend of mine a few weeks ago while we were picking our DFS lineups together. Clearly, he didn't understand the importance of finding value in players, and had tunnel visioned on point output. What he was missing was that Giannis was also the most expensive player on the slate, and by a wide margin. As such, there were significantly better value plays across the board.
Now, there is always a scenario where you have a gut feeling and just have to pick that top guy, and I respect and understand that, but at Analytic City we aim to fill our lineup every night entirely with guys who we predict to have a positive projected +-. Why? Because if our projections are accurate, these players will all return incredibly high value, and thus a higher point total.
Here you can see players with positive projected +- values in the final column
Projected +- is the difference between a player's projected points, and the points you would expect a player at that salary level to score. You can define expected points either through an arbitrary threshold, (such as a 250 point total on Draftkings) or by looking at historical data from this season, and seeing how players with identical salaries have scored.
Let’s look at an example to make this more clear:
From the above table you can see that Jimmy Butler costs $8,400, and Joel Embiid costs $10,000. We want to know, purely from their salary, how many points we should be expecting players who cost that much to produce. Obviously, on average, we would expect a player who costs $10,000 to produce more points than a player who costs $8400. So, we can find Butler and Embiid’s salary expected points in 1 of two ways. Using Butler for this example:
- We can look at all players who have been priced at $8400 during the season so far, and average how many points they have scored.
- We can establish an arbitrary threshold with a desired point total for our entire lineup. For now let’s call it 250 points.
Option 1 seems more intuitive, but we run into some issues. When we find averages for salary levels that not many players have been priced at (say $10,500), outliers can skew those averages above the averages of higher salaries, or below the averages of lower salaries. So, using our example again, we could run into the case where we find that players priced at $8300 have actually scored more on average then players priced at $8400 this season. Does this mean we want to say that players at $8300 are worse value plays because we expect them to score more points? Of course not!
We fix this by using option 2. By establishing an arbitrary point total we wish to reach (let’s say 250), we can standardize salary expected points so that as a player's salary decreases, we expect them to produce fewer points. When we now compare those values to projected points, we will have an accurate measure for projected +-
Along with projected $/point (how many $ you are spending to acquire a single point), projected +- provides one of the best metrics to look at whether a player is a potentially high value play on any given night!