Understanding NBA Advanced Stats: The Core Drivers of DFS Expectation
Brief Introduction
If you have ever read one of my NBA Core Plays articles, you have almost certainly seen me reference a player’s fantasy-point-per-minute (FPPM) rate. To try and prove why/how that might go up or down, I probably then talked about that player’s usage rate, assist rate, rebound rate, etc.
But what exactly are those rates? Do they really matter? And if they do, how much do they matter?
Before we answer that, let’s do an overview of the most basic part about NBA DFS: the scoring system(s). Once we get past that, it will become much clearer why you see the above phrases thrown around all of the time.
NBA DFS Scoring
What stats are even used for us and the players we roster to accrue fantasy points? The easy way to think of it is, if it was in the box scores that some of us looked at in the morning newspaper as we grew up, it probably is counted as a fantasy stat for our purposes (luckily for Jaren Jackson, personal fouls are not counted).
For your reference, here is the NBA scoring for FanDuel, and here is the NBA scoring for DraftKings. As you might expect, there aren’t many things listed. We basically have points, rebounds, assists, blocks, steals, and turnovers. While you can argue it falls out of the scope of this article, there are a few differences between the two sites that I want to make note of just in case.
The first thing is turnovers. On FanDuel, a whole fantasy point is deducted for a turnover, while it is only a half of a point on DraftKings.
Logic tells us that the more often you have the ball in your hands, the more likely you are to turn it over. That is why we see guys like Trae Young (4.8), James Harden (4.5), Russell Westbrook (4.5), Luka Doncic (4.2), and LeBron James (4.0) as the top five in turnovers per game at the hiatus point of the 2019-20 NBA season (note: this article is being written during the COVID-19 hiatus, so any remaining reference to 2019-20 “season stats” will be up to and including that last night of action before the hiatus). So these types of players will get docked more severely for turnovers on FanDuel compared to DraftKings.
One thing that DraftKings has that FanDuel does not is a bonus category. On DraftKings, a player can get 1.5 fantasy points if they record a double-double (sorry you five guys from the paragraph above, turnovers do not count). If they happen to get a triple-double, they get 3.0 fantasy points. And before you go and double-check, no, you do not get credit for both bonuses if your player gets a triple-double since they technically also got a double-double too.
The important thing to take away from the brief bonus discussion (you also get a 0.5 fantasy point bonus on DraftKings for a three-pointer made, by the way) is thinking about the types of players that are most likely to get those bonuses. Guess which two guys have the most triple-doubles in the 2019-20 season? Luka Doncic with 14 and LeBron James with 13. Did you notice that both of those guys were also on that list of averaging the most turnovers per game? This is why you will always see/hear me say that DraftKings has a scoring system that is far more advantageous for the superstars. Not only do they get docked less for turnovers, but that are also more likely to get the double-double or triple-double bonus.
The last most noteworthy difference between the sites that I want to point out are blocks and steals. On FanDuel, you get three points for each, while you only get two points for each on DraftKings. One of my favorite examples to use for illustrating this scoring difference (along with the others already mentioned) are the two studs for the Los Angeles Lakers, LeBron James and Anthony Davis.
As you can see here, LeBron James has gotten 1.50 FanDuel points per minute this season and 1.55 DraftKings points per minute. Anthony Davis has gotten 1.50 FanDuel points per minute and 1.45 DraftKings points per minute. So they produce at essentially the same level on FanDuel, but LeBron is drastically better on DraftKings.
Why is that?
The different scoring systems! Blocks and steals are weighted more heavily on FanDuel, and AD (4.2 blocks/steals per 36 minutes) easily outpaces LeBron (1.8) in those two categories. LeBron also averages two more turnovers per 36 minutes, and we already know those are docked more on FanDuel than DraftKings. Yet on DraftKings, that triple-double bonus is far more attainable for LeBron (13 times this season, as we already noted), and he is docked less for turnovers on that site as well.
So there you have it! Now that we know exactly which stats are being tracked, it is time to move on to how exactly we predict those stats will be accumulated.
Definitions
I highly recommend giving this page on Basketball Reference a skim. It defines a ton of words/phrases/stats that are used in basketball lingo. There are going to be some that you have almost certainly heard of, but you might be surprised at how many seem like a foreign language. For our purposes, we are interested in the ones that are predictive of all of those stats we just talked about that will accrue fantasy points.
Usage Rate
This is referred to as ‘Usage Percentage’ in the glossary I just linked above, but usage rate is the far more common term (same for assist percentage —> assist rate, etc.). I am not going to copy the exact mathematical formula for usage rate from that glossary (although you should absolutely study it to gain a better understanding), but I will copy the definition that is listed after that formula:
“Usage [rate] is an estimate of the percentage of team plays used by a player while he was on the floor.”
If we combine that sentence with the stats used from the mathematical formula, here is the best definition, in my opinion, for usage rate in layman’s terms:
Usage rate is an estimate of the percentage of times a player ends his team’s possessions with a field goal attempt, free throw attempt, or turnover while he was on the floor.
We know we get fantasy points when a player scores actual game points. Thus, usage rate is telling us exactly how often a player is attempting to score actual game points when they are on the floor.
For fun, here is the list of usage rate leaders for the 2019-20 season. It is no coincidence that the guy with the highest usage rate on that list (Giannis Antetokounmpo) also has the highest FPPM rate (1.85 FD / 1.94 DK) in the entire league. And remember all of those nights that you rostered P.J. Tucker and cursed at him endlessly for standing in the corner and doing nothing? Yeah, well, he is last on this list amongst qualified players with an 8.6% usage rate!
So when you factor in how often a player tries to score (usage rate) with how efficient they are at scoring (I prefer using true shooting percentage since that takes into account field goals, three-point field goals, and free throws, but there are plenty of other metrics that calculate a player’s shooting), that is going to be our most direct route to figuring out how a player gets actual points (which translate directly into fantasy points).
Assist Rate
The definition for assist rate is much more clear-cut and is probably exactly what you think it is too. From the Basketball Reference glossary, assist [rate] is defined as “an estimate of the percentage of teammate field goals a player assisted while he was on the floor.” Easy enough!
Not so fast my friends. -Lee Corso
This metric is actually depends on other players. If you are James Harden, you can do everything right in setting up your teammate for a good look, but that player could just miss the shot. This could obviously happen on a multitude of occasions throughout a game, causing his assist rate (and ultimately, his assists) to be lower than they should have been.
Conversely, a guy like Mitchell Robinson (i.e. the guy with the lowest assist rate in the league amongst qualified players) could luckbox his way into four assists one night (like he did against the Spurs on November 23rd) if everything falls just right, even though he didn’t record more than two assists in any other game all season. It goes both ways.
So let me introduce you to a friend I like to visit with every day when I am looking for abnormalities in players’ assist rates. And his name is Potential Assists. NBA Stats defines a potential assist as “any pass to a teammate who shoots within 1 dribble of receiving the ball”. So this metric is a way to layer in some more information to mitigate the variance associated with whether or not a teammate was actually able to hit the shot.
If you look at that list of potential assists per game leaders (from the paragraph above), and this list of assist rate leaders, you see a lot of the same names. LeBron James is first in both. Trae Young is second in both. Luka Doncic is third in assist rate and fourth in potential assists per game. I think you get the picture.
But how do we use this to affect our thinking when it comes to NBA DFS? Let’s use Bogdan Bogdanovic as an example.
In the 2017-18 season, Bogdanovic had an 18.3% assist rate. It went up to 19.3% in his second season (2018-19). And if you look at the start of this season, it was trending up even more. In those first 18 games (10/23 through 12/2), it was all the way up at 29.3%!
But then, something happened. The assist rate absolutely plummeted from December 4th up until the hiatus. In those 35 games, the assist rate was only 12.9%.
I wish I could tell you exactly what happened and why, but I can only speculate. It could have been any number of things, ranging from an injury (Bogdanovic started dealing with a wrist injury at a certain point in the season), to the coach’s strategy/gameplan, to Bogdanovic entering the starting lineup as of January 24th (and thus seeing less playmaking opportunities with the second unit). I don’t know why it happened, but I do know that it happened.
So how do we use potential assists to help explain things? Let’s do some time traveling.
I noted that 18-game stretch from 10/23 to 12/2 in which Bogdanovic had a 29.3% assist rate. He then missed the next game (12/4) with a knee/hamstring injury. Guess what his assist rate was in a little over 21 minutes in their next game on 12/6? 0 percent. 0%. Zero point zero percent.
Of course, anything can happen in one game, just like I said up above with James Harden and Mitchell Robinson. But if it is me on the morning of 12/8 (their next game), I am trying to figure out why he didn’t record an assist. If you don’t have access to the recording of the previous game (on NBA League Pass, for example), these stats are your next best resource (and maybe even the best resource depending on how comfortable you are deciphering game film). So what exactly would I be looking at?
The first thing I’d do is try to get a feel for how many potential assists he had averaged per game leading up to that point. You can find pretty much anything on NBA Stats, and this link tells us he averaged 7.3 potential assists per game from 10/23 to 12/2.
So the next obvious step would be to look at what the heck happened in that game on 12/6. Of course, we can do that too on NBA Stats, and we see that he had exactly one potential assist that game! At this point, my antennas are up, but I am not treating it as gospel.
With each game that passes though, you start to get a little more worried about the trend.
12/8: 13.5% assist rate
12/9: 4.6% assist rate
12/11: 0.0% assist rate
12/13: 5.5% assist rate
The potential assists in that four-game span show a drop to 5.8 per game. While that may not be enough of a drop to warrant those extremely low assist rates, it still shows that he is trending down. And if you bring in the entire sample from 12/6 through 3/8, you see that his potential assists per game in that span (5.1) were well below the 7.3 per game from the early-season sample.
Another stat we can use in conjunction with potential assists are touches. And yes, NBA Stats track that too!
NBA Stats defines touches as “the number of times a player touches and possesses the ball during the game”. You catch the ball from a teammate? That is one touch. You pass it to someone else and get it thrown right back to you? That is two touches. We are basically looking at how many times you have that orange round thing in your hands. Easy peasy.
So let’s use those exact same timeframes (10/23 to 12/2; 12/6 to 3/8) to see what Bogdan’s touches per game looked like.
10/23 to 12/2: 49.7 touches per game
12/6 to 3/8: 45.9 touches per game
The fewer times you have the ball in your hands, the fewer chances you have for generating assists (this metric can be used for points too, obviously). As you can see, that is exactly what happened with Bogdan at a certain point in the 2019-20 season.
Rebound Rate
These next few are pretty straightforward. I will still highlight rebound rate in its own section because there are still variables that can cause it to go up and down. On the Basketball Reference glossary linked above, total rebound percentage (i.e. rebound rate) is defined as “an estimate of the percentage of available rebounds a player grabbed while he was on the floor”. I bet you could have guessed that word for word if I asked you to! Easy enough.
The one noteworthy thing I will point out is that who a player is/isn’t on the court with can affect this. For example, Steven Adams had a rebound rate of 14.7% in the 2018-19 season. You know, the most recent year he was on the Thunder with Russell Westbrook? Now, I will try to put this next sentence as delicately as possible. While it has never been officially acknowledged by anyone, it has long been understood that it would behoove Adams to let Westbrook get as many of the rebounds as possible. What was Adams’ rebound rate this season, you ask? 19.3%. And the rebound rates for Westbrook the past two seasons? 15.8% in his last season with the Thunder and 11.7% this season with Houston.
While these situations are much easier for projection systems to keep tabs on, that is not the case for the assist rate example I laid out up above. Projection systems usually need a larger sample of games before they start heavily moving a player’s assist rate one way or another. A rebound rate is much easier to pin down.
Block/Steal/Turnover Rates
These three are probably exactly what you think they are. The definitions for each per the Basketball Reference glossary are…
Block Percentage (i.e. block rate): An estimate of the percentage of opponent two-point field goal attempts blocked by the player while he was on the floor.
Steal Percentage (i.e. steal rate): An estimate of the percentage of opponent possessions that end with a steal by the player while he was on the floor.
Turnover Percentage (i.e. turnover rate): An estimate of turnovers per 100 plays.
There are a few curveballs here. The first is that block rate only factors in two-point field goal attempts instead of all field goal attempts (like, you know, three-pointers). I am not joking when I say that I literally just learned that was the case when typing out these definitions.
The other curveball is that turnover rate is “per 100 plays” as opposed to “while a player was on the floor”. I have no idea why that is different than all of the others, but it ultimately doesn’t matter since we are using the same turnover calculation rate for all players.
Lastly, there are a few noteworthy things I want to touch on for the rates stats in this section. The first is that block/steal rates come with way more variance than the other three rates (usage, assist, rebound) we have already discussed, and that is mainly because they happen in games far less frequently compared to the other three rates stats. Let’s take Hassan Whiteside for instance.
Whiteside leads the NBA this season with an 8.3% block rate. That has equated to 3.1 blocks per game (3.5 blocks per 36 minutes). Compare that to the leader in points per game being 34.4, assists per game being 10.6, and rebounds per game being 15.2. It’s just not in the same ballpark (arena?). We can say the same thing for the league leader in steal rate, which is Kris Dunn at 3.8%. However, that has only equated to 2.0 steals per game.
Yes, block rate and steal rate matter, and they are always factored into projections. However, they just fluctuate wildly from game to game.
The other noteworthy item is one I have already briefly alluded to, and that is turnover rate tends to increase as usage increases (unless you are Luka Doncic). So when you are using CourtIQ to see how much the usage rate of Player X goes up when Player Y is off of the court, I encourage you to also toggle on the ‘TOV%’ stat as well. That will ultimately tell a little bit of the story too when it comes to how the FPPM increases/decreases.
Conclusion
It shouldn’t come as much surprise when you really think about it. The stats that accrue fantasy points are points (usage rate), assists (assist rate), rebounds (rebound rate), blocks (block rate), and steals (steal rate), with turnovers (turnover rate) deducting fantasy points. Doesn’t it make sense to try and figure out how often a player gets those stats when they are out on the court? Those rates are the bulk of the inputs that go into a player’s fantasy-point-per-minute rate that you so often hear/see me reference in the Core Plays article or on the Crunch Time show.
And do you know what the best part is? We have CourtIQ, the best tool in the business, that allows you to see how all of those rates fluctuate depending on who is on or off of the court.