I’ve posted a few recent articles on NBA strategy, including the importance of minutes and typical aging for NBA guards, because I’ve recently jumped into the sport. I’m a total novice without a clue right now, so I’m doing a lot of research on which overarching strategies might be best, developing certain heuristics I can use to hopefully not get crushed as I learn what the hell I’m doing. I’m basically just trying to take you on that journey with me to show you what sort of stuff I’m doing to try to reach a respectable level of play.

As I wrote in my initial NBA article, “I just started playing daily fantasy basketball this year and I’ve never actually written a single NBA article in my life—until now—so if we were betting on the usefulness of this post, I’d take the under on Barely Useful.”

Anyway, one simple way to see if I’m on the right track is to download my DraftKings game history and see what sort of scores I’m posting relative to the crowd. I did this a few weeks ago when I broke down my entire NFL game history. I’m a huge proponent of spending as much time looking at past results as assessing the future because if we don’t evaluate and critique our approach, it’s impossible to evolve and improve as a daily fantasy player.

I actually don’t even care if I’m a crappy fantasy basketball player right now because I know I can use a scientific approach to the game in order to improve. If my results are even average, I think I can develop into a profitable player down the line.


Downloading and Assessing Game History


First of all, you can find your game history by clicking on “My Contests” and then clicking the tab “History.” On the right side of the page is a link called “Download Entry History,” which will give you a CSV file with information on every league in which you’ve ever played.

Within that file, you can easily sort by sport, total points, and so on. The way that I determine how I’m performing relative to the rest of DraftKings users is to divide my finish by the total entries in a league. That will let me know in which percentile I’m finishing.

For example, if I finish 100th in a league with 400 people, I’d divide 100 by 400 to get 0.25. That means that I finished in the top 25 percent.

Note that this strategy doesn’t take overlay into account (since you’re assessing actual entries and not the total number of available spots in a league), so it’s a really accurate reflection of how you’re performing relative to others.


Early NBA Results


I’ve played in 634 NBA contests thus far in 2014. Here’s a look at my distribution of total points scored.

NBA Scores 1


This doesn’t really tell us a whole lot on its own because we have no idea what type of score is needed to win each league type, on average. I added a vertical black line, which I’ll address in one second.

Now, here’s a look at how I’m finishing in terms of percentile.

NBA Scores 2


Note that I inverted the results, so if I finished in the top 25 percent in a league, I charted that as the 75th percentile.

Here, the black line represents my median finish. That’s means my median score in 2014 is in the 63rd percentile, which is a good sign, especially since I’m just testing things out without much knowledge of, you know, anything at all.

Looking at my score distribution, I can determine the average median score for DraftKings users. That’s what I represented with the black line in the first graph. On average, a score that finished right in the middle-of-the-pack is about 234.75 points (based on my game history, anyway). In the average night of NBA, you’ll need about 235 points or so to win a 50/50. I’ve scored that many points in 63.6 percent of leagues, and my personal median score is 242 points.

Meanwhile, the average score in the 80th percentile (which is around what you’d need to cash in a tournament) has been 256.5 points. My personal 80th percentile score is 265.5 points, so I’ve performed much better than average thus far. That’s a surprise.

Another way to look at this data that might be more aesthetically pleasing is to sort it into buckets. Below, I charted the distribution of my scores into various percentiles and compared it to what we’d expect from randomness alone (which would just be 25 percent of scores in each quantile).

NBA Scores 3


Good news: I’m finishing in the bottom 25 percent much more infrequently than if things were totally random (only 12.3 percent of the time), and I’m finishing in the top 25 percent much more often (35.8 percent of leagues). Those are pretty awesome early numbers, so I think I’m on the right track, but I also think I’ve gotten a little lucky thus far.

One reason I believe I’ve been lucky is that the difference between the two middle quantiles isn’t that great. I could be wrong, but I think that with the difference we see between the two most extreme buckets, there should be a larger discrepancy between the 26th-50th percentile and the 51st-75th percentile.




So I’ve probably been on the right side of variance early in my daily fantasy basketball career—especially since I won a tournament for a few grand early in the season—but I’m probably on the right track as well. I don’t think that this sample size is large enough to claim that I’m a long-term winner by any means, so there’s plenty of work left to do, but it probably means that I’ve submitted +EV lineups up until this point.

Also note that I’m really big on searching for overlay, so I primarily enter leagues that don’t fill. I haven’t represented any overlay at all in this data, so my actual results are better than what’s shown. If you can take advantage of overlay so that you need only a top 25 percent finish to cash in a GPP, that’s a huge edge.

Moving forward, I’ll be interested to see how these results change—particularly if I continue to finish in the top 25 percent so frequently. If that happens, I’m going to continue to pump entries into GPPs because, as long as I’m cashing at or near that rate, I’ll be able to maintain bankroll growth while waiting on some high tournament finishes.

I think these results also suggest that I might be able to ramp up my level of play a bit, too. I’ve become a lot more comfortable with creating lineups as the season has progressed, and the data suggests that I’m at least not a horrific player.

Thus far, I’ve been putting right around two percent of my bankroll into play each night as I test stuff out, which is a pretty low percentage. It’s important to note that my early results aren’t at all a guarantee of where I’ll finish in the future—I think some natural regression is certainly in store and I suspect my next 634 contests won’t run as smoothly—but the early numbers are definitely a good sign that I’m going about things in a potentially +EV manner.