Thinking in multiples
One useful mental model I have when playing DraftKings is multiples. A multiple for an NBA player is simple to compute. The basic implementation in the DraftKings library I maintain is below:
Why is this a good mental model? Multiples are my means of quickly finding out how much a player contributes towards the final score of a lineup. Let’s assume every player in a 50K lineup of 8 players costs $6,250. If each has the following multiple, here are the scores we end up with:
300+ scores I have generally found to be good enough to win money in an NBA contest, so one rule of thumb is to shoot for each player in your lineup to have at least a 6x multiple.
This rule of thumb unfortunately oversimplifies the reality of NBA DFS. There exist only a handful of players in the league who are consistent in their point outputs — think LeBron James, DeMarcus Cousins, James Harden, Russell Westbrook, Giannis Antetokounmpo and Kevin Durant. Because of this consistency (along with the obvious 50+ point outputs), DraftKings prices each of these players at ~11K+. Thus, to achieve a 6x multiple, a top player needs to hit 66+ points.
Does that mean you should avoid playing the best players on an absolute point basis? Absolutely not. Paying for consistency is a great use of funds. It has smoothed out my winning percentage over time. I’ll gladly accept a 5.5x multiple on $11K with a 90% probability (all of the above average close to 55 points per game) versus a chance at a 7x multiple at 50%.
Let’s look at the table above with some small modifications. I’ll assume below 11K comes in at a 1.5x multiple lower than the other 39K (11K + 39K == 50K), and what it then takes for the remaining lineup to hit the final multiples.
By the way, if you want to see how I arrived at these numbers, the Google Spreadsheet is linked here.
The key takeaway here is that evenly distributed lineups are nearly impossible, and you need to structure your lineups accordingly. The NBA is not a normally distributed talent pool; there are probably 5–10 legitimate stars in the league guaranteed to get significant touches. Everyone else is variable in comparison.
Fortunately, the high variance of individual players moves faster than the DraftKings team can adjust salaries. My experience has been salaries lag behind performance.
Kyle Kuzma of the Lakers is a great example of this phenomena. Check out Kuzma’s multiples over the last several games:
Players like Kuzma offer a margin of safety. They can score below or at their point totals for the last several games, and still offer a strong multiple. Pairing an on-the-rise player like Kuzma with an established but expensive player I have found to be the key to consistently scoring over 300+.
Importantly, the variance principle also applies to players in free fall. The DraftKings team cannot drop the price as quickly as the player’s performance declines. Carmelo Anthony offers a good case study here:
Clearly, Carmelo’s performance is hard to project. Salaries always play catch up to reality.
My focus in recent weeks has been attempting to detect when salaries are falling out of lines with a player’s recent performance. One strong signal here is which players are appearing in winning contest lineups, as high total scores are strongly correlated with high multiples. DraftKings offers full contest results as CSV downloads for all its contests. How to leverage this data and find mispriced players will be the subject of a future post.
I use multiples to guide my lineup building strategy because they represent an easy-to-understand way to gauge player values. As a rule of thumb, 4.5x is an acceptable multiple for a 10K+ player, and players under 10K should be capable of achieving a 6.4x multiple. Hitting these numbers will get you at least to 275; a little luck might push you to 300.
