On our recent podcast with DraftKings Pro Drew Dinkmeyer, I talked with him about the possibility and challenges of predicting ownership. First, we talked about projections in general and how even a sport like baseball in which nothing is very predictable – even the best hitters only get a hit 3 out of 10 times, on average – is equally unpredictable for everyone.

Drew brought up a good point – there’s no threshold you have to hit when projecting player performance. In our player projections, we aren’t trying to hit a specific number. Rather, we’re just trying to outperform the competition. While a top hitter only gets a hit 3 out of 10 times, it’s more useful to think of them in percentiles (Bryce Harper’s .487 wOBA would be the 100th percentile) and adjust accordingly.

Unfortunately, projecting ownership doesn’t have that caveat. There is a specific number we have to hit: the player’s actual ownership in whatever contest. This brings quite the challenge as there is no room for error. In projecting players, you just have to be more right than someone else. In projecting ownership, you’re either right or wrong. There’s no almost.

The first step is finding out what’s important in projecting ownership. What I mean is: what variables correlate strongly with ownership? Unfortunately, there are many variables that affect this. Just brainstorming them for MLB batters, you have batting order, salary, implied runs (Vegas), recent play, public money (moneyline), popular players, good matchups, positional scarcity, and the list goes on.

Some of these things are hard to quantify for a single slate. Drew mentioned on the podcast that he believes he’s better at projecting ownership for NFL because there’s a whole week in between contests and he can more-or-less feel out where the public is leaning. Is there a way to quantify this with data, though? Will there be a projection system one day that accounts for Twitter and Google trends, article hits, Facebook followers, and other ways of “measuring” the public’s perception on a player?

That’s probably down the line, but for now we can start to look at some quantitative stuff that’s easily accessible. Just to start the thinking process, I looked at three variables for a slate on DraftKings – batting order, Vegas projected runs (or implied runs), and a player’s salary. This will be a small sample because it’s only one slate, but I picked the biggest contest I could find recently – the Swing For The Fences tournament on June 22nd that had 95,825 entries.

In case you were wondering, here was the winning lineup by Boopscoop4:

swing for the fences winner

Obviously, ownership played a huge factor in the win, as the Phillies went off that night but were only owned in 1% or less of lineups.

I looked at every player in the slate, made a big Excel sheet, and found the correlations between player’s actual ownership percentage and their batting order, Vegas projected runs, and salary. Here were the results:

Batting Order 0.1328
Projected Runs 0.1868
Player Salary 0.2891


None of these are strongly correlated (a perfect correlation = 1), but it is interesting to see which of these is stronger than the others. And the winner in this slate is salary. I guess this makes sense – the best players are going to be high-priced and especially so in good matchups. I’d like to think it also has to do with the popularity of the top players, but I can’t definitively say that’s true.

There’s an issue correlating batting order – there’s not a straight line of value. What I mean is that batting order value isn’t 1>2>3>4>5>6>7>8>9. Instead, it’s probably more like 4>3>1>2>5>6>7>8>9. To account for this, I put them in three tiers:

Tier 1: spots 1 through 4

Tier 2: spots 5 through 6

Tier 3: spots 7 through 9

This isn’t perfect obviously, but it hopefully should create a stronger correlation. And it does (0.1491), but it’s not a significant jump and is still less than Vegas projected runs and player salary.

The final formula or algorithm – whenever some smart person finds it – will definitely combine all of these variables. They relate to each other as well. Take salary and batting order. You could have two players that are priced at $2,900 on the same team (thus keeping Vegas run projection constant). However, if one is hitting in the first spot and the other is hitting in the eighth spot, the former will likely be much higher-owned.

It will be interesting to see how long it takes for someone to be able to project ownership, even if just kind-of well. Perhaps NBA or NFL will be easier than MLB in this regard because there isn’t the issue of batting order. You still have to worry about a ton of other variables, but even just one less could make things easier and perhaps we’ll see it sooner as a result. So is projecting ownership possible? I’m becoming more optimistic, but it’s certainly very complicated, and probably more art than science at this point.