We have been tracking NBA Ownership Percentages in Cash games (Double-Ups) and GPP tournaments all season long and using that data to project ownership percentages based on the number of times a player is recommended by the DFS experts that we grade. I decided that since we over halfway through the NBA season it was time to analyze the data we have collected and determine how accurate the projected ownership percentages have been so far. If you want to skip the analysis and get right to changes being made for “smarter” projected ownership percentages then jump down to “The Solution” section.

The Analysis

Overall, our projected ownership percentages have performed pretty well with our Cash game projections averaging 1.61% higher than the actual ownership percentage and our GPP projections averaging 0.83% higher than the actual ownership percentage.

However, when I broke it down by the number of times a player has been recommended, I noticed that we performed pretty well for players recommended by eight or fewer experts but didn’t perform as well for those recommended nine or more times. Here is a table of the results (through February 1), the average difference between our projected ownership minus their actual ownership.

# of Experts Cash Own % Diff GPP Own % Diff Occurrences
15 (0.65%) 4.55% 2
14 10.14% 6.90% 5
13 3.91% 4.65% 17
12 9.62% 6.73% 17
11 6.99% 5.52% 35
10 4.84% 1.84% 64
9 9.68% 3.83% 80
8 0.85% 0.45% 113
7 1.46% 2.08% 144
6 3.26% 1.75% 223
5 2.68% 1.62% 285
4 2.44% 1.64% 335
3 3.24% 1.79% 496
2 (0.38%) (0.52%) 666
1 0.28% (0.08%) 1186
Total 1.61% 0.83% 3669

When analyzing the data further to determine why the most recommended players were off by more, I found that we are under projecting cheaper priced players compared to higher priced players. Here is a table with the average Cash game ownership percentage differences by salary range. The average GPP ownership percentage differences were similar. Prices are using DraftKings salaries. Here is a table of the results (through February 1), the average difference between our projected ownership minus their actual ownership.

# of Experts $9K+ $7k – $8.9k $5k – $6.9k $3k – $4.9k
15 (5.10%) 3.80%
14 19.53% (3.95%)
13 10.28% 0.95% (0.92%) (23.30%)
12 13.98% 4.00% 5.93%
11 13.07% 3.41% 14.09% (18.03%)
10 12.08% 5.66% 6.24% (21.73%)
9 11.99% 4.43% 14.23% 3.41%
8 5.84% 4.71% 0.82% (10.65%)
7 4.34% 1.91% 1.74% (2.64%)
6 3.77% 4.51% 4.01% (2.06%)
5 2.93% 2.86% 4.61% (2.84%)
4 3.34% 3.55% 3.90% (1.57%)
3 3.81% 4.36% 3.94% 1.32%
2 1.70% (0.65%) 0.23% (0.90%)
1 2.32% (0.79%) 0.34% 0.34%
Total 6.79% 2.60% 2.50% (0.44%)

Looking at the table above the biggest differences occur with the high-priced players being projected for too high of an ownership percentage and the low-priced players being projected for too low of an ownership percentage. So we need to examine what adjustments we can make to better project ownership percentages.

The Solution

In order to more accurately project ownership percentages, we are going to use the historical ownership percentages by the number of expert recommendations AND by salary range. This will result in reduced projected ownership percentages for high-priced players recommended by a lot of experts and increase projected ownership percentages for low-priced players. For example, a player that costs $9,500 and is recommended by 10 experts will have a 27% projected ownership, while a player that costs $3,500 will have a 58% projected ownership.

Here is an example of what the Projected Ownership Percentage tables for Cash games will look like going forward with a similar breakdown for GPP tournaments:

Count $9k+ $7k – $9k $5k – $7k $3k – $5k
15 63% 60% 65% 77%
14 51% 50% 64% 76%
13 50% 49% 53% 75%
12 44% 48% 49% 64%
11 33% 45% 41% 63%
10 27% 34% 33% 58%
9 21% 28% 25% 36%
8 17% 20% 24% 35%
7 15% 19% 18% 26%
6 15% 14% 15% 22%
5 12% 12% 11% 19%
4 9% 9% 9% 15%
3 6% 7% 6% 9%
2 3% 5% 4% 5%
1 3% 4% 3% 3%

Look for these “smarter” projected ownership percentages in the NBA Industry Consensus and Weighted Consensus articles going forward. We will continue to analyze the data and make improvements as needed.

Short URL: http://tinyurl.com/hswzb58