Page 1 of 1

Razzball vs NumberFire - DraftKings points projection test

Posted: Wed Jun 17, 2015 11:48 am
by Rudy Gamble
Objective:
Test Razzball DraftKings Hitter and Pitcher Point projections vs. NumberFire (available at http://www.numberfire.com/mlb/fantasy/f ... rojections). Use DraftKings salaries as a baseline.

Timeframe:
May 28th - June 13th

Sample
Hitters - 3,096 (only includes hitters in starting lineups where both services had projections)
Pitchers - 429 (only includes starting pitchers where both services had projections)

Hitter Results (Correlation (r) / Average Pts / Average Absoute Error):
Average Points/Player (Actual): 6.8
Razzball - 0.187 / 7.3 / 5.13
DraftKings - 0.164 / 6.8 / 5.03*
NumberFire - 0.148 / 7.3 / 5.17

Pitcher Results (Correlation (r) / Average Pts / Average Absoute Error):
Average Points/Player (Actual): 14.98
Razzball - 0.299 / 14.8 / 8.92
DraftKings - 0.249 / 14.9 / 9.08
NumberFire - 0.228 / 13.0 / 9.38

Conclusions:
- Razzball DraftKings point projections tested better than NumberFire for both hitters and pitchers.
- Razzball's edge vs DraftKings salaries is consistent with testing since May 2014 (razzball.com/ombotsman)
- NumberFire point projections tested worse than if one used a formula to convert DraftKings' salaries into points.

Notes:
The underlying data is available at:
https://docs.google.com/spreadsheets/d/ ... sp=sharing

DraftKings salaries were converted to Points via the following formulas (based on a regression of projected salary to actual points)
Hitters - 1.003956938 + 0.001539136 * Salary
Pitchers - 4.40163481 + 0.00150703 * Salary

DraftKings' AAE advantage for hitters vs Razzball is a result of the regression formula using actual points. If you adjusted Razzball projections to reflect for the lower average points/player (6.8 vs 7.3), the AAE would be 5.009.