Early motivations

A lot of my interest in starting analytics came out of Reddit discussions. It seemed like there were a lot of theories floating around, but there was no data to support them.

Here's an example referencing an old discussion-- Did QBs score more fantasy points if they were on bad defenses? There were arguments on both sides at the time. A bad defense meant the QB had to throw more to catch up.... or was it actually that a good D got off the field more quickly so the QB had more chances? As you can see, I found no trend when just looking at fantasy scores. To me this meant there was a lot of room to try and clarify trends.

Anyway, I first got interested in making rankings because there were 3 things I found missing from many ranking sources:

• I felt there must be room to improve their predictive accuracy. 

  • With D/ST, I felt like in-season matchups were not being compared correctly, while last-season biases lingered, etc.

  • With Kicker, I thought rankers were biased toward "feel good" candidates based on prior years. And I guessed there wasn't enough investigation of how kicking depends on various offensive qualities.

  • And with QB, I strongly suspected that there was a lot of crowd-pleasing in the rankings. I especially thought there's not enough attention to the strengths of the opposing defense

• Rankers were not disclosing their methods. Without knowing what information they assessed, it was hard to judge how much stake to put into them.

  • Did they make guesses? Use their eye-test? Borrow external stats? Crunch their own numbers?  It seemed a lot like there was intuitive guess work, instead of measurements and tested math. 

• I wanted projections beyond the current week. I wanted to see, at the very least, 1 extra week look-ahead of D/ST projections, for planning purposes.

This is why (1) I went deep into model development, (2) why I tell you my methods and what numbers I consider, and (3) why I put some focus on future week forecasting.