Data scientist, physicist, and fantasy football champion

Week 1 QB Results

This week was basically flipped upside down. Drew Brees, Aaron Rodgers, and Tom Brady were all expected to do very well , but they just didn’t cut it. Alex Smith, Matthew Stafford, and Sam Bradford were the top 3 QBs this week, and none of my models called that. Some were higher on Stafford and Bradford than others, but absolutely none put Smith even in the top 15. Heck , Goff and Siemian even ended up in the top 15.

The way the scores shook out this week, there were basically 3 tiers: Smith, Stafford, and Bradford score in the mid-20s for fantasy points, then the next 11 all scored between 13 and 19 points. Basically, a huge step between tier 1 and then a slow taper for tier 2. No models predicted the top 3, so the best scoring models are all just predicting tier 2.

Let’s see which models did a good job last week:

[note: I dropped my predictions for Jameis Winston and Jay Cutler. Keeping them would have unfairly dropped the accuracy of my models. Nobody who was paying attention started them last week.]

.accuracy graph-1.png


It’s pretty much Model A, then everyone else, then Yahoo. That was unexpected since I really like my Bayesian models, Models D and E. I also would have thought that Model F would have done a little better since it had Brady outside those top 3 and Bradford in the top 10. This was actually so unexpected I went back and checked my scoring function to make sure I hadn’t done anything wrong. I didn’t, it’s just that Model F got a few big names right but really jumbled up the middle. Model A missed some of the top and bottom players, but did a great job the 4th through 13th ranked QBs for the week.

So that’s it. One week down, and it looks like Model A is the model to beat, though I’m still going to keep being a little excited about Model F. Models D and E have room to come back; maybe next week won’t be so crazy.

Week 1 K Results

Week 1 DEF Results