Don't Trust the Machine: NFL Week 9
NFL gambling as been the bane of my existence for the last.... 9 weeks. Weird how those numbers work. I have been experimenting with different models and so far there really isn't much to be proud of. No matter how accurate these models are in testing, when it comes to actually performing these things crap the bed. So this week I am going to post these crap outputs and see if there is a trend in wins and losses for the algorithm. Worst case scenario, obviously, is a model that performs around 50%. If the model performs at 0% then at least you know you can bet against it.
Here are the outputs below:
Model estimating points differences against total spreads:
Model predicting point totals by team:
For the sake of averaging I essentially ran every game twice from the perspective of each team. So what to make of this? Nothing at the moment, I am not going to bet on a model I don't fully understand. I plan on giving an update tomorrow detailing where the model was accurate and where it needs work.
Data provided by www.pro-football-reference.com
Odds provided by www.bovada.lv