I didn't change my prediction as to who would win in the matchup between the New England Patriots and the Arizona Cardinals. I changed my prediction of probability from 53% to 61%.
Why? Not because I suddenly had higher expectations of New England -- because up to that point, I was at -9.8 points for the week, and a New England win at 61% probability would net me 9.8 points and make the week a wash instead of a loss. I decided to throw better "money" after what I already thought was decent "money."
New England won, 27-13.
This week, I beat FiveThirtyEight's model. I neither picked up nor lost any points, it lost 11.6 points (but it's still way ahead of me for the season, 423.5 to 61.6). I picked eight of 13 games correctly. The model picked six of 13 games correctly. In both cases where my picks disagreed with the model's, my picks were the right ones. And I'm doing better than 66% of humans playing the game.