I thought it might be interesting to compare my efficiency model's predictions with the consensus predictions. Probably the best estimates of the "consensus" predictions are the Las Vegas betting lines. Although I'm am not interested in gambling, betting odds and point spreads are useful as benchmarks for evaluating the usefulness of all this number crunching.
Note: The over/under lines are often in .5 increments. In these cases I rounded to the win total that is costlier to bet on, and therefore considered more likely. For example, if the line for a team is 9.5 wins and to bet on the 'over' it takes $120 to win $100, and to bet on the 'under' it takes $115 to win $100, I rounded to 10 wins. When the over and under were equally costly I left the prediction at a .5 increment.