Below please find the CSML prediction results for week eight of the 2015 NFL season. This model uses a learning algorithm fed by data from the last seven weeks.
This past week, the model predicted the outcome of 10 of the 14 games correctly for an accuracy of 71.4%. The crowd, on the other hand, was correct on 9 of the 14 games for an accuracy of 64.2%.
Four of the same games were incorrect for both the model and the crowd. The one that diverged was the SD @ BAL game which was close and could have gone either way.
As a caveat, it is worth pointing out again, that these predictions were based on last year's method, which uses a variety of averaging and weighting procedures based on game scores. The model we are currently building is a multi-layered, feature detection network that uses AI rather than statistical procedures. Time will tell if it does better than this method. Please see my recent post regarding the "Predicting the NFL" Update for more.