2006 Game Predictions

The first week I tested the game-by-game model was week 5 of 2006. I was initially encouraged because it correctly predicted all 14 games that week. That streak of luck would not last long, however.

But keep in mind it doesn't really predict a winner, it produces probabilities. So for the BAL vs. CIN game, it might say BAL 0.60 CIN 0.40. Baltimore is favored with a 60% chance of winning. So if CIN wins, that doesn't necessarily mean that the model was wrong. But for the statistical purpose of judging the "fit" or validity of the model, we'll say that if the predicted favorite wins, the model was correct.

For week 5 the model produced the following probabilities: