Below is the summary of the model’s performance for Round 1, 2016. Some key points:
- The model predicted 7/9 winners based on the final teams named in each game.
- The model predicted the winner on the line 4/9 times.
- The model made a loss of 15%, mainly due to big losses on Essendon and Brisbane on the line. Caution should be exercised when betting on Essendon in the first few rounds as the model probably overrates Essendon, given the WADA suspensions.
- The errors of the model were actually relatively small compared to the average, but the errors on the two largest bets (Essendon and Brisbane on the line) went against the recommended bet.
- The predictions for total points (not shown) were poor, with much higher scores than last season. I will wait discontinue this model from now, at least until several rounds into the season.
AFL Predictions 