Part of what TMQ does in his predictions is cheating by mathematics. By taking predicting one of a large number of unlikely outcomes he manages to make predictions that look unlikely but are actually quite good. On the other hand many of his points do not involve such cheating. For example predicting the way sporstcasters do the superbowl winner is in fact worse than randomly choosing. Similarly, over the past several years, picking the division winner to repeat gives you 1/8 odds, while randomly picking another team gives you better than 1/5 odds (7/32 to be exact).

However, TMQ then proceeds to precisely ignore all the predictions which he has made. In his yearly Haiku predictions he predicts scores. From these scores you can see who he is predicting to win each division: Eagles, Bucs, Packers, Rams, Patriots, Titans, Steelers, Oakland. Notice that he has thus prdicted that 5 of last years 6 division winners will repeat. Although he might be saying he only predicts one of these to actually win but he doesn't know which one, this prediction itself runs against his numbers. As I explained above, he should actually just be picking random non-division winners and he'd do better than his own predictions.

Consistency people, consistency... Is it too much to ask for?