It comes as no surprise that last weekend _ full of big wins and startling upsets _ has seriously reshuffled the college football polls.
Houston, Wisconsin and Texas made huge jumps after unexpected victories while Ttop-10 losers Oklahoma, Louisiana State and Notre Dame tumbled almost as far.
Now that voters have seen the teams play, as opposed to making summertime predictions based on talent evaluation and gut instinct, they have a better feel for the situation, right?
Not necessarily.
Ed Feng, a Stanford-educated chemical engineer who runs the Power Rank sports analytics site, believes that preseason polls are better at predicting eventual success than anything voters produce later on, and he backs that claim with numbers.
Examining the past 11 seasons, Feng used bowl games _ neutral-site contests between quality opponents _ to see how often the higher-ranked teams won.
The preseason Associated Press poll predicted winners 60.2 percent of the time. The coaches poll did even better at 60.9 percent. But the AP's final regular-season poll predicted only 56.6 percent of the winners and the success of the coaches poll dropped to 55.8 percent.
Habit and emotion might explain at least some of the discrepancy. As Feng puts it, "When the regular season begins, you start making systemic mistakes."
In many cases, he believes, voters drop losing teams a few spots too far. Oklahoma went from No. 3 to No. 14 after its stumble against Houston, a defeat that turned on a missed field goal attempt returned for a touchdown.
"There's the image that Oklahoma has fallen off," Feng said. "The game was really pretty tight."
This dynamic works the opposite way too, with upset winners receiving a strong push. Wisconsin's rise from unranked to No. 10 is among the biggest one-week jumps in the AP poll's history. Texas similarly went from unranked to No. 11 after defeating Notre Dame in double-overtime.
"That was a very exciting game, but I don't think you can put that Texas team in the next category yet," Feng said. "The reality is, they still have a freshman quarterback who is going to make mistakes."
So what would the statistics guy tell voters as the season progresses?
Less emotion, he says. More logic.
