NOD said:
SMTTEM -- I've always wanted to ask, but I have trouble thinking of how to ask the question without seeming like an ass -- why do you care so much about the line? Do you think it correlates to reality in some way? Do you think that it's an accurate approximation of the true spread? I mean, after all, all it is is an attempt from the gambling community to maximize betting on both sides of the line. In actuality, its correlation to game results is nothing special in the realm of statistical significance.
I think the gambling community does do a good job in assessing the two teams and predicting what should happen in the game. When I see UNC favored by 10, it makes sense to me that UNC should be 10 points better than Duke at Duke. Statistically, the Vegas spreads in NFL games are an average of 9 points off, which is really good when you remember that 0 points off is the best possible outcome (there are no negative numbers weighing the average variance down). Historically, they beat every computer or human predictor. So for every crazy result like Packers by 18 in Houston or Giants by 23 in San Francisco, or Lehigh over Duke, the oddsmakers are nailing it in several other games. In college football, there will be more variance than in the NFL, but I don't know the exact number - probably something like 11-12 points off on average. Which means Duke could very well win this game by a point or two and no one in Vegas would be surprised at all, but if UNC and Duke played 10 times on 10 consecutive days starting this week, with no significant injuries, everyone but the irrational homers would expect UNC to win by an average of around 10 points.
They ideally want 50/50 betting, which means they're not necessarily "predicting" what will happen, but how the betting public wagers their money will generally will set a good "prediction" anyway. When real money is on the line, people almost always act rationally. When they don't, we get the once a decade or so disaster like dotcom or the housing bubble. For example, if Vegas sets the opening line of Alabama vs Notre Dame as Notre Dame being favored by 100 points, zero people in the world are taking Notre Dame -100. Anyone with any money and an internet connection is putting their life savings on Alabama to stay within 100 of Notre Dame, and the casinos are going bankrupt. So obviously, on the extremes, it makes no sense for the casinos to make a bad "prediction." If you get closer to a normal example, like say Duke being only 6-point underdogs vs UNC, Vegas is still going to be taken to the cleaners with that kind of line. They also might win big with a great Duke performance, with everyone losing on UNC -6, but it's much more likely that they lose big when Bernard goes for 200 yards rushing and UNC wins by 7-14. They have to go with something within 2-3 points of the best possible prediction, or else they lose money, and they generally have no reason to deviate from the best possible prediction at all. Extreme cases of unequal homer support, like Dallas Cowboys vs Jacksonville Jaguars, might result in a line being 1-2 points off the best possible prediction, but not enough people care about Duke or UNC football on either side to make this line deviate.