Favourites Bias in American Point Spreads
Posted 19th November 2015
The following material is adapted from my forthcoming book Squares & Sharps, Suckers & Sharks: The Science, Psychology and Philosophy of Gambling.
By applying a points or goals handicap to a favourite in a 2-way contest, the idea is to equalise the probability that either side will win, and in doing so, balance the volume of money bet to help the bookmaker manage his liabilities. In 2004, the American economist Steven Levitt (famous for the book Freakonomics), challenged this traditional viewpoint, and in doing so, the notion that point spread bettors were collectively acting as a wise crowd. Point spread bookmakers, Levitt argued, do not, in fact, play the traditional role of market makers, matching buyers and sellers but, rather, take positions with respect to the outcome of games, systematically exploiting bettor biases by setting spread prices that deviate from the market equilibrium price. The primary bias in question is an overbetting of favourites, that is to say, teams to which a negative point handicap has been applied.
Analysing the actual number of wagers placed by bettors as part of a handicapping contest offered at an online sports book during the 2001 NFL season (285 entrants making 19,770 bets on 242 games), Levitt found that over 60% of bets were placed on the favourite. Such a systematic bias towards favouring a point spread favourite argues against the hypothesis that bookmakers are doing the best they can to even out the bets on each game. On the contrary, with only 48% of favourites covering the published spreads in the preceding 21 NFL seasons, Levitt speculated that, despite the additional liability of unbalanced markets, bookmakers are intentionally biasing spreads very slightly against the favourite, thereby taking a position with respect to the outcome of a game, in order to exert their superior talent with a view to yielding greater profitability over the long term. Of course, the extent to which they artificially move the spread, however, is constrained to prevent smarter bettors from exploiting such a price distortion, in much the same way that fixed odds bookmakers are limited by how much they can shorten underdogs and lenghten favourites relative to market efficiency, the so-called favourite-longshot bias. The implication from Levitt's research is that point spreads do not represent the crowd's prediction of game outcomes and, therefore, that bettors in such a market are collectively unwise.
Of course, Levitt's sample of bettors was small, the data came from a competition (not real betting activity) and concerned the number of wagers rather than actual volumes of money staked. A bias in the number of bettors backing favourites to cover a spread may not necessarily be equivalently expressed in monetary terms. Levitt, furthermore, failed to offer any explanation for why bettors might be predisposed towards favourites on a point spread market, merely observing that such a systematic bias exists. Following Levitt, Joseph Simmons, associate professor at the University of Pennsylvania's Wharton School, and Leif Nelson, associate professor at the Haas School of Business at the University of California, Berkeley, have provided a psychological account for where the bias originates from: intuitive confidence. Tracking data from thousands of predictions of 850 professional and college football games on Yahoo.com for the 2003 and 2004 seasons, Simmons and Nelson found, as Levitt had previously, that bettors prefer to back the favourite. 65% backed an NFL favourite to cover the point spread. This rises to an aggregate of 70% when college games are included as well.
Crucially, the more people believed a certain team would win, the more likely they were to also choose that team to cover the spread. In other words, the intuitive confidence bettors felt in picking the winner translated into an unrelated belief that the winner would cover the spread. By contrast, the weaker the intuition (for example, where two teams are considered to be more evenly matched), the weaker the intuitive bias. The researchers proposed that such intuitive bias arises because intuitions often spring to mind with subjective ease (particularly when one team is considered to be much better than the other), leading people to hold their intuitions with high confidence. This represent an example of attribute substitution, where an individual has to make a judgment of an attribute that is computationally more complex (determining which team will beat the point spread), and instead substitutes a more easily calculated attribute (which team will win).
As for Levitt's study, however, Simmons and Nelson's initial research investigated only the percentage of bettors backing favourites, but did not consider actual money wagered. In a follow-up study, the authors attempted to address this by tracking the wagers bettors submitted to a website created for the purposes of the experiment. The 178 participants were asked to assign one of five possible wager amounts ($0.50, $1.00, $1.50, $2.00, or $2.50) to each of the 226 Sunday games during the 2007 NFL season. For every game, the size of the point spread presented to the participants had been artificially tilted in favour of the underdog, meaning favourites lost (124) more games than they won (98).
Participants were also assigned to different experimental conditions. Some were asked to predict which team would beat the spread. Of those, about half were actually informed that the spread price had been manipulated. Others were instead asked to simply predict the match winner, along with an estimation of the actual point differential of the game. Simmons and Nelson found the same bias towards point spread favourites as in their initial research. For nearly 90% of games, participants uninformed about the price manipulation wagered more than 50% on the favourite. That figure was only marginally smaller (83%) for those who were told that the spread had been increased. Whilst they haven't reported total volumes of money wagered on favourites versus underdogs, their findings nevertheless are indicative of an unwise crowd, even when given added hints about teams that held value.
More intriguingly, however, those bettors asked to predict a match winner and point differential elicited superior betting choices. By converting the average predicted point differential for each game into predictions against the point spread, the authors found that this crowd more wisely (given the manipulated spreads) predicted the underdog 83% of the time. As the authors suggest, paradoxically it seems that, asking people to estimate the point differentials directly may cause them to focus on the very dimension (the point differential) that receives insufficient weight when making cognitively simpler choices.
These findings would appear to support the authors' theory of intuitive confidence. When forced to engage in a slower and more cognitively taxing task (estimating the actual point differential) bettors are less likely to make lazy and biased judgements based on intuition. Thus, the authors conclude that "although systematic biases may ruin the crowd's judgments when judgments are elicited in a manner that encourages intuitive responding, those biases may be absent from logically identical methods of eliciting the same information, and the crowd may emerge wiser." It would appear, then, that the presence of crowd wisdom is highly context-dependent and can to a significant degree be maniuplated by the bookmaker. In summary, point spread bettors might be acting unwisely simply because they have been given the opportunity to do so (by being lazy).
Intriguingly, the same logic, that is to say, bettor maniuplation by the bookmaker, might also account for the present of the favourite-longshot bias in fixed odds betting markets. I'll take a look at this possibility in the next article.