There is still plenty of room for disagreement on this subject. I take the position (PDF) that, while point shaving in amateur sports may be somewhat more common than previously believed, the suspicious statistical trend that he identifies (heavy favorites win yet fail to cover the spread more often than expected) is unlikely to arise primarily from widespread corruption.

I base my opinion on the observation that the same statistical pattern also appears in NBA and NFL game outcomes, yet athletes in major professional leagues are unlikely to engage in point shaving.

NFL and NBA players who are sufficiently talented to reliably influence game outcomes are already wealthy and would have little to gain from such conspiracy. The enormous amount of money that millionaire athletes would have to wager to make manipulation worthwhile would raise suspicion among casinos and, consequently, law enforcement. Were corruption of this sort to be exposed, the conspiring players would not only forego their substantial future salaries, but would also serve time in federal prison.

So, it seems inarguable that point shaving by professional athletes must be far less prevalent than is point shaving by amateurs. Then why in major professional sports do heavy favorites also win yet fail to cover the spread more often than expected?

Admittedly, it is far easier for me to poke a hole in an intriguing theory than it is to come up with an alternative one that fully explains this odd statistical pattern. Nevertheless, I need a good answer to my question before I will become a believer in widespread point shaving.

Rick Borghesi

The true values of prediction market assets, on the other hand, are unambiguously revealed (with few exceptions) upon contract expiry. For instance, Tradesports lists NFL contracts which, once the underlying games end and final scores are observed, are each worth either $0 or $10. Thus, in this setting the joint hypothesis problem is mitigated.

By matching the time at which NFL scores occur with price movements of Tradesports NFL contracts, I have come across an unusual trend (PDF) – selling contracts during active events is profitable, and even more so when the transaction is executed right after a touchdown occurs.

Why?

The structure of Tradesports assets is such that an understanding of short selling is required to make an informed trading decision. On this exchange, when Team A plays against Team B, the contract is framed relative to Team A. That is, a trader must either buy a bet on Team A or else sell a bet on Team A. This is not the case in traditional casino-style markets, where one can buy a bet on Team A or else buy a bet on Team B.

If I were to propose that NFL bettors are generally less sophisticated than equity traders, there would be few objections. Short selling is a procedure with which few NFL bettors are familiar. This deficiency causes a reluctance to sell, creates a supply shortage of bets on Team A, and produces an unwarranted price increase of bets on Team A. The end result is that NFL contracts expire at $0 more often than expected, and the sellers cash in.

Rick Borghesi

• The Late Season Bias: Explaining the NFL’s Home Underdog Effect (PDF file) and;
• The Home Team Weather Advantage and Biases in the NFL Betting Market (PDF file).

Although the latter was published first, I began writing it after the former was accepted for publication (unfortunately long backlog at the publisher). Each paper deals mainly with home underdogs, as I make the point that at least part of the underdog bias is a function of whether the underdog plays at home or on the road. That is, bets on home underdogs win at a rate greater than do bets on underdogs. The main idea, which is nicely summarized on this post published on the “Odd Numbers” blog, is that the home underdog bias in football occurs almost exclusively in the final two months of the NFL season. While this may or may not be the case in other sports, the timing of the bias potentially reveals something about the behavior of bettors and/or bookmakers.

In the first paper, I dedicate the majority of the analysis to show that removing late-season games makes the home underdog bias disappear, and briefly offer that bettors may fail to take into account the impact that winter weather conditions have on game outcomes. In the second paper, I gather weather data that enables me to contrast the temperature to which players are acclimatized and those which occur on gameday. It turns out that cold-weather acclimatization is an important determinant of game outcome, which is not surprising. What is surprising is that cold-weather acclimatization is also an important determinant of bet outcome.

The important question that remains unanswered is ‘why?’ Is it simply that weather is not part of the equation for gamblers? Probably not. Is there a relationship between this finding and that in Levitt (2004 – PDF file), who shows that bookmakers may exploit bettor biases? Do bookmakers take a naked position on the acclimatized team just as they do on heavy underdogs?

Incidentally, this weather bias is not limited to sides bets. In ‘Weather Biases in the NFL Totals Market (forthcoming),’ I show that a simple and profitable strategy can be implemented to take advantage of the relatively slow speed with which totals lines move in response to changing weather conditions. Temperature, rain, humidity and, even more so, wind, all affect the total number of points scored. These determinants are known with certainty only immediately prior to kickoff, yet gamblers place totals bets in the days prior to each event. What is the point of placing a totals bet before this information is known? Unless you know what no one else does, your risk (but not payoff) is greater.

Obviously, gambling is entertainment, and to some the risk must be a significant part of the reward.

Rick Borghesi