Friday, September 05, 2008

No NFL Bettor Bias at Tradesports?

Philip O'Connor and Feng Zhou offer me a sigh of relief by inferring that the already low odds that my Browns will win the Super Bowl are not inflated by long shot bias in the Journal of Prediction Markets. Here is the abstract:
We investigated 1,587 Tradesports point spread contracts for NFL games during the 2005/06 season. Differing point spreads create differing odds, meaning we could test for the traditional favorite long shot bias in NFL betting. We found that there was no favorite long shot bias. However, the market underestimated the chances of the favored team winning by about 10% across all odds categories, and this bias persisted throughout the season. We found relatively low transaction costs. For a price-taker, the Tradesports "Vegas-line" point spread had a 2.2% total takeout including exchange fees, about half of the 4.55% takeout of traditional legal bookmakers. Contracts with a price around 50, creating even money returns to bets on both teams, and higher volume contracts, had lower transaction costs. Participants were found to prefer the Las Vegas line point spread contract followed by the straight-up contract. Trading volume during the game (in-running) was about twice the trading volume leading up to the game. Teams with better season records and from cities with larger populations generated a higher volume of trades. Sunday night and Monday night games generated about four times more volume than regular Sunday games.

2 comments:

Matt E. Ryan said...

"Teams...from cities with larger populations generated a higher volume of trades."

I really believe there's something here as a source of predictable bias, and I've been trying to get at it with my own data in baseball but haven't had a ton of success as of yet.

Thomas said...

You may also be interested in this:
http://scienceblogs.com/cortex/2008/09/football.php