Behavioral biases in the NFL gambling market: Overreaction to news and the recency bias

Abstract

This paper examines the recency bias and overreaction in the NFL betting market from 2003 to 2017. Consistent with the recency bias, bettors are more likely to bet on teams who have won previous outcomes. We add to the literature and find that the magnitude of prior wins and losses in the previous weeks’ plays a greater importance than the sole outcome of wins and losses in betting behavior. Additionally, our results show that bettors wager 2.1% less on the home team when their first-string quarterback does not play, and 3.1% more on the home team when the visitor’s first-string quarterback does not play, which is consistent with overreaction. Finally, our results show that bookmakers earn “over the odds” thanks to bettors’ quasi-rational behavior as they commit the recency bias.

Introduction

The field of behavioral finance has identified a plethora of behavioral and cognitive patterns affecting rationality during financial decision-making. These predictable patterns of quasi-rationality or ‘biases’ are rooted in our psychobiological makeup. The related literature has traditionally focused on monetary decisions made within the framework of traditional financial markets. In this paper, we step outside of the conventional focus on ’financial investors’ and concentrate on gamblers betting on National Football League (NFL) games. We find that NFL gamblers are afflicted by biases identified in the behavioral finance literature, specifically, the recency bias and overreaction to news.

The recency bias refers to the cognitive pattern of over-emphasizing recent data or events while overlooking less recent ones. The recency bias is also related to the availability heuristic examined by Tversky and Kahneman (1973), as individuals consider the probability of events based on the relevant information that comes to mind, and thus overweight more current information over less current information. Overreaction, like the recency bias, is related to Tversky and Kahneman’s availability heuristic. The overreaction bias refers to the overt exaggerated response to new and recent information, which results in securities being overbought or oversold in financial markets. Therefore, intuitively, said biases are often studied together.

In this paper, we explore how previous games’ outcomes influence bettors’ wagers during the thirteen NFL seasons that took place between 2003 and 2017. The bookmaker sets the spread and total (discussed below) with knowledge of the previous game’s outcomes. As such, the bookmaker has already incorporated the recent success (or lack thereof) prior to the gamblers placing their wagers. In fact, research shows that betting on teams with recent success has a negative and significant impact on game outcomes (Camerer, 1989, Paul and Weinbach, 2005, Shank, 2018). Thus, betting on teams with recent performance can be described as a bias given that it results in a further negative expected return.

Consistent with the recency bias, we find that bettors are more likely to wager on the home team if the home team covered the spread more often in the previous three weeks, and the visitor team covered the spread less often consistent with prior literature. We add to the literature by demonstrating that the magnitude of how many points each team covers, or fails to cover, the previous weeks’ spread plays a greater role in gambling behavior than simply if the team has won or lost its previous game. Moreover, NFL bettors also overreact to news about the availability of the first-string quarterback. We find that gamblers wager 2.1% less on the home team when their first-string quarterback does not play, and 3.1% more on the home team when the visitor’s first-string quarterback does not play. Additionally, these results are corroborated when examining the totals market. Finally, we show that bookmakers enjoying profits, after transaction costs, in excess of what might be expected in a fair game due to bettors’ quasi-rationality wagers from committing the recency bias.

The sports betting market provides several benefits over traditional financial markets in considering behavioral biases. First, the potential payouts of gains and losses are known – that is – the expected returns — can be observed. This stands in contrast to financial markets where investors’ expectations may only be inferred by complex models of expectations. Second, the games on which the bettor places a wager have a short duration with a known end date, whereas, for equities, investors’ time horizons cannot be observed.. Third, sports betting markets share similar essential features with the stock market, such as liquidity and wide information availability (Thaler and Ziemba, 1988, Avery and Chevalier, 1999, Durham et al., 2005). A further argument proposed in favor of using sports markets to study market efficiency is that they are idiosyncratic. “That assets’ risk is idiosyncratic implies that an asset pricing model is unnecessary. Therefore, tests of market efficiency and/or superior individual performance do not suffer from the joint hypothesis problem (Fama, 1970)” (Andrikogiannopoulou and Papakonstantinou, 2018 p. 1957).

The two major forms of betting in the NFL are the spread and total, which are both set by the bookmaker. The spread specifies the least margin by which the favorite team must win. As such, bettors can wager on the favorite to win by this margin (i.e., the spread) or the underdog to lose by less than the margin or win outright. The total refers to the number of points both teams will combine to score. The bettor can then bet the game outcome will be “over” or “under” this total. For example, in a game between the Pittsburg Steelers (the favorite) and the Cleveland Browns (the underdog), if the spread is -10 (i.e., the Steelers must win by more than 10 points) and the total is 50, a bettor who bets on the Steelers to win by more than 10 points (also known as covering the spread) and on the “over” will win if the actual score is Steelers 31–Browns​ 20.

The bookmaker does not payout at a 1:1 relationship and is not a pari-mutuel betting system. The vast majority of NFL games against the spread and in the totals market have a payout of −110, which means that the bettor must wager $110 in order to win $100. The $10 (4.55%) commission the bookmaker charges to make a bet is called the vigorish.¹ Due to the vigorish, a gambler would have to consistently win 52.38% of their bets to earn a profit.² If the bookmaker has no private information about the market and is uninterested in taking a side, it should set the spread and total in each game, so the percentage of bets on each side (i.e., the favorite “covering”/ “failing to cover” in the spread market and “over”/ “under” in the total market) is between 47.62% and 52.38% to ensure a profit. However, most people are unaware that bookmakers do not function this way, as they do not aim to have equal amounts wagered on each side of the bet. Levitt (2004) examines the behavior of bookmakers and finds that the bookmaker rarely has an equal amount bet on each side of a gamble, which is confirmed by Paul and Weinbach, 2007, Paul and Weinbach, 2011, and Shank (2019a). Levitt (2004) argues that bookmakers use their expertise to set the betting lines in order to maximize their profits as they can set the lines to get more than 52.38% on the losing side on average and earn more than a 5.45% profit on average.

Our results have considerable implications due to the size of the sports gambling industry. As it relates to the NFL, the American Gambling Association estimated that nearly $100 billion was wagered in 2016, with almost $5 billion occurring on the Super Bowl alone. However, in 2016 only Nevada and Delaware had fully legalized sports betting. Conversely, as of the end of 2019, 17 additional states had legalized sports betting, with more currently working on new legislation. As such, the $90 billion gambled in 2016 has likely already increased significantly due to being more available and will likely continue as more states make it legal. Furthermore, as it relates to all sports, H2 Gambling Capital (2013) estimates that the total amount of global wagers on sports was near $1 trillion in 2012. As such, the ramifications of our results could have large monetary impacts.