An innovative approach to National Football League standings using bonus points

Abstract

We investigate the inclusion of bonus points in the National Football League (NFL) using a prediction model built on league points. Both touchdown-based and narrow-loss bonuses are shown to be significant determinants of match outcomes. This implies that including bonus points in league standings generates a more accurate ranking of teams from best to worst than a system that only rewards wins and ties. Our preferred system awards four points for a win, two for a tie, one point for scoring three or more touchdowns and one point for losing by seven or fewer points. Such a system would also make it easier for supporters to identify playoff contenders and place importance on otherwise meaningless end-of-game plays in some matches, which would increase spectator interest.

Keywords:

• tournament design
• sports predictions
• sports ranking systems
• NFL

JEL Classification:

• C20
• C60
• L83

Notes

¹ See Stefani (1997, 1999) for assessments of SRSs and Szymanski (2003) for a review of tournament design issues.

² Several related studies scrutinize National Collegiate Athletic Association (NCAA) Division 1-A college football ratings (Bassett, 1997; Harville, 2003; Mease, 2003; Stern, 2004; Frey, 2005; Beard and Caudill, 2007; Caudill, 2009).

³ In related studies, Dare and Holland (2004), Boulier et al. (2006) and Borghesi (2007) examine efficiency in NFL betting markets.

⁴ We do not account for changes in player and/or coach behaviour induced by bonus points. Notably, a touchdown bonus may induce coaches to choose a passing or running play instead of a kicking play. In a theoretical setting, Banerjee et al. (2007) show that the introduction of the NHL overtime-loss bonus induced player behaviour changes. However, as a win is worth four times the amount of a bonus in our preferred NFL specification, we speculate that introducing bonuses will not stimulate major tactical changes. Investigating this conjecture presents an interesting avenue for further research, perhaps building on Romer (2006). We elaborate on bonus-induced behaviour changes in our conclusions.

⁵ Our decision to use a small number of home advantage parameters does not allow us to include team fixed effects to account for unobservable characteristics. In unreported simulations, we verify that our results are robust to the inclusion of these parameters.

⁶ Although N^MA  = 13 and N^TV  = 12, our bonus recommendations are relevant for determining playoff contenders based on a 16-game regular season as the bonus system should be invariant to the period evaluated. N^MA and N^TV are selected to determine the rate at which previous results should be discarded so that strength measures are current, rather than to reflect season performance as required by end-of-season rankings.

⁷ Our preferred allocation of competition points is similar to that used for most rugby competitions. However, Winchester (2008) finds that the try bonus used in rugby is not significantly correlated with team strength, but he does find a significant relationship between strength and a net-try bonus. This disparity is probably due to differences in how players function across the two codes. In rugby, the same set of players is responsible for both offence and defence, so a team with a comfortable lead may relax defensively. In contrast, NFL teams have separate offensive and defensive squads and each player is carefully evaluated by coaches, making dominant teams less susceptible to defensive lapses. Ultimately, a weak NFL team is less likely to gain a touchdown/try bonus than a weak rugby team.

⁸ Classification changes observed under the away win rule that are reported in Table 3 include classification exchanges between Miami and Denver in 2003, and New Orleans and St. Louis in 2004; and classification changes involving Chicago, Carolina and Tampa Bay in 2005. The classification change under the away win rule not observed under the bonus system involves Denver claiming wild-card qualification at the expense of Kansas City in 2006.

⁹ Classification changes when net points are used as a tiebreaker include the classification exchange between New Orleans and St. Louis in 2004, the classification changes in 2005, and the classification exchange between Cleveland and Tennessee in 2007.