The NHL's overtime-loss rule: empirically analyzing the unintended effects.

In 1999, the National Hockey League (NHL) altered the payoff structure for regular season games that went into overtime. With this so-called "overtime-loss" rule, teams that lost during the five-minute sudden death overtime period were now awarded one point in the standings, rather than zero

points, as had previously been the case. Prior to this rule change, risk averse coaches were reluctant to adopt offensively aggressive strategies during overtime. Instead, they were content to play for a tie--i.e., to take the relatively certain single point that resulted from playing defensively, rather than gambling on the all-or-nothing outcomes (i.e., either winning or losing) that came from playing offensively aggressive. With the rule now guaranteeing teams at least one point from overtime, the NHL was inducing teams to play more offensively aggressive during overtime, thus both increasing excitement for fans and resulting in fewer games ending in ties.

However, we assert that there will be a second, unanticipated effect of this rule change--namely, that it induces even more defensive strategies during regulation time. For games that are tied in the late stages of regulation time, teams now have an incentive to defer any offensively aggressive strategies until overtime, where they are guaranteed one point even if they lose. This effect should manifest itself in more games being tied at the end of regulation time, causing more games to go to overtime, and partially negating the desired impact of the rule.

To empirically test these two effects, we employ team-level panel data (489 observations) for the 1983 84 to 2002-03 seasons. The first (intended) effect is estimated by: PW = 0.078 + 0.182 OTL + 0.038 GOAL + 0.001 DIFF, with [R.sup.2] = 0.26, and with all coefficients of the expected sign and significant at the 5 percent level. PW is the percentage of a team's overtime games in a given season in which a winner is produced; OTL is a dummy variable set equal to 1 for those seasons where the overtime-loss rule existed; while both GOAL and DIFF are control variables. GOAL measures the total (for and against) number of goals scored per-game for a team during a season, while DIFF measures the extent to which a team's quality (i.e., points in the standings) deviates above or below the league average for that season. Consistent with the NHL's objectives, the results indicate that the overtime-loss rule increases PW by 18.2 percent.

The second (unintended) effect is estimated by: OTP = 0.245 + 0.036 OTL - 0.011 GOAL - 0.0008 DIFF, with [R.sup.2] = 0.29, and with all coefficients of the expected sign and significant at the 1 percent level. This result indicates that the percentage of team's games that were tied at the end of regulation time, and were thus forced to go to overtime (designated by OTP), increased by 3.6 percent during those seasons that the overtime-loss rule existed.

This latter effect works counter to the NHL's original objective, It serves to at least partially offset the intended effects of the overtime-loss rule. (JEL L83)

NEIL LONGLEY AND SWAMINATHAN SANKARAN

University of Massachusetts--U.S.A. and University of Regina--Canada