From time to time, sports fans discuss overtime.
In baseball, extra innings are easily understood and implemented. In a sport filled with tradition (except for the one bizarre exception that MLB’s two conferences play with entirely different lineup structures and rules), wrinkles like starting the 12th inning with a runner on second base don’t get a lot of support.
Basketball has so much scoring that continually adding five-minute tie-breaking overtime periods has never been controversial. Double overtime is rare enough that alterations don’t generate much support.
Hockey has tweaked its rules the most. It’s hard to remember how many players play for how many minutes and how a shootout works from season to season in the NHL. There are even alterations in the overtime rules that make calculating the standings less than intuitive. Thankfully, so many teams reach the playoffs after a marathon 82-game season that casual fans are content to trust the system and wait until playoff hockey in April when everything is sudden death and six-a-side and shootouts are relegated to a distant and vaguely unpleasant memory.
Soccer embraces the playoff shootout concept because games can remain 0-0 seemingly longer than a cricket test match and overtime is largely absent from the regular season. While periodically (every four years, with the World Cup), interest spikes in the US, this might be part of the reason MLS is a distant fifth in the US. While the “Big Four” North American sports leagues are 1st, 2nd, 3rd and 5th in the world in annual revenue, the MLS is 14th – seventh among world soccer leagues. Whatever the reason, we view sports very differently from the rest of the world.
And football is our sport. The NFL is the world’s largest league in terms of revenue – about $13 billion per year. It also supports more franchises – 32 – than any major league in the world. The Super Bowl is the world’s largest single-day sporting event.
But how does football handle overtime? That’s a tough one. College football has an alternating possession rule which somewhat resembles a shootout. The NFL allowed ties in the regular season for a long time, then implemented a single overtime period, then has tweaked that rule a couple of times in the last decade. Playoff games are not limited by time, though there’s so much scoring in the game that this has never been a problem.
Recapping the rules: in 1974, the NFL added a 15-minute sudden death overtime period for regular-season games. Overtime starts with a coin toss and a kickoff. If the game is still tied after 15 minutes, that’s the final result.
People complained the coin toss was too arbitrary and the risk of scoring quickly too great, so in 2010 for playoff games, NFL teams were guaranteed a possession in overtime unless the team receiving the initial kickoff scored a touchdown. In 2012, that exception was extended to the regular season. This year, the overtime period was shortened to 10 minutes for regular-season games.
To make matters even more complicated, in 2011 the NFL moved the kickoff from the 30-yard-line to the 35 and in 2016 touchbacks after kickoffs were moved from the 20-yard-line to the 25. These are both changes that affect overtime.
That’s an awful lot of rule churn for a league that’s number one in the world by a large margin.
Does it matter? I looked in depth at the 324 overtime games in the NFL regular season and playoffs over the last 20 seasons. In terms of studying the game, I like to consider 1978 and the modifications to passing rules the birth of the modern game and 1998, when I started work on Front Office Football and Peyton Manning entered the league, as the beginning of the time period when studying play-by-play can generate results relevant to today’s game (though it’s always best, in my opinion, to focus on the most recent five years of results).
Here are some of my findings:
Since 1998, 6.0% of regular-season games have gone to overtime and 9.1% of playoff games have gone to overtime. The difference doesn’t surprise me since playoff games are limited to the top twelve teams. The average difference in skill level between teams should be lower in the playoffs.
One factor I considered is momentum in games. Announcers and fans like to talk about momentum being significant. Some of that might be more effective strategy changes during the course of games and some of it might be the mental attitude of players or maybe even the results of injuries.
I’ve often wondered if I should add a momentum factor to Front Office Football. In the original design for the college game, I drew an elephant named Mo who was going to appear on the scoreboard to indicate a team had momentum. But ultimately, I felt it was something that would annoy people more than make them feel more invested in the game. To date, I’ve never added a momentum factor or catch-up factor to either game and it’s unlikely I ever will. My design theory is that results should always be ratings driven and there should only be tangible adjustments to ratings. Most importantly, no ratings adjustment aside from injury should ever be, on its own, more than a small percentage. That keeps the game “honest” and understandable, though the “die rolls” associated with any kind of simulation mean individual play results can vary considerably and sometimes give people the impression that something else could be going on. I can only assure you that I test this kind of thing extensively and that everyone has runs of bad luck from time to time. Anyway…
I counted teams that scored last in regulation as having momentum. These teams won 54.8% of overtime games from 1998 until the two-possession rule took effect (essentially with the 2011 playoffs, as no 2010 playoff game went into overtime). With the new rule, 63.3% of the teams with momentum have won.
Out of everything I looked at with this study, this was the most surprising. It suggests that not only is momentum real, but that the rule change – guaranteeing that the team with momentum gets the ball in overtime – has made it more significant. Maybe even made it more fair if fair is defined as treating overtime as an opportunity to extend a game. Is the difference worth noting? I think so. That’s an 8.5% difference with 98 results since the change. Running a test assuming a normal distribution of results, I get a p-value of 0.079 on the comparison. This means I can’t reject the assumption that nothing has changed with a confidence level of 95%, but it’s awfully close. Another two years’ worth of data will help.
What about the role of the first possession? Before the change, the team with the ball first won 58.5% of overtime games. Since the change, that number has dropped to 53.1%. If the goal was reduce the impact of the coin toss, it looks like that has happened, too.
Before the rule change, teams won on that first possession 33.9% of the time. The goal was simply to score points. Nothing wrong with that strategy, but it heightens the impact of the coin toss. Of the 75 first-possession wins, 61 came from field goals and 14 came from touchdowns. The TD percentage was 18.7%. In 2017, for example, that TD percentage (TDs divided by TDs + FGs) was 58.0%.
Since the change, teams win on that first possession 19.4% of the time. They’ve scored 20 touchdowns and 18 field goals (the field goals allow the other team a possession). That percentage is 52.6%. So what we’re getting is a first possession in overtime that feels a lot like a normal in-game possession. Good or bad? I don’t know.
Before the change, a team’s odds of winning if it failed to score on the first possession was 38.1%. That has decreased to 34.4%. If a team scores a field goal on its first possession, it wins 73.3% of the decisions (61.1% if you factor in ties – three of the five ties in the NFL since the change are among the 18 games where both teams started overtime with field goals). Overall, the winning percentage once a team has failed to win (or lose) on its first possession is 42.6%.
Before and after the change, teams had about a 2% chance of losing the game on the first possession due to a fumble or interception return.
Another factor to examine is home-field advantage. Since 1998, the home team wins 57.9% of the time. Before the change, the home team won 52.5% of overtime games. Since the change, the home team has won 59.4% of overtime games. Because that’s a bit closer to the overall mean, it also suggests the new overtime rules are helping with the problems that motivated the changes.
What is the tradeoff? More ties in the regular season, undoubtedly. We’ve seen ties in 5.2% of regular-season overtimes since the change. That’s compared to 1.0% of overtimes before the rule change. Now that overtimes are reduced to 10 minutes, I expect even more ties. That didn’t happen in 2017. But since 11.7% of overtimes since the rule change last four possessions and 12.6% last five or more possessions, I wouldn’t be surprised if a high percentage of those games end in a tie. My guess is the tie percentage will go to about 10% of overtimes in the long term. It’s hard to figure in strategy changes, but that seems like a reasonable estimate. The net result will be an average of about 1.4 ties per season. That’s not going to break the NFL.
Another topic I examined was weather. Is it ever to a team’s advantage to kick off to begin overtime? Between 1998 and the rule change, teams chose to kick off twice after winning the overtime coin toss. In 2000, in a snowstorm with winds gusting to more than 40 mph in Buffalo, the Bills kicked off to New England. The strategy should have worked as the Patriots failed to score and Buffalo drove to the 12 on its first possession. But a 30-yard kick was blocked, then New England drove down the field and kicked the winning 24-yard field goal with just a few seconds left on the clock. Then in 2002, Marty Mornhinweg pretty much sacrificed his job as Detroit’s head coach by choosing to kick off to Chicago in overtime, only to see the Bears kick a field goal on their first possession.
One problem with using weather is that the NFL doesn’t consistently report data for each game. For instance, the stories I’ve read about the Mornhinweg game indicate winds were at 17 mph when he made that fateful decision. My research suggests that wind significantly reduces offense, making it particularly hard to pass. But the NFL reported a wind speed of 4 mph for that game, which shouldn’t trouble anyone. Using the reported wind speeds for games is obviously tricky. How do you determine what it must have been like on the field? Unfortunately temperature, where we have more accurate data, doesn’t have much effect on NFL games. Mostly it’s heavy snow and wind that makes life difficult on the offense.
Since the rule change, four teams have chosen to kick off to start overtime – all in games with high winds reported. The receiving team scored on the first possession in one of those games, and won two of the four. That’s far too low a sample size to make any conclusions, but since Bill Belichick has been the coach to make that choice in two of those four games (winning once), the reaction to this decision in the future will not be anywhere near as strong as it was when Mornhinweg tried it (and the rules made it much more of a losing proposition).
I scored games with a wind speed of 11 or more mph and a couple of other games with heavy snow as having difficult weather. That was 25.1% of all overtime games. About 4% of overtime games were marked as having extreme weather. What I found was that with no weather issues, teams scored on their first possession 36.4% of the time. With difficult weather, that dropped to 29.6% and with extreme weather, 15.4%. However, I found no significant weather difference in the percentage of times teams scored on the ensuing possession.
Overall, since the rules change, teams are 40-32-5 with the first possession in games with no weather marked and 12-14 in games with difficult weather. Of the nine overtime games with extreme weather since the rules change, the team with the ball first is 4-5.
I think in extreme cases, the decision to kick off in bad weather might help, but not much. And in difficult weather, at best the coin-toss advantage is removed. So my advice to coaches is that move should only be considered if you’re Belichick and you’re so far ahead of any living being in terms of understanding the game that this trivial analysis couldn’t possibly help or if conditions are absolutely brutal – which is something you probably have to determine from being out there on the field during that game.
This brings me to my last set of observations. Since I suggest that weather shouldn’t alter this decision unless your smaller players are having trouble remaining upright in the wind, is there any condition to examine? I didn’t want to data-mine the spreadsheet, because the results resemble astrology if you go in that direction. Obviously, if you look at enough samples, you’ll find something that passes a high significance test simply because of the nature of confidence intervals. Perhaps you should always kick off in Sunday night games in the NFC in odd-numbered years. Or always bet on horses with 12-letter names that start in position five. Data mining can bring endless entertainment in any sport.
I looked at one factor that might be a result of bad weather or, more importantly, why it might be worth kicking off in bad weather games. What is the score going into overtime?
Obviously tied. But if it’s 3-3, maybe the field position you potentially gain with a three-and-out on defense is more important than the advantage of going first. The average score going into overtime is 21.8 points per team. Are the odds of scoring on that first possession significantly worse in low-scoring games?
Unfortunately, scores are so tightly clustered at 17-all, 20-all, 23-all and 24-all (combined 55 out of the 103 overtime games since the rules change) that sample sizes are very tiny. All I’ve noted is that teams have scored on their first possession in only three of the 12 games that were tied at 13-all or lower headed into overtime. However, that low ratio didn’t carry over to the 1998-2011 period for low-scoring games. I can’t conclude anything based on the score itself. I’ve also noted that in games that are 30-all or higher since the rules change, teams have scored on their first possession in five of 18 games. Both high-scoring and low-scoring games aren’t favoring the first possession. In between, teams have scored on that first possession in 30 of 73 games.
And while the winning percentage of teams possessing the ball first (3-8-1) in games at 13-all or lower is unusually low since the rules change, that also doesn’t hold true for the 1998-2011 period (24-18-1).
All of this seems like low-sample-size gibberish to me, and perhaps looking at the weather is gibberish as well and you should simply take the ball first no matter what. Hard to say.