≡ Menu

Is Week 1 Really An Outlier?

Week 1 always seem to have some really rare results. Last year, the 2-14 49ers won 28-0 in the late MNF game on opening weekend. The Redskins finished last year with a winning record, but lost at home by 22 points to the Steelers, easily Washington’s worst performance of the season. And the Falcons won the NFC last year, but you wouldn’t have known that watching week 1: Atlanta lost at home to Tampa Bay.

In 2015, the Titans blew out the Bucs, 42-14, in the season opener; Tampa Bay finished the year 6-10, while Tennessee went 3-13. And the 49ers, who wound up going 5-11, again were week 1 superstars: in 2015, San Francisco shocked the 11-5 Vikings, 20-3, on Monday Night Football.

Many of these characters were part of the shocking week 1 results in 2014, too. That year, the 49ers beat the Cowboys in Dallas, 28-17: Dallas finished tied with the best record in the league, while San Francisco went 8-8. The Titans, as they did in 2015, were week 1 superstars in 2014: despite going 2-14, Tennessee beat the 9-7 Chiefs, 26-10, on opening day. And the Vikings and Rams show up here, too: in 2014, Minnesota won in St. Louis, 34-6, in week 1; both teams went 6-9 the rest of the year. Oh, and Miami upset the Patriots in week 1; New England won the Super Bowl, while Miami missed the playoffs.

So is week 1 really an outlier? Well, for the 49ers it obviously has been. The last three years, based on expected results (using location-adjusted SRS ratings to predict final scores), San Francisco has exceeded expectation by over 20 points in each of its last three week 1 games. But there are also teams like the Jaguars. In week 1, 2016, Jacksonville lost at home to Green Bay by 4 points, and Jacksonville finished about 8 points behind the Packers in the SRS. In 2015, the Jaguars lost by 11 at home to the Panthers in week 1, and finished about 15.5 points worse than Carolina. In 2014, Jacksonville lost by 17 in Philadelphia in week 1, and finished 2014 a little over 14 points worse than the Eagles in the SRS. In other words, Jacksonville’s week 1 performance came within 2 points of expectation — based on the full season results — in each of the last three years.

So instead of using anecdotes, let’s look at the full set of data. I generated full season SRS ratings and expected results for all games since 2002. Then, I calculated the extent by which each game varied from those expectations.1 For example, the biggest outlier was this game between the Dolphins and Titans in 2012. In 2012, the Dolphins had an SRS rating of -2.6, while Tennessee had an SRS rating of -10.0. So Miami was 7.4 points better, which means in a home game, the Dolphins should be expected to win by 10.4 points. Instead, Tennessee won by 34 points, meaning the game varied by 44.4 points from expectation.

That was a week 10 game, though, so it doesn’t really stick out in our memory. The Titans/Chiefs game from 2014 (expected result: Kansas City +20.5, actual result: Tennessee +16, difference: 36.5) was the biggest week 1 outlier in the sample, with the 2003 Bills shutout of the Patriots (expected: NE -4.9, actual: Buffalo +31, difference: 35.9) being the second biggest week 1 outlier.

I calculated the difference from each game in each week since 2002. The graph below shows the average results:

The average game differed from expectation by 9.4 points, with week 1 being at 9.1 points. In other words, week 1 appears to be perfectly average, while the final games of the year are outliers. This is consistent with previous results using a different methodology, so that makes me feel more confident in our results.

  1. Since the ratings represent the results from the entire season, this can only be calculated at the end of the year. These are not results based on SRS ratings at the time the game was played. []
  • Adam

    Interesting study. I think our perception of week 1 being wacky is due to the primacy effect – we remember the first event in a sequence more vividly than the others. Thus, week 1 upsets tend to stick in our memories better than the same upset happening in week 9.

    • AgronomyBrad

      That, and there’s probably more viewership and coverage of opening weekend, and talking heads saying how “Team A shocked Team B in a massive upset based on our preseason rankings.” As opposes to something happening mid season when a lot of casual fans are tuning out and there’s other stuff to hold our attention (world series, election, etc)

    • I think week one also seems wacky because it occurs before we have a chance to put the entire season into context. Games we think of as upsets sometimes end up not being upsets at all, but we view them as such in real time based on our expectations of how the two teams should perform.

      • Richie

        Yeah, doing the same study but using point spread instead of SRS may show week 1 as more of an outlier. I feel like maybe Chase has already done that before. (Since point spread is more representative of perceived strength than actual strength.)

        • As a person who tries to think of original topics to research and write about, I can say that it feels an awful lot like Chase has already covered every topic there is.

  • Dr__P

    Week 1 means all veteran’s salaries are guaranteed, so teams wait until after that game to sign additional FA talent

  • Tim Truemper

    I think Adam has the best explanation (my psychologist bias showing here). The primacy effect is pretty powerful, well-supported under other circumstances, and is the most parsimonious.

  • John

    After reading the headline I instantly thought of the 2003 Bills-Patriots Week 1 game. Otherwise known as the “Lawyer Milloy” game.

  • Pingback: New York Jets Flight Connections 07/09/17 – New York Jets News()

  • Pingback: Weekly Sports Analytics News Roundup - July 11th, 2017 - StatSheetStuffer()