clock menu more-arrow no yes

Filed under:

The StatHawk, Week 1: Regression is not always a naughty word

New, comments

Sometimes it is downright encouraging.

Steven Bisig-USA TODAY Sports

If the term "regression" sounds negative to you, you're not the only one. It's often used as an antonym of "progression."

Examples!

"Colin Kaepernick regressed as a passer in 2014."

"The typical espn.com commenter shows unmistakable signs of regression to age 11."

But statistically speaking, regression has a different, far more neutral connotation. When we talk about "regression to the mean," what that means is outliers will likely get drowned over time by the ocean of more typical events to follow. Think of it this way: if a quarterback throws into double coverage ten times, and gives up nine interceptions, that's bad for him, but it's also unusual. After 20 throws into double coverage you would expect him to give up fewer than 18 interceptions.

It works just as well in reverse: if a QB (let's say 2013 Nick Foles, for no reason whatsoever) throws into double coverage ten times and nets eight touchdowns, that's good, but it's also unusual. After 20 throws into double coverage you would expect him to have fewer than 16 TDs.

That's regression. Isn't that right, Richard?

(Statisticians, feel free to give a better definition below, with extra points awarded if you cite nerdy terms like "independent variables" and "linear" and "the intercept.")

Short story long: if a player or a team has something weird happen in a year, chances are that weird thing will unweird a little over time. Because regression. It pulls numbers back to the mean. Commentators will often say a player or a team is "due" for good luck or bad luck. They're wrong when they say that: while the gambler's fallacy is very attractive at first glance, it is still a fallacy. But the Phil Simmses of the broadcasting world are also right, if you give them enough credit -- they're probably just trying to express regression to the mean, in their own imperfect way.

Lecture complete. What Hawk stats from 2014 are ripe for this kind of welcome regression? I thought you'd never ask.

Regression Candidate 1: Takeaways

The mean, in this case, will be set at the average amount of takeaways logged by Carroll's Seahawks, from 2011-2014. Why not 2010? Because that defense was 25th by points and 27th by yards, whereas its successors have all been top 10 (or top 1) in those categories since.

Chart time, yes?

2011 2012 2013 2014
Forced Fumbles 13 18 24 19
Fumbles Recovered 9 13 11 10
Interceptions 22 18 28 13
Total Takeaways 31 31 39 23
Turnover Margin +8 +13 +20 +9

Interesting. So our Hawks' average numbers for the past four seasons are:

Forced Fumbles: 18

Fumbles Recovered: 11

Interceptions: 18

Takeaways: 31

Well, the 2014 numbers fall far short. It would be statistical silliness to expect an uptick in the fumble departments this coming season; however, last year looks like an aberration for interceptions specifically and for ball-hawking in general.

If the 2015 numbers revert to something close to 18 picks and 31 total takeaways, that's a significant upgrade, probably worth more than one win on the season.

We'll take it, thanks.

Regression Candidate 2: Opponent's FG performance

Across the National Football League of Football, kickers made 84 percent of their kicks last year and 61 percent of those over 50 yards.

Against the Hawks, they upped their game.

NFL Kickers

2014 season

NFL Kickers

2014 vs. Hawks

Overall FG percentage 84.0 87.5
Percentage >50 yards 61.0 60.0
Percentage 19-48 yards Pretty High 100 percent

Some random numbers plucked from across the league, and from this fascinating article at fivethirtyeight.com:

Kickers from 44 yards: 86 percent

Kickers from 48 yards: 71 percent

Kickers from 47 yards: 55 percent

Kickers from 40 yards: 88 percent

But then, from our table above --

Kickers from 48 yards or less against Seattle: 100 percent

It'd be reasonable this season for opposing kickers to regress back toward the 84 percent league-wide number this season, and to miss occasionally from inside 48 yards. That's the kind of thing that, again, could easily swing a close game back in the Hawks' favor.

Regression Candidate 3: Punt returns

The Seattle Seahawks did not have a banner year returning punts, or covering them last season. You watched the games. You know how bad it got. The eye test is still a thing. But here are four telling examples, quantified:

Even if the punt return unit had not been massively upgraded with the addition of Tyler Lockett, you'd expect some regression back to the mean from those extreme performances. In 2011-2013, the story told had been drastically different:

  • Out of all qualified returners, Hawks were 8th, 20th and 8th in return average (11.5, 8.7, 11.3);
  • Their shortest "long return" in those three years was 37 yards;
  • They never finished with negative scores in both FO categories at once;
  • They covered punts better than most teams, placing 11th, 31st and 21st in yards given up.

A couple well-timed punt returns -- or heaven forbid, a special teams score -- just might, you guessed it, be good for another win somewhere along the way.

Come on regression. Get to work. Tomorrow's good for us.