Regression to the Mean

Analysis
+ posts

The 2017 season win total predictions have come out and the Lions have been given an over and under of 8 wins through most betting establishments and national writers. Many use the term “regression to the mean” and mention Detroit’s wins in close games to somehow discount them for 2017 and justify less losses despite an improved “on paper” playoff team. This enraged me and noted further evaluation. Let’s take a look at the actual numbers for NFC North teams in games decided by one score last year.

Minnesota
The Vikings were 2-4 in one score games in an 8-8 season. They have room for improvement and proved in 2015 they can win double digit games. The Vikings either won handily in 6 games in 2016 or got beat pretty badly in their other four losses.

Minnesota improved their offensive line (Reiff, Remmers, Efflein) and RB situation (Murray, Cook) but still have questions for 2017 (weapons, QB, offensive line despite additions, new defensive holes). We will see if the 2015 Vikings or the 2016 Vikings are more accurate for 2017. Sam Bradford has a big year to prove things and earn a long term deal. The Vikings defense was still good in 2016 but it couldn’t get Minnesota into the playoffs alone. We’ll see if the Vikings defense stays as strong and if Minnesota’s offense can take a much needed step forward.

Chicago
The Bears were 1-6 in close games with a 3-13 overall record. They were blown out in 7 games and had two other decent size wins (Min by 10 points, San Fran by 209 points). The one close win was home against Detroit by 3 points.
The Bears had major QB issues and secondary issues. They had some injury issues as well (QB, McPhee, and Kyle Fuller) but Chicago just lacked the talent to win close games or hang with seven opponents. The Bears should be closer this year and took some gambles at QB (Glennon and Trubisky). We will see if the Bears secondary additions (Fuller health, Prince Amukamara, Marcus Cooper and Quentin Demps) can get them back to a competitive team. Chicago split with Detroit and the Vikings last year and is capable of splitting with Green Bay. The Bears have a dark horse chance to make some noise in 2017 and win division games but they still may be a year away while their QBs develop.

Detroit
The Lions won 8 games by 8 or fewer points. Some of those wins were late come backs and some games the Lions led and had the game in hand while staving off a come-back (Bears, Jaguars, Rams). The Colts game Detroit had a big lead and the Colts took the lead late only to have the Lions take it back away. The Redskins and Eagles game, the Lions led most of the game only to have the Eagles and Redskins take a late lead before Detroit had clutch fourth quarter play by Stafford and Slay. The Lions came back twice on the Vikings but both games were close and down to the wire like most divisional games.

The problem with the experts and using “regression to the mean” is that Detroit did lose many close games by one score (5) already. The national writers seem to take into account the Lions close wins but ignore the close losses. Winning 8 out of 13 one score games is a pretty normal winning percentage (61 % winning percentage).

Detroit went 8-5 in close games and beat the Saints handily in the Super Dome. Detroit lost on the road to the Giants and Cowboys while very injured (Stafford, Slay, Levy far from 100%, Swanson, Riddick, and Abdullah). Detroit lost those two tough games by double digits but the Lions were far from 100% in the toughest games of the year (at NYG, at Dal, at Seattle).

The five close losses show Detroit can not only win more games (as they get players healthy and added more talent) but also can have a much better win loss percentage in close games than 61 percent. They could go 9-4 or 10-3 next season or win more games by double digits and have less close games than 13. The regression to the mean doesn’t fly as a reason Detroit will lose more close games because they already lost five close games in 2016 and had a normal winning percentage in the 13 close games.

Regression to the mean can go the other way too. Detroit’s defensive DVOA and run game can regress to the mean and be solid instead of absolutely poor. Lions have made plenty of moves to make those improvements happen. Acting like Detroit will continue to struggle in pass rush, coverage and run game while now losing close games (despite winning 8 of them despite all those things) is just being naïve. The national writers only use regression to the mean for close game wins and ignore regression to the mean for the Lions weak areas from 2016. It is lazy journalism at its finest.

Also, why is it a given that Detroit will lose more close games? Stafford is an excellent late game quarterback and they have the talent to make big plays on offense and defense. Detroit earned every single win and led many of those close games for most of the contest. In the NFL, there will be close games and the Lions are a team that knows how to win them. Anyone who says the Lions lucked into the playoffs or several wins should be shot and not allowed to watch football for eternity.

Green Bay
The Packers were 6-3 (66 % winning percentage) in the regular season and 7-3 (70 winning percentage) including the playoffs in one score games. Green Bay had a better win percentage in one score games than Detroit. The Packers did win four games by double digits and one playoff game by double digits (at home). If you are going to use the regression to the mean argument, then Green Bay is the one team who should be losing more games due to regression of the mean than Detroit due to a higher winning percentage in close games. Also, Green Bay was blown out 3 times (at Redskins, at Titans, Cowboys) compared to only twice for Detroit (while injured at Giants and at Cowboys). Green Bay was also blown out by a team they had a close loss to in the first game (at Atlanta in Champ game).

Regression to the mean should be Detroit and Green Bay splitting not a Green Bay sweep. Regression to the mean should show Rodgers not having an MVP type performance in 2017 (see 2015 season). Regression to the mean is missing the playoffs and not making it 10 years in a row. If you want to talk about regression to the mean, let’s start right there.