Wednesday, August 08, 2018

Binary Spread Record: Over-Performing

One of my favorite topics to discuss on this blog in the betting line. The betting line is a mix of bookmaker ‘power ratings’ adjusted for venue and market corrections based on bets placed that combine to give even the casual fan a decent idea of which team will win a particular football game. But what happens if the betting line continues to make incorrect calls about a certain team? And by incorrect, I mean not just in reference to covering the spread, but actually winning the game. If a team pulls off a lot of upsets or is upset a lot, what tends to happen the next season? Being a resident of Columbia, South Carolina, I have a particular curiosity about teams like this.

The hometown Gamecocks, a team I am rather ambivalent about, are getting a decent amount of preseason love by the media and especially the hometown radio station (107.5 The Game). The Game is not quite the propaganda machine of Fox News, but their takes are generally pretty rosy even when the team is struggling. Vegas is more lukewarm toward the Gamecocks’ prospects with an over/under win total of 7.5. Using their surprising performance last season as a frame of reference, what should we realistically expect from the Gamecocks in 2018? To begin answering that question, let’s review South Carolina’s schedule from 2017. Below I have included their entire 2017 regular season schedule with the venue, opponent, spread, and game result (straight up, not against the spread). The final column (Net) rewards the Gamecocks when their game result differs positively from the spread and penalizes them when their game result differs negatively. Major thanks to Phil Steele for providing spread data in his annual magazine.
Using the spread as a binary (win or loss) and not probabilistic predictor, the Gamecocks ‘should’ have won five games last season. They won eight. They pulled four outright upsets (NC State, Missouri, Arkansas, and Tennessee) and lost once as a favorite (Kentucky) giving them eight total wins and a net differential of +3.

I looked at every team from the previous three seasons (2014-2016) that exceeded their binary spread record by at least three games. I then looked at how much their win total changed the next season. The results are listed below.
Eighteen teams met the +3 criteria and two thirds of them (twelve) saw their win total decline the following season with the average team declining by about 1.4 wins. What might be driving these declines? For starters, these teams were probably fortunate to win as many games as they did in what I will dub their ‘upset’ season. If a team wins games as an underdog, but continues to be an underdog in their upcoming games, ‘luck’ could be the driving force in their good record. We’ll define luck or good fortune as being on the good side of more random events like turnovers, huge special teams plays, and general clutch play in tight games. Maybe an analyst would describe it as ‘finding a way to win’. You get the idea.

Before we look at all teams that exceeded their binary spread record by at least three games in 2017, I wanted to look at a team that exceeded their preseason expectations significantly, but did not do so by pulling off a lot of upsets.

Fresno State finished 1-11 in 2016 with their lone win coming against an FCS team. Despite hiring an established coach in Jeff Tedford, expectations were muted in 2017. Their preseason win total of 4 reflected a small expected bounce, but not much else. However, the Bulldogs surprised nearly everyone by winning nine regular season games and playing in the Mountain West Championship Game. Let’s take a look at how their schedule played out.
After an exhibition against Incarnate Word, Fresno was cannon fodder for a pair of elite Power Five programs. Relative to the spread, they performed admirably in those two games and found themselves favored in their first three conference games. After alternating an upset of San Diego State with a head-scratching loss to UNLV, Fresno was chalk until they upset Boise State in the regular season finale. While the Plexiglas Principle certainly to applies to teams that make a seismic leap like Fresno State, they may be more disposed to hold onto their gains than a team like South Carolina since most of their wins did not come at the expense of ‘superior’ teams.

As promised, here is a listing of every team from 2017 that exceeded their binary spread record by at least three games. This doesn’t guarantee regression, but keep this in the back of your mind if you plan on backing these teams this season.

No comments: