Madness: Upsets Are Predictable

Here at Sharp Sports Analytics we focus on college football. Yet, there is no doubting that the NCAA basketball tournament is the best tournament in sports. 64 teams (I don’t count the play-in teams), 63 games (I don’t count the play-in games), 3 weekends, and absolute madness. One of the selling points of this tournament is that anything can happen and that any team can win. The results, so it goes, are completely unpredictable. While the tournament is truly spectacular, the unpredictable claim is just false.

Picking the Right Dog for the Fight

2018 TournamentSharp DogsSquare Dogs
Avg Spread11.57.5
Against the Spread12-6-03-6-0
Win Count51

We’re 48 games into this tournament and there have been 10 upsets*. Of those upset only 1 team, Marshall, was a public team in Vegas. By a public team I mean a team that is receiving 60 percent or more of the bets in a game. Of the 41 games where an upset was possible*, 9 times the underdog was the public team in Vegas. The lone win, Marshall over Wichita State, leaves the public dogs with an 11 percent upset percentage, not so hot.

The sharp dogs have performed much better. By sharp I mean the side that has 40 percent or less of the bets on it. UMBC over Virginia, Buffalo over Arizona, Loyola-Chicago over Tennessee, Florida State over Xavier, and Texas A&M over UNC were all sharp upsets. There were a total of 18 sharps dogs and those dogs came out on top 5 times. This equates to an upset percentage of 28 percent, much higher than that of square dogs.

In any analysis such as this, sample size is a huge issue, so we must dig further to validate the claim that upsets are predictable. Fortunately, there is plenty of evidence to support my claim. Sharp dogs not only won at a higher rate than square dogs, but they were much larger underdogs. The average sharp dog was an 11.5 point dog while the average square dog was a 7.5 point dog. This means that based on the spread alone, we’d expect square dogs to win at a higher rate than sharp dogs.

Additionally, we can look at against the spread performance to see if sharp dogs are beating expectations more often than square dogs. The data speaks for itself here. Sharp dogs are 12-6 against the spread in this tournament while square dogs are 3-6 against the spread. Once again, we run into the issue of sample size, so I collected data over the last 9 tournaments to see how sharp dogs have performed. Since 2010, sharp dogs are 145-110-5 against the spread. This equates to a cover percentage of 57 percent.

While the NCAA tournament is certainly madness, the sharp eye recognizes that the sharp dogs are the ones who pull off the big upsets.

* I define an upset as a non-neighboring lower seed who is a point spread under dog that wins. A 9 over an 8 is not an upset. A 5 over a 4 is not an upset. When a 10 seed is a two point favorite over a 7 seed it’s not an upset. Thus the upset count is 10 instead of 15.

Follow Jordan on Twitter @JDSports84

About Jordan Sundheim 154 Articles
Jordan Sundheim is a veteran of the sports analytics field with a background in both athletics and statistics. He earned his master's at Duke, coached football at Stanford, and is currently a sports analytics consultant.