How Good Are We At Predicting Blowouts?
With All-Star weekend approaching, we are more than halfway through the 2016-17 NBA season. This means we have a healthy dose of data at our disposal. One topic worth exploring is that of blowouts, so I thought it’d be interesting to explore the question: “How good are we at predicting blowouts?”
If you’re like me, you rely on the Vegas spreads in your daily research. Vegas is significantly smarter than I ever will be at predicting scores and spreads, so why not leverage their knowledge? One thing I focus on is whether a game has a large spread, which signals a possible blowout. This might be a red flag for me because if a game blows out early, the thinking is the starters on both sides will get pulled early and not reach value. For this article, I’m going to explore how well Vegas has been this NBA season at predicting blowouts, and how often the favored team has covered the spread in a predicted blowout.
Special thanks to @sethayates for helping me pull the data for this article. Just some housekeeping things to take care. First off, I’m going to focus on the 2016-17 NBA season, specifically the games that have been played between October 25, 2016 (the start of the season) through January 31, 2017. Some of you may argue we need a large sample size to draw more concrete conclusions. Trust me, I’m all for larger sample sizes, but by focusing on this season’s data, we can work with current results using current rosters. Secondly, I am going to define a “predicted blowout” as any game with a Vegas spread of 10 or more points. I realize this is somewhat arbitrary but there is no official definition of “blowout” I’m aware of and I have to pick a cutoff, so double-digits feels like the most reasonable approach. Alright, let’s dive in!
The Set-Up
Using the data from this season, I first identified every game with a Vegas spread of 10 or more points. Through January 31, 2017, there have been 123 games that fit this category of “expected blowouts.” Just to help visually, here’s how they have been distributed between October 2016 through January 2017 (keep in mind the season started on October 25, 2016, which is why October has a smaller value):
One fun fact, of the 123 expected blowout games, the Warriors account for 35 of them. The Spurs are next with 15 games, followed by the Cavs at 14 games, the Clippers at 12 games, and the Rockets at 11 games. Another fun fact is that only 18 different teams have been expected to blow out an opponent so far, meaning 12 teams have yet to be favored by 10 or more points this season.
How Often Teams Cover The Spread
Now we get into the juicy stuff. From the chart above, we saw how many games each month had a double-digit spread. So what percentage of those games actually covered the spread? This is where I was surprised. If we look at those 123 games that were expected to blowout, only 60 of them (i.e. 49%) actually covered their spread. It shocked me that only half of these expected blowouts covered the spread. I came in with an assumption that if Vegas predicted a blowout, the team would just blow out their competition without question, but that hasn’t always been the case this year. If you’re wondering, the Warriors have covered the spread only 16 out of the 35 times they were double-digit favorites, which is 46%. Although the crazy thing is when they have covered the spread, they’ve crushed their opponents by an average of 13 points more than the spread. Case and point – they beat the Lakers on Nov 23 149-106; they beat the Blazers on Dec 17 135-90, and most recently, they crushed the souls of the Clippers on Jan 28 144-98.
How About Just Focusing On Blowouts?
I decided to look at the data from another angle. Rather than taking those 123 expected blowout games and finding out how many covered the spread, I decided to see how many of those games ended up with the favored team winning by 10 or more points. In other words, let’s not worry about covering the spread. If Team X was favored by 14 points but only won by 13 points, they would not have covered the spread but you could make the argument they still blew out the other team based on our definition of blowouts being a victory of 10 or more points.
Below is a chart that looks at the data from this perspective. The numbers went up as you would expect given we’ve loosened our inputs, but they didn’t rise by that much. What the chart is saying is that among those 123 expected blowout games between October 2016 – January 2017, 68 of them (i.e. 55%) ended in an actual blowout with the favored team winning by 10 or more points. To view it from the other side, in 45% of the games where we’re expecting Team Y to get blown out, Team Y either kept the game close within single-digits or actually won the game.
So there are a lot of obvious follow-up questions: Why aren’t more teams that are expected to blow out their opponent not doing so? How do these numbers compare to historic Vegas trends? Do star players on either team still reach value in blowouts? Who benefits most in blowouts?
I’m going to give you an honest answer – I don’t know yet and will need more time digging into the data to uncover them. For goodness sakes, to give you a sense of how pathetic I am, I’m writing this at midnight while eating chicken nuggets dunked in Chick-Fil-A sauce that I grabbed from their store #dadlife #dadbod. But hopefully in time, I’ll be able to do more articles that look more into blowouts and how we can better approach them based on Vegas spreads. As for now, the takeaway for me was that maybe we aren’t as good at predicting blowouts as we think we are.
Now one last point. I know some people will argue that my definition of 10 points as a “blowout” is too strict. I did some Google searches on what people consider a blowout in the NBA and it varied all over the place. I toyed around with taking those 123 games with a double-digit spread and seeing how many ended up with the favored team winning by 20 or more points, in case that’s your definition of a blowout. Using those inputs, I found about 30% of those games ended up with the favored team winning by 20 points. So in the end, we’re looking at something around one-third to one-half of these expected blowouts actually blowing out, depending on your definition.
You can reach me on Twitter. You can also leave me a Comment below if you have thoughts on these findings, thoughts on how Vegas spreads influence your DFS approach, or what you consider an NBA blowout. I do read every Comment and appreciate you taking the time to read this.