Nylon Calculus: Would shorter NBA playoffs be a recipe for upsets?

The league may have to consider shortening playoff series to squeeze in the 2019-20 NBA playoffs. Would that be a recipe for chaos?

Since the NBA season was suspended on the evening of Wednesday, March 11, fans have wondered if the 2019-20 season would come back, and if it does what form it would take. There has been much talk in recent days about to the feasibility of finishing the season in a “bubble”, a sealed-off area where play could theoretically resume safely. Great reporting has been done on what would need to happen for this to work — including an ESPN report by Tim Bontemps and Brian Windhorst.

While there are many changes that the NBA might need to implement to pull off this idea, a version that fascinated me was reported in late March by Jabari Young of CNBC, among others.  Young reported that the NBA was not only exploring the possibility of holding its playoffs in empty stadiums in a centralized location — Las Vegas was discussed at the time and now Walt Disney World in Orlando is also being reported as an option — but also considering holding shorter postseason series. These series would potentially be best-of-five (or even shorter), rather than the standard best-of-seven.

While we do not know if the 2020 postseason will ultimately take this form (or any form at all), this got me thinking about how such a format could affect the playoffs.  Specifically, how many more upsets might we expect under these conditions — no homecourt advantage and each series of shortened length?

In order to determine how a new format would alter a playoff series, I first built a general model for an NBA playoff game, using data for the 2003-2018 playoffs from Basketball-Reference. For a really simple model, we can use just two variables: which team is playing at home and the difference in regular-season point differentials between the two teams (home-team point-differential minus road-team). Using these two variables, we can build a simple linear regression model for the outcome of any one playoff game, and in turn, get the probability of each team winning the game.

What do we see from this model? Well, as one might expect, the difference in regular-season point differential between the two teams is a highly significant predictor of the final margin in a playoff game. In fact, each point of point differential difference is worth about one point in predicted game outcome. Taking an example from this year: the Toronto Raptors, who posted a +6.5 point differential, might be expected to beat the Brooklyn Nets, who posted a -0.6 point differential, by an average of 7.1 points on a neutral court.

A second interesting nugget we get from this model is that homecourt advantage in a playoff game was worth about 4.4 points on average from 2003 to 2018.  This is a good bit higher than the 2.2-point average margin of victory for the home team in the 2019-20 regular season. It makes sense intuitively that the typical homecourt advantage in a playoff game, with increased fan enthusiasm, would be greater than this advantage in a regular-season game. An increased NBA homecourt advantage in the postseason is consistent with what others have found too, including Neil Pane of FiveThirtyEight in 2017.

Of course, we are leaving a lot of stuff out with this simple model.  Regular-season point differential is not always the best proxy for playoff strength.  Ask the 2015-2018 LeBron James– led Cleveland Cavaliers, who notoriously had a second gear for the postseason.  But this simple playoff game model does give us a place to start in analyzing potential differences between a typical best-of-seven series and a shorter series on a neutral court.

How much likelier are upsets?

Using our simple playoff game model, we can simulate a playoff series between two teams of arbitrary regular season strengths under different conditions. To do this, we can treat individual games as independent and then run 100,000 simulations of a playoff series. The chart below shows the results:

The x-axis is the difference in regular-season point differential between the team with homecourt advantage in the series and the team without it. Even if we are in the neutral court scenario, we can still look at the graph from the perspective of the team that would have had homecourt advantage in the series. The y-axis gives the probability that the team with homecourt advantage (or which would have had homecourt advantage) will win the series.

So using our example from earlier, say the +6.5 point differential Raptors play the -0.6 Nets.  We look to 7.1 on the x-axis (6.5 minus -0.6). In the usual best-of-seven games series with the Raptors having home court, we would look to the red line and find that Toronto has about a 91 percent chance of winning the series. But if we play the same series as a best-of-five on a neutral court in Las Vegas, we instead look to the green line and find that Toronto’s chance of winning the series falls to 86 percent. In a best-of-three on a neutral court, the Raptors become just 81 percent favorites.

Is this a big deal? Well, a 5 or 10 percent change in win probability is certainly not nothing, especially when squads are searching for any meaningful edge in a playoff series. So the underdog Nets would, of course, prefer a shorter series on a neutral court. But the difference between 91 percent and 86 percent would be hard to feel over just one playoff series. We would see five more upsets over the course of 100 theoretical playoff series between the Nets and Raptors, but in real life, we only get to witness just the one actual series.

Interestingly and perhaps intuitively, we see that large underdogs would be the biggest beneficiaries of having a three-game series rather than a  five-game contest. The fewer games the better for a heavy underdog who needs a few breaks to pull off the upset against a stronger foe.

On the other side of the spectrum, when the teams are of equal ability, the length of the series is of less importance. The neutral court is of course still a welcome change for the team which would have been without homecourt advantage in the traditional series. I estimate here that if two teams of equal ability play each other in the regular format, then the team with homecourt has a 55 percent chance to win the series.  So the neutral court costs the team with the better record about 5 percentage points of win probability in this scenario.

What if the playoffs were single elimination?

Ok, now what if we get really crazy?  Suppose the NBA institutes a single-elimination tournament, similar to the NCAA’s March Madness postseason competition. How many upsets could we see then?

Looking at the gap between the red line, our typical best-of-seven, and the black line, single-elimination, we see some real upset potential.  The Brooklyn Nets, who had just a 9 percent chance of winning our hypothetical, standard series against the Raptors, now have roughly a 28 percent chance of pulling off their own One Shining Moment upset.  Of course, the effects of this change would be most pronounced for large underdogs like the Nets. If the teams are more evenly matched to begin with, the single-elimination format is less of a big deal.

What we find here is that a best-of-five or a best-of-three on a neutral court would increase the chance of an upset, but not dramatically so. Heavy underdogs gain 5 or 10 percentage points in increased win probability. A single-elimination tournament would be another order of magnitude of craziness, with large underdogs gaining roughly 20 percentage points in increased win probability.

Of course, the one guaranteed thing about a playoff series played on a neutral court would be a less exciting viewing experience. And we would all be losers because of it. But seeing any playoff basketball at this point, if it can be done safely, would certainly be a joy to us all.

Nylon Calculus Rookie Review: What did the Heat see from Kendrick Nunn?