What makes an NBA game fun to watch? (Hint: It’s not the basketball) | by Daston Arman | Jan, 2021
Important: Close Games
The one thing has been subpar about this season so far has been the large number of blowouts. As a perfect example, there was a lot of talk about how boring the Christmas games were because every game was lopsided.
The data backs this up. All models universally point to close games being BY FAR the most important aspect of a fun game. A simple plot of Margin-of-victory (MOV) vs. popularity makes this very clear.
Not all close games are fun, but pretty much all blowouts are boring.
Of the >70 variables I tested, only 6 actually were significant, with 4 of those being related to how close the game was in the end: MOV, whether the game went into overtime, 4th qtr lead changes, and the largest 4th qtr comeback.
My Random Forests model gave MOV a 71% feature importance while the next most important variable only had a feature importance of 9% (woah!). As well, MOV was the 1st variable selected in my stepwise model, giving by far the most improvement in R².
Using my LASSO model, I estimate the difference between a 20pt blowout and a 1pt overtime game to be at least ~2.5 popularity points (on a 10pt popularity scale).
This is one big reason the 2019–20 Bucks had so few fun games. They were just too good in the regular season, winning too many games in boring blowouts. While we know superteams can draw big ratings, it’s important that another team can at least challenge them or else the game will be a dud.
Important: Big games from All Stars
The only other two significant variables in my model were (1) the number of All Stars playing and (2) the number of 30pt scorers in the game. These are both fairly simple measures of star power and individual scoring.
My model estimates the difference between a game with 0 and 4 All Stars is ~0.9 popularity points, all else equal. Similarly, the difference between a game with 0 and 4 thirty point scorers is ~1.1 popularity points.
Not Important: Basketball
In this analysis, I made a big effort to look at a ton of variables about games to try and tease out what was fun to watch. To my surprise, no variable about the actual game of basketball was even close to being significant.
Pace, total PTS, passing, fouls, points in the paint, three pointers, etc.… I did not find evidence that any of these affect how much people enjoy a game.
As a big basketball fan, my prior opinion was that maybe people liked high scoring games with very little fouls, but nope! Bummer for me because I thought this analysis would have a ton more nuance and interesting findings!
Read More …