What makes for an exciting game? Sensational plays, sure. Some lead changes, absolutely. But there’s also a feeling you get when watching one — that the unexpected just keeps happening.
But this is FiveThirtyEight, and intuitions won’t do! What if we found a way to quantify that feeling of surprise? What if we made the NBA Excitement Index?
Having long been a fan of Brian Burke’s win probability model and graphs at Advanced Football Analytics, I built a similar model for the NBA and debuted it on my site, Inpredictable, just over a year ago. The latest version of the model is based on detailed play-by-play data from 13 NBA seasons (from 2000 to 2012). This is not a simulation model; it is based on game results — for example, how often do teams win when trailing by two with 25 seconds left on the clock?1
By tracking swings in win probability, I was able to translate those swings into excitement. When visualized, games with a high Excitement Index are easy to spot. Their win probability graphs are choppy and turbulent.
Take, for example, the triple overtime thriller between the Memphis Grizzlies and the San Antonio Spurs on Dec. 17.
The Spurs fell behind early, trailing by as much as 23 in the second quarter before tying the game with just under nine minutes left in regulation. From there, the lead changed hands several times, and with just 15 seconds left in regulation, the Spurs appeared on the verge of completing the comeback. The Spurs’ Boris Diaw rebounded a missed Vince Carter 3-point attempt, giving San Antonio possession with a three-point lead.
Most games in that situation fizzle out rather undramatically (a combination of intentional fouls, free throws and time-outs as the trailing team attempts to stave off the inevitable). But this game featured nearly a whole game’s worth of drama squeezed into those final 15 seconds of regulation — a last-second turnover, a Mike Conley game-tying shot with seven seconds left on the clock; what appeared to be a game-winning shot from Danny Green with two seconds left, and a highly improbable Marc Gasol 30-footer to tie the game at the buzzer.
That’s not even to mention the three overtimes that followed.
The Spurs-Grizzlies win probability graph clearly has its share of dramatic turns, and its Excitement Index of 17.4 is the highest value of any NBA game this season.
San Antonio has had one of the most “exciting” seasons in the league thus far; a fact somewhat at odds with its no-nonsense public image. The table below ranks each team by the average excitement index of their games. The Spurs, with an average index of 6.45, trail only the Phoenix Suns for the most exciting season heading into the All-Star break.
New York Knicks fans should take note: This may be the only ranking this season in which your team finds itself adjacent to the Golden State Warriors.
| TEAM | WIN % | EXCITEMENT INDEX | |||
|---|---|---|---|---|---|
| 1 | PHX | 0.547 | 6.75 | ||
| 2 | SAS | 0.642 | 6.45 | ||
| 3 | TOR | 0.692 | 6.35 | ||
| 4 | CHI | 0.630 | 6.29 | ||
| 5 | LAL | 0.245 | 6.26 | ||
| 6 | MIL | 0.566 | 6.25 | ||
| 7 | MEM | 0.736 | 6.23 | ||
| 8 | IND | 0.389 | 6.22 | ||
| 9 | POR | 0.673 | 6.21 | ||
| 10 | BOS | 0.392 | 6.18 | ||
| 11 | OKC | 0.528 | 6.08 | ||
| 12 | SAC | 0.346 | 6.01 | ||
| 13 | HOU | 0.673 | 6.01 | ||
| 14 | MIN | 0.208 | 5.98 | ||
| 15 | NOP | 0.509 | 5.95 | ||
| 16 | DET | 0.389 | 5.94 | ||
| 17 | ORL | 0.304 | 5.92 | ||
| 18 | WAS | 0.611 | 5.89 | ||
| 19 | DAL | 0.655 | 5.81 | ||
| 20 | DEN | 0.377 | 5.80 | ||
| 21 | CHA | 0.412 | 5.73 | ||
| 22 | BKN | 0.404 | 5.69 | ||
| 23 | ATL | 0.796 | 5.61 | ||
| 24 | LAC | 0.648 | 5.61 | ||
| 25 | MIA | 0.431 | 5.57 | ||
| 26 | UTA | 0.358 | 5.55 | ||
| 27 | PHI | 0.226 | 5.48 | ||
| 28 | CLE | 0.600 | 5.46 | ||
| 29 | NYK | 0.192 | 5.45 | ||
| 30 | GSW | 0.820 | 5.35 |
I think most fans, if given a choice, would prefer to watch a game with a high Excitement Index (although I imagine Warriors fans are content with their “boring” season). But excitement is a fickle thing, hard to predict and nearly impossible to ensure. There is one metric that reliably correlates with excitement: the Vegas point spread. But that correlation is not particularly strong, and the result should not be too surprising: Games with a low point spread tend to have higher average excitement than games in which one team is heavily favored over the other.
Teams aren’t guaranteed to stay exciting as a season progresses. Using data from the past three NBA seasons, I looked at how a team’s average Excitement Index in its games before the All-Star break correlated with the Excitement Index of games played after it. Some correlation did persist, but the results largely showed a regression to the mean.2
NBA teams can’t avoid boredom for long.
CORRECTION (Feb. 18, 11:20 a.m.): In a previous version of the table accompanying this article, several NBA teams were in the incorrect order.