Video games of professional sports teams are quite popular and are frequently updated to reflect the current team rosters and player skill levels. The players in the video game, let’s call them avatars, have the names of the real players, supposedly have their skills and play their positions. We were interested to determine how much of this is true, i.e. whether the play of the real player and the play of the avatar on a statistical level are the same.
To make the comparison we analyzed the first five real games of the Phoenix Suns in the 2015-2016 season and compared them to games run autonomously in the “NBA 2k16” video game between the same teams .
These games were between October 28 and November 2, where the Suns played the Dallas Mavericks, the Portland Trailblazers (twice), the Los Angeles Clippers, and the Sacramento Kings. The lineup of avatars and for the real games was identical. For all shots, in both the video game and the real game, we recorded the shooter and several other interesting facts associated with the shot. For instance, we recorded the number of passes prior to a shot; the time on the shot clock when the shot was taken; the time the shooter held the ball before shooting; the region where the shooter received the ball and where he took the shot; and, most importantly the result of the shot (successful or unsuccessful).
We find that the “NBA 2k16” video game in many ways is very good at representing the real game. This indicates that the developers of the game must have utilized strong statistical analysis of real NBA games to model their gameplay.
Two of the statistically significant results are: Shots taken in the first eight seconds of the shot clock were more successful in both the real games and the “NBA 2k16” games.
Holding the ball longer increased the probability of success in both the real games and the “NBA 2k16” games. This finding in particular is very interesting since it contradicts the basketball literature: Usually, when the ball is held longer, the likelihood of a forced shot, which has a lower probability of success, increases. Apparently the Phoenix Suns have figured out a way to score on long ball possessions. The interesting fact for our analysis is that the video game knows this.
A category where the video game misses, but not by much, is the shot allocation within the team, i.e. the percentage of all shots of the team in the complete game taken by the specific players. Here “NBA 2k16” did not correctly model the fraction of the team’s shots for each player, however, the differences were found only in bench players. We suspect that the statistics for the starting line-up is well documented and has been used by the game developers but the bench players either have little play time, are new to the team, or are new to the league and hence their behavior is not as well documented.
In another categories, the video game did not do as well: The literature on NBA games suggests that more passes per possession decreases the likelihood of a successful shot, a fact apparently known to the game developers since our video games showed the same trend. However, this trend was not present in the real games of the Phoenix Suns.
Figure 1: Hot and cold zones. The success probability of a shot is <30% in the blue region and >50% in the red.
Beyond the timing of the shots for avatars and real players we were interested to see whether real players and their avatars have particular positions on the court from which they are shooting much better than from others. We define a hot zone as a region where the probability of a successful shot is > 50%. Conversely a cold zone is a region where the player has a less than 30% success rate.
Using the data from the five real Phoenix Suns games and their video replications, the shot chart in Figure 1 were created. As two examples we show in the two top charts of Figure 1 the hot/cold zones for the starting point guard, Eric Bledsoe and his avatar. In the bottom two charts we show them for the starting center, Tyson Chandler and his avatar.
Although the agreement between the shot charts is not perfect, they illustrate a similar trend. In real Suns’ games, Eric Bledsoe shot well in two point regions and poorly in the three point regions. His avatar shoots extremely well from two-point range and poorly from the corner three point shots. The main difference occurs at the three point shot at the top of the key where Bledsoe shot poorly in real Suns’ games but his avatar is doing great. Eric Bledsoe was the only starter, where statistically, the allocation of shots across the seven regions as well as the success probabilities across the regions in “NBA 2k16” differed from the real games.
This is illustrated by the very good match for the hot/cold zones of Tyson Chandler. As a center, he has the highest probability of success close to the basket. The game developers clearly know this and Chandler’s avatar did not shoot in any other region besides the one closest to the basket. The real Tyson Chandler in general behaves as expected and shoots well close to the basket, but he also took one shot and missed from the top of the key accounting for the blue region.
Overall, the results of this paper show that the video game development team knows the literature on NBA basketball statistics. In addition, these results suggest that the developers specifically modeled the Phoenix Suns and typically missed only for new or bench players. It will be interesting to extend this research to determine whether the video games also correctly describe team play, using more sophisticated network based metrics .
In general, these video games allow almost realistic simulations of a game. If they improve a bit more, we may find that coaches and Monday morning quarterbacks may use video game simulations to study the effect of different lineups and player trades on specific games. It remains to be seen whether the real games are unpredictable enough to remain fun to watch.
 Harrington, J. (2016) Are NBA Video Games representing the real game? A statistical comparison of Phoenix Suns’ Shooting Patterns and their Video Game Counterpart. Undergraduate Honors Thesis, Arizona State University
 Chang, Y. H., Maheswaran, R., Su, J., Kwok, S., Levy, T., Wexler, A., & Squire, K. (2014). Quantifying Shot Quality in the NBA. In Proc. 8th Annual MIT Sloan Sports Analytics Conference (pp. 1-8).
 Fewell JH, Armbruster D, Ingraham J, Petersen A, Waters JS (2012) Basketball Teams as Strategic Networks. PLoS ONE 7(11): e47445. doi:10.1371/journal.pone.0047445.