When I was playing my DOS NBA video game I certainly did not think I would be writing this following post. However, the evolution of the video games – and in particular of sports video games – has been impressive to say the least. Play the latest NFL Madden or FIFA and you will get a sense of realism that you might actually forget you are just playing a video game. The simulation of game strategies is also impressive and here is where things are getting interesting.
Simulation games have been used for years to help train at a low cost and in complex settings pilots – both civilian and military. The military has also used video games to train soldiers and treat PTSD when they return from their deployment. Even surgeons are trained through surgical simulators in minimally invasive techniques. If surgeons, pilots and the military are using video game technology (and not for entertainment only :)), why can’t the sports community do the same? We have started seeing computing technology being employed towards training (e.g., robotic tackling dummies in American football training, QBs training through virtual reality technology during the offseason etc.), however, these efforts are more of a look to the future and certainly far from being mainstream (e.g., virtual reality is still trying to take off as a mainstream technology).
Can we use current video game technology for facilitating sports analytics tasks?
While there are many ways to look into this question I am focusing on two points of view; (i) the leagues’ point of view, as well as, (ii) the sport’s analyst/researcher point of view.
Video games as the leagues’ tool for designing new rules
Many times the leagues are changing the rules of the game in order to achieve a specific objective. For example, this year the NFL changed the touchback rule in an effort to minimize the hard hits from the collisions between the kick returner and the defense. The rationale is that by adding 5 more yards to the touchback the returner will decide to take the touchback, thus, eliminating the chances of serious injuries. However, how is the league sure that this will not lead to kickers trying to pin the ball inside the 5 or 10 yard line and thus, forcing the returner to still return the ball? The short answer is that it doesn’t know. Having though a video game that implements this new rule can allow the league to observe how people react on this and what they are doing. Video games might not be able to realistically simulate hard hits but it certainly can reveal what people will decide to do – i.e., take the touchback or try to return the ball, kick the ball as normal or try to pin it within the 5 yard line. Of course, who is playing the video game is also important. While it is possible to learn from the casual user, observing a trained in football person playing will reveal more. Similarly, for basketball one can simulate what will happen if the three-point line is moved further away (or even adding a four-point line). One of the issues with this however can be simulating the field goal success rate. Current studies can form a blueprint for the FG% as seen for example from the following figure:
However, it might be hard to simulate the real-life adoption of the players to the distance, which might lead to a different function between FG% and distance. Nevertheless, important data can be collected by observing how people react to it. In general, I firmly believe that video games is an elegant, quick and cost-effective way to obtain a first reaction to changes in the rules.
Video games as the analysts’ tool
Video games can allow us simulate many more “rare” scenarios that appear in a game. For example, in NFL the sample of two-point conversion attempts is too small and censored towards situations where teams trail in the score in the end of the game. Similarly, fourth-down conversion attempts are biased towards “manageable” situations and this can impact the analysis of the best decision in fourth-down situations. What if through video-games we could obtain a larger dataset of similar situations? Never punt or attempt a field goal. How does this impact the game? Never take the kick PAT but attempt the 2PT conversion. How does this impact the result? (with the significance of a single data point, my Madden experience says that 2PT conversion should be the default PAT strategy and kick PAT situational football). Clock management strategies can also be evaluated. Should an NFL team take its time-outs 4 minutes before the end of the game or let the opponent burn some time? Should you take a 5-yard penalty for delay of game in the beginning of the third quarter or burn a timeout that you might need later? Similar clock management issues appear in basketball too. With 40 seconds left on the clock for the period, what is better, go for a quick shot in order to get two total possessions or go for one well-developed possession? These are situations that either do not appear very often in a real game or more frequently there is not enough experimentation from the coaching stuff in order to draw conclusions. Again video games will not provide a definite answer, but they can provide additional evidence and guidelines for the analysts.
The bottom line is that video games can be of great help in a variety of sports analytics tasks! After all we have already seen crowdsourcing being used in real-life sports analytics tasks, with the Draft 3.0 from the Sacramento Kings!