Virtual Tennis: Serving Up A Simulator

Abstract

As with many facets of modern life, the game of tennis has become increasingly technology-driven at the highest levels. At the junior level there is more and more video analysis going on with players and their coaches, at the collegiate level there is often technology, such as Playsight, that can give statistics and design drills for players, and at the professional level they use all of these and much more, just to gain any slight edge over their opponents. Another technological tool that is growing in the tennis community is a mobile phone app called SwingVision, which gives ”real- time, automated, video analysis” for tennis players at all levels and will be written about more in Chapter 3. All these resources offer more potential to understand and analyze the sport, yet one of these technologies that has been relatively unexplored is simulation. My thesis will look to change this and explore a simulation as it pertains to the sport of tennis. Some of the more “marquee” sports in America, like football and basketball, have simulation technology that is used and that people can experiment with on the web, but in tennis there seems to be no similar sort of resource. Creating a simulator that can predict how a match will play out, gather statistics, and help players better understand their opponents’ tendencies will give plenty of interesting insights, especially as the model becomes more and more complex and realistic with added shots and shot probabilities. There will be plenty of applications as well as room for expansion in this endeavor so I hope to create something that has a lasting impact. Being able to test theories as to how a match could go differently if certain things were to change is clearly a very valuable tool for any player. This paper lays out the structure of the simulator and details the variables that go into running a simulation of two players with their input probabilities, as well as analyzing the effectiveness of several different decision-making processes and match- ups. Coded using Python, this simulator produces complete singles matches between two players, with each point being recorded shot-by-shot, and each shot categorized by depth, location, and speed. For any given incoming shot in a tennis match, a player realistically has only a few decisions that they should be considering, and these decisions can become even fewer depending on the player’s style or preferences. For the first model of this simulator and to not be predictable, each player is given between two and five different shots they are choosing between, with weights given to each of these shots based on my own knowledge of the sport and the likelihood of making the shot. We go on to examine several other decision-making processes and how changing a player’s strategy or shot probabilities can affect the results. Additionally, we examine a series of simulated matches between a virtual version of myself and my advisor, Professor Ahmadi.