Shoot Like Ronaldo: Predict Soccer Penalty Outcome with Wearables

A penalty shot is a crucial opportunity of scoring a goal in soccer and often becomes a game-winning factor. The goalkeeper has to guess the ball flight trajectory within a fraction of a second and typically jumps to either side of the goalpost taking (often mistakenly) cues from the kicker. These cues are postulated in goalkeeper's mind based on the kicker's posture, run-up, and angle of attack at the ball. Statistical analysis of historical data helps pundits to identify beneficial strategies in such a competitive environment but these tactical knowledge is of marginal help to a novice soccer player for improving their penalty kickout and blocking skills. Oftentimes, players are not equally skilled at successfully placing the ball around different sections of the goalpost. To empower players with a retrospective view of their performance and identifying the strong versus weak shots, and blocking capabilities, it's crucial to detect the direction, force, and trajectory of the shots. In this work, we propose a wearable sensor-based approach to detect the outcome of various goal shots from the kicker's dominant foot movement profile. We empirically assign six hot-zones inside the goal post and collect data on a real-life penalty shoot-out using economically available accelerometer sensors from four participants. We develop a deep learning approach for the shot classification and we report superior (53%) accuracy over traditional approaches (47%) in a challenging setting of recognizing different goal shots from the segmented data stream.