top of page

Projects

Identification of Prominent Researchers in the AI research community using Co‑Authorship Networks

Co-Authorship Networks are social networks that map the collaborative relationships between researchers and scientists. By using methods of network analysis, I have tried to uncover the influence of a particular researcher in the AI research community. A detailed report can be found here

Measuring the Tactical Creativity of Football Players using Exploratory Breadth 

My primary project during the internship with Dr. Memmert at the German Sport University in Cologne. The fundamental idea is that by treating the a sports team as Complex Adaptive Systems (CAS), the dynamic change in spatial and functional roles of the agents (players) can be computed by modeling the properties of the system as a whole. Dynamic overlap, a measure of statistical entropy was utilized to represent the tactical creativity of the team in terms of the different states of the system 'explored' by the team, hence the term 'Exploratory Breadth'.

Effect of warm-up decrement on the Free Throw Accuracy of elite NBA atheletes

Another project that I was fortunate enough to contribute towards during my internship in Cologne. In this study, we tried to determine the psycho-physiological factors affecting the free-throw accuracy of NBA players. The concept of ‘calibration effect’ implies that human neurological systems require to ‘calibrate’ themselves while performing repetitive actions thus resulting in the fact that the probability of making an error is higher in the first of many attempts while repeating a sensorimotor task. Free throws in basketball presented themselves as an ideal avenue to examine the calibration effect, with the particular case of two or three free throws being taken in succession. We analyzed the free-throws data from ten NBA seasons and statistical tests yielded a significant result, indicating the presence of the calibration effect. I co-authored the manuscript with the results of this study, which was accepted and published by the German Journal of Exercise and Sports Research

Expected Goals (xG) Model for the Indian Super League

While working with SportsKPI, a sports analytics startup I led the creation of a Machine Learning based pipeline which could help better understand the quality of chances created in the game of association football.

 

By tracking the variables involved in a shot in football, such as the distance & angle to the goal, number of defenders blocking the shot, etc., I was able to create a logistic regression model which could predict the probability of a shot getting converted into a goal. I was responsible for the creation of this pipeline in its entirety starting from data collection using annotation software such as Longomatch, cleaning and then modeling of this data to generate actionable insights.

bottom of page