Who Am I?
I'm a research scientist at NVIDIA where I study applications of artificial intelligence and machine learning in esports. To put it succinctly, I investigate how to use data and machine learning to better understand human performance in competitive environments. You can see my profile on NVIDIA Research for more.
Research Areas
My primary research interest is investigating human performance in competitive video games. I seek to answer problems that have demonstrated need in the esports and artificial intelligence communities.
The underpinning of my work is that I use large amounts of data. This means anything from spatiotemporal player tracking and event data to videos, text and even audio. I focus on making interpretable and performant models.
In general, I'm interested in human performance in competitive environments, and that includes conventional sports. I've worked across most major sports, including baseball, basketball, American football, ice hockey, and soccer.
I'm a big fan of open-sourcing as much as I can, and I try to conduct my research with the broader community in mind. To that end, I like to develop code, tools, and data that other researchers might find helpful. Check out examples of my tools and code.
Publications
CHI PLAY (2022). Xenopoulos, P., Rulff, J., & Silva, C.
arXiv preprint (2022). Xenopoulos, P., & Silva, C.
WWW (2022). Xenopoulos, P., Freeman, W. R., & Silva, C.
AI for Sports Analytics Workshop (IJCAI, 2021). Xenopoulos, P., Coelho, B., & Silva, C.
AI for Sports Analytics Workshop (IJCAI, 2021). Petri, G., Stanley, M. H., Hon, A. B., Dong, A., Xenopoulos, P., & Silva, C.
Big Data (2020). Xenopoulos, P., Doraiswamy, H., & Silva, C.
PMLR (2022). Xenopoulos, P., Chan, G., Doraiswamy, H., Nonato, L. G., Barr, B., & Silva, C. Spotlight Talk
TVCG (2022). Xenopoulos, P., Rulff, J., Nonato, L. G., Barr, B., & Silva, C.
Big Data (2021). Xenopoulos, P., & Silva, C.
arXiv preprint (2022). Hogan-Hennessy, S., Xenopoulos, P., & Silva, C. (2022).
Tools & Data
pip install awpy
Awpy is a Python library to parse, analyze and visualize Counter-Strike 2 demofiles. Since its release, Awpy has garnered over 20,000 downloads, and has an active Discord community. The library is well-documented and works on Google Colab.
git clone https://github.com/pnxenopoulos/esta
The ESTA dataset contains about 1.5k parsed CSGO demofiles using awpy. ESTA contains detailed spatiotemporal data on player actions (damages, kills, grenade throws, etc.) and trajectories. In total, ESTA is about 4 GB of compressed JSONs and contains 8.6m million actions and 7.9m game frames.