Real-time forecasting within soccer matches through a Bayesian lens published in Journal of the Royal Statistical Society Series A: Statistics in Society, Volume 187, Issue 2, April 2024. (with Soudeep Deb & Rishideep Roy) (ArXiv) (Journal)
Abstract :
This paper employs a Bayesian methodology to predict the results of soccer matches in real-time. Using sequential data of various events throughout the match, we utilise a multinomial probit regression in a novel framework to estimate the time-varying impact of covariates and to forecast the outcome. English Premier League data from eight seasons are used to evaluate the efficacy of our method. Different evaluation metrics establish that the proposed model outperforms potential competitors inspired by existing statistical or machine learning algorithms. Additionally, we apply robustness checks to demonstrate the model’s accuracy across various scenarios.
Presentations:
August 2024: Joint Statistical Meetings, Portland, USA (Accepted)
December 2023: 3rd All India Research Conference, IIM Lucknow, India
November 2023: Indian Institute of Management Bangalore (IIMB) Decision Sciences Area Brown Bag Seminar
Dynamic Stake Allocation for Live Tennis Betting: A Finite-Horizon Markov Decision Process Approach (with Rishideep Roy & Soudeep Deb)
Abstract :
In this paper, we consider in-play betting in tennis, where the hierarchical scoring system induces a large but finite state space with frequent, well-defined transitions. Match progression is modelled as a time-inhomogeneous Markov process and embedded in a finite-horizon Markov Decision Process. The state combines the current score with the market participant's remaining budget and outcome-contingent exposures, while the objective is to maximise expected CRRA utility of terminal wealth. Transition probabilities are estimated from observed match dynamics, allowing the optimal policy to respond jointly to the evolving match state, existing financial exposure, and risk preferences. Results show that the dynamic policy achieves favourable expected terminal-wealth performance while substantially reducing drawdowns relative to benchmark betting strategies.
Presentations:
June 2025: 11th MathSport International Conference, Luxembourg
June 2025: 34th European Conference on Operational Research, Leeds, UK (Accepted)
August 2025: Joint Statistical Meetings, Nashville, USA (Accepted)
December 2025: 2025 IMS International Conference on Statistics and Data Science (ICSDS), Seville, Spain
Model-based Kelly Allocation in In-Play Soccer Markets (with Rishideep Roy)
Abstract :
This article considers in-game betting in soccer, formulating the problem as one of Kelly allocation under uncertainty. In contrast to ideal betting models that assume frictionless or perfectly fair odds, the analysis treats in-game odds as exogenous, time-varying quantities that may embed vigs, asymmetries and other market frictions. Betting decisions are made sequentially over the course of a match, subject to evolving budget constraints and uncertain returns.