A statistical theory of optimal decision-making in sports betting
The bettor’s problem is quantile estimation: the median outcome decides the side, and further quantiles decide which games to bet at all.
I study how noisy systems — brains and markets — turn information into decisions. I’m an Associate Professor at CCNY and lead machine learning at Optios, a neurotechnology company.
Three threads run through my work: the statistics of sports betting and prediction markets, where I publish on what’s actually true about beating the market; machine learning for neurotechnology, decoding human state from biosignals in real time; and theoretical work using stochastic thermodynamics and causal analysis to find structure in neural and financial data.
The bettor’s problem is quantile estimation: the median outcome decides the side, and further quantiles decide which games to bet at all.
A sportsbook’s expected profit is an exact affine function of the share of money on the losing side. Advanced to the full-paper round of the 2027 MIT Sloan Sports Analytics Conference Research Papers Competition.
Every forecast is posted before the game and scored afterward, and the full record is reported. Judged on calibration and proper scoring rules, not on a good week.
A research experiment in market efficiency, not betting advice.
Code and data behind both betting papers.