ST. PAUL, Minn., Oct. 24, 2017 /PRNewswire/ -- Las Vegas is notoriously good at stacking the odds against gamblers. However, a new study by the Computational Cognition Group (C2-g) demonstrates that gamblers who exploit oddsmaker decision-biases can now beat the house.
The study author, Dr. Erik Schlicht (C2-g Founder), used real-world gambling data to train and evaluate machine learning algorithms under different conditions that approximate sports betting.
The study evaluated models that either exploited or ignored decision-biases. These biases could result from the human and/or algorithm estimating the spread since both are known to periodically occur. The findings demonstrate that algorithms that exploited oddsmaker biases against the house, had after-the-spread (ATS) win rates greater (65-66%) than chance and models that do not factor biases into their predictions.
Dr. Schlicht stated, "The result is interesting for a couple of reasons. First, it demonstrated that oddmakers exhibit decision-biases that are known to adversely impact humans and algorithms in other contexts. Second, it demonstrated these decision-biases can be exploited by gamblers to improve their own wagering decisions and maximize profit. This is poetic-justice for gamblers, as it allows them the rare opportunity to turn the odds against the house."
The study author mentioned that findings were realized in the context of sports gambling, but the principle could be generalized to other competitive economic situations where humans and algorithms make forecasting decisions, such as stock selection.
Schlicht, E.J. (2017). Exploiting oddsmaker bias to improve the performance of NFL outcome prediction. arXiv: Statistical Applications.https://arxiv.org/abs/1710.06551
Dr. Erik Schlicht is the founder of the Computational Cognition Group (C2-g), a small data-driven consulting company located in Minnesota. His research utilizes quantitative methods to investigate human performance under uncertainty and risk. He leverages methods from machine learning and computational cognitive science to understand real-world decision-making. He has conducted research at Harvard, MIT Lincoln Lab, and Caltech, and received a Ph.D. in Cognitive and Brain Sciences from the University of Minnesota.