Welcome to Pythia
Data Sciences
About
Pythia develops proprietary predictive models of sporting events for the purpose of helping our clients achieve staggering returns on wagering. Our coverage of sports and locations has grown considerably since our inception, and currently we predict results globally for a range of competitions. We make use of everything from classical statistics, to modern machine learning techniques, and work mostly in Python and MATLAB, but also R, VB, and C++, on a mix of on-premises and cloud-based machines.
New Approaches
Keeping up with new developments in sports analytics and machine learning are vital to what Pythia does. We are constantly on the lookout for new methods or approaches that we can use, from abstract ranking systems to metabolic power models, we draw upon a large range of advanced techniques to drive our predictions.
NEW DATA
New or unique data sources are very important to maintaining our predictive advantage. Therefore, testing and evaluating new data sources makes up a good part of what we do. In addition to technical analysis of the data, we have access to a range of subjective sports experts who help with fundamental analysis to allow us to get the most from the new information.
MODEL MAINTENANCE
With a significant number of models deployed, containing a large amount of sub-models, keeping on top of model retraining, hyperparameter tuning, and dealing with data or rule changes as they occur is an important task. From performance reporting to tweaking sub-models, we are always monitoring and improving our live models as needed.