Date: 2016 to present
Gambling is hard - the vast majority of traders will run to a long-term loss. Deep knowledge of a market might be sufficient for some to develop a profitable approach, but our view is that data, objective analysis and computerised implementation are the principle requirements. Removing/limiting human input in the betting process is generally best.
We've been heavily involved in most aspects of P2P betting. This covers the direct collection of very large amounts of market data, the restructuring, storage and access of these, and sourcing of auxillary information. We have built predictive models for odds estimation and market neutral algorithms for automated trading. We also maintain our own trading platform. In addition, we have a range of tools for the testing of algorithmic "edges", to guard against spurious profitability and accurate evaluation of risk. This is also key for sensible staking strategies.
Data is collected directly from P2P APIs or from web-scraped sources. A wide range of techniques are employed, covering: machine-learning and statistical predictive modelling; time-series modelling and more general probabilistic approaches. Simulation and computer-intensive inference are used extensively. All work in this area is bespoke and coded in-house.
Trading applications are bespoke and can be hosted on the cloud for low maintence 24-hour trading. Estimation of edges, profit and risk are from a mix of theoretical calculations, historical trading scenarios and extensive simulation.