A Trading Agent Framework Using Plain Strategies & Machine Learning

• Autor(a) principal: João Pedro Araújo Santos
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://repositorio-aberto.up.pt/handle/10216/76151
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

The world of online sports betting exchange (trading) is growing every day and with that people are trying to improve their trading by using automated trading. In analogy to the financial markets the buy and sell operations are replaced by betting for and against (Back and Lay).This thesis describes a framework to be used to develop automated trading agents at Betfair sports markets using a Java programming interface. Betfair processes more than five million transactions (such as placing a bet) every day which is more than all European stock exchanges combined. Betfair is available 24 hours a day 7 days a week. For this thesis were developed two trading agents, DealerAgent and HorseLayAgent, accordingly with the presented framework. The agents mentioned above act on To Win horse racing markets in United Kingdom. They use plain strategies together with machine learning methods to improve the profit/loss results. The developed agents were submitted to viability tests using data from Betfair To Win horse racing markets from January, February and March of 2014.