O uso de Auto ML nas apostas esportivas

Abstract

The sports betting market has been growing effectively, and of the 40 teams that compete in the A and B series of the Brazilian Championship, 35 are sponsored by bookmakers. Observing the number of gamblers that are profitable, it was found that there is a great need for solutions that help these people to become profitable. In order to achieve this goal in the long term, the first step of this project is to evaluate the use of Auto Machine Learning (AutoML) solutions in two betting scenarios: both teams score, and the match has more or less than 2 goals. For this work, data from the 2017 Serie A Brazilian Football Championship were used. The championship has 38 rounds in total and 10 matches in each round, totaling 380 matches. To calculate the profit scenarios, two forms of evaluation were established: without restriction and with probability restriction. Comparing the results obtained in these two scenarios, scenario 2 in all results presented a profit, unlike what was achieved in scenario 1, in addition, we have the assertiveness rate that was much more positive in scenario 2, reaching a value of 73.68% while the maximum rate obtained in scenario 1 was 63.88%.