I downloaded ten seasons of Premier League football results and betting odds data (Bet365) from https://datahub.io/sports-data/english-premier-league (2009-2019 inclusive in ten CSV files).
After drawing some charts with the data just looking at the relationships between the betting odds and the results, one potential strategy that stood out was betting on matches to be drawn under certain circumstances – in particular when the odds of a draw are greater than 2 to 1, but the difference between the odds of the home or away team winning are small (<1).
Such matches stand out because the betting market is saying that a draw is not very likely (offering high odds of a draw), but the home and away teams have a relatively equal chance of winning the match…which should indicate a draw is likely because the teams are seen as being so evenly matched.
Below is a summary of some of my findings. At the top is detailed what would have happened if betting on all Premier League matches in the seasons from 2009-2019 to result in draws if the [decimal] odds of a draw were >=3.216 and the difference in the odds between the home and away team were <= 0.939.
Various different values of home and away odds difference (HA) and draw odds (D) were tried out on the training set of data; then I tested the models on the 288 matches played so far during this 2019-2020 season (match data again sourced from datahub.io).
All models would have made a profit and only one of the four ever dipped into losses before bouncing back into profitability a couple of months later.