The Poisson model, as a variation of the original Dixon & Coles model, was the first model I implemented and used to predict the outcome of football matches. I submitted picks based on this model during Feburary 2017 and May 2018.
Following pictures shows the profit line:
The model showed just a very small profit of 14.4 units during this time. Based on the backtest of the model I expected a yield between 6% and 7%. But as the productive usage showed, the model suffers some disadvantages, which indicate that the model is not able to beat the bookie in long term.
In following blog posts I analysed the submitted picks and tried to look behind the results and spot the weaknesses of the Poisson model:
- Bundesliga match days 1 – 10 (2017/18)
- Bundesliga match days 11 – 20 (2017/18)
- Bundesliga match days 21 – 30 (2017/18)
- Retrospective for Bundesliga season 2017/18
If you have further questions, feel free to leave a comment or contact me @Mo_Nbg.