Ouch. After a promising run in testing, the Poisson model managed the incredible feat of getting every Premier League prediction wrong. How could this have happened? The only possible explanation is match fixing orchestrated by Dan Tan. Surely it's not these models.
In the other leagues we saw small losses across the board except for some profits in League Two, where both models tested horribly. On the whole, I'm optimistic that they will net a profit in the long term as they seem to be spitting out fairly reasonable probabilities for each game.
After the action Wednesday I will produce a new set of ratings and run simulations for the remainder of each season in order to predict the battles for the Champions League spots, relegation, and promotion. In the meantime, here are today and tomorrow's fixtures for the lower leagues now that odds have been posted.
This format in Google docs will probably work better than the fusion tables I was previously using. Both models like the value for Millwall to beat Sheffield Wednesday.