So far have been looking at the goals scored/conceded from
as input
to a Poisson model to estimate the results?

Here we get bet365 odds from to feed into rpart, a machine learning
package in r to predict results?

The idea here is that the market takes into account all opinions
and so should do better than other models? If I had a magic
model (assume there are well resources groups out there
who do) then I would keep betting wherever the odds offered
value until the odds were the right ones? There is of
course risk since players can get injured, referees make wrong
calls and so on.

This model looks at all the games played in the premier league
so far and fits the best pattern it can! So it should be pretty
good, though we haven’t tested it yet which we’ll do in our next
video?

The main thing is that we are using the wisdom of the market
which probably knows more than say my Poisson model?

Vídeo Completo