My 2 cents on the second article:
The Dixon Coles model is, despite being the most widely known, flawed:
The defensive and attacking ability of teams are not independent so you have to define them simultaneously, which is done in the first article I've attached that uses the good stuff of the DC model (the modification of the pure Poisson framework to take into account the low scoring possibilities underestimated by Maher's model) but goes deeper. I'm still working on back testing it.
The second one tries to go back to more simplicity by using a regression based approach. I've only read it once but it seems pretty awesome...
Attachment 41560
Attachment 41558