If you are interested in how Football in Cider works out its predictions then have a read of this document:
Although Football in Cider does not use this library, as it was written back in 1990, it uses exactly the same technology.
Libraries like this have come on in such a way that many companies are now using this technology in their own production environments without their customers even knowing. For example, your supermarket loyalty card gathers so much information it would be impossible for humans to learn your behaviour and respond in a way to maximise their profit margin. Instead, neural networks parse all the data and learn how a different promotion affects the bottom line. After the learning process the supermarket then targets their customers in whatever way has the most positive impact on profits. This could mean reducing emails and increasing billboard advertising for example because the AI machine learnt that email promotions affect sales in the medium term. These AI neural networks discover patterns and create rules that human may never be able to understand.
From our experience here at Football in Cider we actually struggle to explain why our predictions have come about. Looking inside of the neural network is extremely complex. The data is extraordinarily large and the decisions paths can twist even the most talented of brains. Also, each game ultimately creates its own instance of a neural network. So what is true for the goose isn’t necessarily true for the gander.