Accounting For Two Teams In Training Data Set For my ML course, I am planning on building a soccer match predictor using ANN.
Some of the input variables I have chosen are:


*

*Form in last 5 games

*Current Table Position

*Number of Injuries

*Club Ranking
My question is, how can a training set account for two teams, if each team has a different value for each variable. 
Any input is much appreciated. 
 A: 
My question is, how can a training set account for two teams, if each team has a different value for each variable.

You could give as input to the neural network the feature values for both teams.
A: Machine learning can use almost anything as an input, from the color of the team's Jersey to the number of people in the crowd, or if it was home or away.
So, the team name could be an input.
Alternatively, just create separate models for each team's performance. 
Incidentally, it often doesn't make sense to think too hard about the predictive variables when using machine learning approaches. Sometimes it is unexpected variables that machine learning shows up as having the mist impact. 
To explore this, consider taking a population of 20+ variables and developing the n combinations of 5-7 variables that you could have. Then run your neural network or other machine learning tool for each group of variables. Choose a metric (e.g. RMSE?) for each model, and see which group of variables does best. I'll bet you're surprised which are truly important. 
