Neural Network: Extra features used for training but not for predicting new data

Question

I want to do a neural network to predict who is going to win a soccer game. I have several features (like the team, the physical shape of the team, etc.) and an output, telling which team won (or if there was a die).

In my training, I want to add features like the number of goals these teams scored. The problem is that I won't be able to include these features when predicting the final result of a future game. Is there a way to do it ?

I'm using Keras in Python as a library to easily build neural networks.

Answer 1

The number of goals would be an output, not an input. For example, you could build a model to predict how many goals each team would score...

If you used goals scored as an input, the model train to be incredibly simple and useless- just comparing the goals, ignoring all other inputs.