The evaluation of a game state depends on structural features of the current state. So the first task is to decide which features are of value. The next step is to provide each feature with a range of possible values. The last step is to devise a set of weights in order to combine all the feature values into a single value.
For example, in chess there are two useful features: material strength and positional advantage. Suppose both range over the real interval [0,10]. Then we evaluate the current state by first finding the values for these two features and combining them. If positional advantage is relatively twice as important as material strength then the formula might be:
eval(Current State)=1*eval(material strength) + 2*eval(positional advantage)
This would yield an evaluation function taking values over the real interval [0,30].