Let's say we want to make a more powerful Enemy object; it will have the same behavior a regular Enemy object will have but with more health. One possible solution would be to derive a new class PowerEnemy from Enemy and give it double health. In this solution the new class will seem extremely sparse; it will use the functionality from Enemy but with one different value. An easier solution would be to have a way to set the health of an Enemy class, whether through an accessor or in the constructor. Inheritance isn't needed at all.

When deriving a class, it is a good idea for it to model the "is a" relationship. This means that the derived class should also be of the same type as the parent class. For example, deriving a Player2 class from Player would fit the model, as Player2 "is a" Player. But let's say, for example, we have a Jetpack class and we derive Player from this class to give it access to all the functionality that a Jetpack class has. This would not model the "is a" relationship, as a Player class is not a Jetpack class. It makes a lot more sense to say a Player class has a Jetpack class, and therefore, a Player class should have a member variable of type Jetpack with no inheritance; this is known as containment.

On platforms such as PC and Mac, the performance penalties of using inheritance and virtual functions are negligible. However, if you are developing for less powerful devices such as handheld consoles, phones, or embedded systems, this is something that you should take into account. If your core loop involves calling a virtual function many times per second, the performance penalties can add up.