Don't Change All the Rules at the End
In some games, the player makes his or her way through the game, getting progressively better at it. He's finally having fun beating the baddies when he hits the end portion and suddenly is confronted with having to learn a new set of rules. Half-Life, for all its ground-breaking innovations and great gameplay, did this: You fight alien creatures and soldiers on earthly soil. But near the end, you find yourself jumping around in a floaty alien space with portals in it. In Undying, you end the game by fighting on an island with newly acquired weapons against a hydra-headed monster that spits fire.
Max Payne, on the other hand, avoids this problem. The final level is grand, breathtaking, with new enemies and set-pieces for action. But the final level still has its feet planted on the ground. There are no new alien creatures, no new weapons, and no new skills to learn. Nor is there any new puzzle-solving or platform jumping. You just do what you've been doing all along, but do it real good.
Obviously, what's both critical to achieve, and difficult to achieve, is balance. You want enough twists and surprises at the end so that it's not simply "more of the same," and rewards the player for having made it through that far. On the other hand, you don't want to give the player the feeling that he or she is starting all over again, with new weapons and new rules.
I always enjoy filmmakers who break established rules and get away with it. Pulp Fiction is a case in point, breaking just about every established rule on plot structure. The same thing applies to these last two points, about the importance giving players cool weapons or cool things to do early on in a game, and about not switching the rules at the end. Some game surely will come along that does both of these, and that is utterly amazing, enjoyable, and fulfilling to play.
In my screenwriting classes, therefore, I tell my students I'm offering tools, not rules, and the same goes for the techniques in this book.