This is another biggie: a very, very wrong and bad TDC in an otherwise good game. I'll let a lady named Jessica explain it: "This damaged the ending of Shadow of the Colossus for me -- and it happens in other games too. You allow the player to control their character during a sequence, but no matter what the player does, the sequence can only go one way. Since it's not clear that it would ever happen, when it does happen, it makes you want to try the sequence again, but that only gives you the same result."
"If it can only go one way, make it a cutscene. If the player has control of the character, let the player's actions make a difference, and affect the outcome. If it's a part of the game's 'style' to let the player 'play' through what are essentially cut scenes, then make it that way throughout the game so that we know the game is going to be this way, and don't just surprise us with it at the end."
John Funderburk adds, "I also hated Tomb Raider Legend's and Resident Evil 4's 'interactive' cut scenes. 'Push the button when I tell you to' -- what game is that?" People play games in order to overcome challenges, make interesting choices, and generally express themselves. Game sequences that don't provide any of those experiences shouldn't be interactive. We expect that when we have control of the avatar, the avatar's actions will affect the game world in some way. If it affects the game world in no way at all, then there's no point in pretending that it's interactive.
A word of caution, though -- this is not an argument against linear stories in games. With a linear story, overcoming challenges earns the player more story (usually in the form of a cutscene), even though the player can't change its content. That's OK -- the very act of overcoming the challenge unlocks the next phase of the story, and the player knows and understands this.
Capcom's Resident Evil 4
The problem arises when we lead the player to believe her actions do matter, and then it turns out that they don't, but the player wastes hours and hours trying. If you want to tell a tragic story -- the doomed hero or the hopeless cause -- you must not lie to the player and tell him that he can escape his fate if he just tries hard enough.
For tragedy to really work, the audience must know in advance that the hero is doomed, or at least come to realize it without spending fruitless hours trying to avoid it. We can still make games about Napoleon, or the Americans in the Vietnam war, even though the player knows the ultimate outcome will be failure.
This is a classic mistake and once again, what's most surprising about it is that people persist in making it. Jacek Wesolowski writes, "One factor that harms my entertainment is that some developers treat mouse and keyboard as secondary setup. The difference between those and gamepads is significant, because usage patterns differ."
"For instance, keyboard is better suited for 'broad' interfaces, assigning a key to each action, whereas gamepads rely on the 'deep' variety, in this case -- button combinations and sequences. Simply mapping buttons onto keys, or vice versa, is often insufficient. But that is exactly what many developers do."
"A good example of this is Assassin's Creed. Its controls make a fairly good sense when playing with gamepad, but the keyboard/mouse mapping is unwieldy and counter-intuitive. Even worse, the developer has imposed an artificial, and very severe limit on mouse sensitivity, probably to match it with the maximum turn rate available with gamepad. There is no gameplay reason for this, because instant 180-degrees turns are available anyway, as well as looking behind avatar's back. In other words, higher mouse sensitivity would not give me any real advantage, other than being able to play comfortably. While playing, I felt as if my preferred control device was sabotaged deliberately."
Jacek has put his finger on one of the reasons I'm a PC gamer rather than a console gamer: I'm not coordinated enough to manage combos, and I prefer to have separate buttons that each do one thing (or better yet, a smart button that does what I mean). However, my preference doesn't make it a Twinkie Denial Condition. The bad mouse/joystick adaptation is one, though.
Mouse-based interfaces work poorly on joysticks, and usually joystick-based interfaces work poorly with the mouse too. They don't do the same thing. A mouse is a pointing device. A joystick is a steering device.
A mouse doesn't automatically return to center the way a joystick does, and a joystick can't move indefinitely in the same direction the way a mouse can. If you're going to make a system for both, don't privilege one over the other or kludge one to fit the other. Design the user interface for each separately as if it were the only input device you will be supporting, and make each as good as it can be.
If you discover that this gives the joystick player a big advantage over the mouse player, or vice versa, don't solve the problem by sabotaging one player's control system! Build in a handicapping system that the players can manipulate themselves and mutually agree upon. It works for golf; I see no reason why it shouldn't work for video games.
Alternatively, under the principle if you can't do it well, don't do it at all, drop support for the device that you can't implement properly. That's better than selling the player an inferior experience.