[Hey, you can read this post at Gamasutra and GameSetWatch. Radical! Thanks to Simon Carless et al. for republishing it.]
Pretend this is the final post.
In video games, the ones that tell the player a long, linear story, the ending is usually an uncertain proposition. Prose and film teach an audience to expect three-act structures and considered pacing in storytelling. Instead, games have what Warren Spector calls the second-act problem; where act one is the intro movie, act three is the outro movie, and in between is the game. Games are structured less like a novel and more like an anthology; an arbitrary number of assembled vignettes, thematically united in post-production. A collection of missions and quests that exist because one designer had a cool idea for a boat chase sequence and another designer had an awesome idea for a stealth mission. It's a problem of pacing, and it relates directly to the presupposed need for games to have fifteen-hour narratives.
I think this issue is compounded by another: players don't know how long a game is. You can hold a novel in your hands and feel the weight of the pages. An album has its track listing printed on the back. A television season consists of a predetermined number of episodes with those episodes at a fixed length. A movie is somewhere between 90 and 180 minutes. No such guidelines with video games. They lack an intuitive metric: it'll fall between one and one hundred hours.
If players don't know when to expect the real ending then they'll have to guess. Maybe after this mission in GTA we'll get to the endgame. Wait, no, one more thing. One more thing after that. With these interminable games that try for an engrossing narrative, players just get tired. Will it ever actually end?
Fallout is based on the premise that the player must find this water chip. It takes a long time, it's an exhausting journey, you find it and return home victorious. And then... one more thing... and you're actually only halfway through.
Objectively, there's nothing wrong with the content. But expectations frame experience, and the game had just prepared the player to say goodbye, not to enjoy another ten hours. Having to take a game at its word, players feel betrayed and jerked around. We react to a piece of content differently if we know it's the ending. When we watch the season finale of a TV show, we know that this time the characters are really in danger. With a video game the player has no idea. Is this thing going to go on for another hour? Or five? Or ten? Where the hell am I in this story? I'm not sure many developers are aware that this can be a problem; like how Ken Levine has said he didn't anticipate the ugly comedown from the stratospheric highs of BioShock's Andrew Ryan scene.
Expectations are everything. The movie Gone, Baby, Gone has a fake ending at about the 70-minute mark, but the audience doesn't start leaving the theater. They know how long a movie is and they're mentally prepared for the remainder of the film. I don't think Fallout players would be as bummed out if they found the water chip at the 70-minute mark. But no one knows how long Fallout is, like how no one knows if Return of the King's running time is three hours and two minutes or three hours and four minutes. The movie continues long past the point where anyone was interested.
One more thing. One more mission, one more quest, one more rung in a ladder carved from monotony and you have only the vaguest of assurances that the ladder ever stops. I wonder why people don't finish games.
Oblivion's core story is paced terribly, which is to say it's paced like a video game. One more thing. One more lost object to find. That's at least consistent with Oblivion's general M.O. as a treasure-hunting smorgasbord, and Mass Effect doesn't handle that dichotomy nearly so well; instead redefining 'sidequest' as a repetitive grind existing at the periphery of the story. BioWare dumps a whole lot of extra content on the player for the purposes of making Mass Effect long enough to count as a conventional video game. It dilutes the tightly focused, very linear narrative that they're trying to showcase. It's also why games like GTA that measure game completion with a percentage stat don't really work, since it can take players five times as long to get from 76% to 77% as it can from 1% to 2%.
Subquests aside, Mass Effect is able to manage player expectations of length. After act one, you get on the spaceship and you're given a certain number of planets to visit. Those are goalposts; checkpoints by which the player can measure their progress in the second act, and theoretically the third act should be as long as the first. See? Easy. Knights of the Old Republic did that, Monkey Island 2 did that. No unpleasant surprises and the player is never unintentionally misled through poor design.
Some games telegraph their length with exceptional results. Right up front, Portal tells you: 19 rooms. Indeed there are, and so the player never thinks that room 15 might actually be a plot-critical gameplay escalation instead of a puzzle chamber. Portal continues after 19, of course, but here it works. It capitalises on the players' perception that the game is over; the "epilogue" comes as an intentional surprise more of the same. When you anticipate player psychology as Valve clearly does, then you can work with it.
You know how everyone in the world is able to pinpoint the exact moment that A.I. should have ended? Spielberg kept telling the viewer "one more thing", and the more times he said it, the worse the movie got. Unless you're Portal, unless you know what you're doing, when players think a game is ending, they should be right. If a game prompts players to say goodbye, then, one way or another, they will.