It's kind of the same thing as wrong medium, when something can be described via a in game mechanism and it's explained via text. Even worse if in the text X is "the most powerful faction" but they really stink when you play them mechanically. (That's just bad setting though, or bad mechanics.)
Non-modular settings. Where you have to take in the whole thing to even get started with the game. To put it in other words, how much do all the players need to know to start playing? If they can read a paragraph, or a page and be ready to play, that's good. If they have to read a chapter(s) to play, that's bad. This is more of the perception of the setting though rather than it's physical size.
A setting that can't be shown in game. This is somewhat the same as the last but a little differently focused. If you can't play the game and learn about the setting, you have to read it all, that's too heavy. Or wrongly heavy.
I think Micheal Moorcock said that anymore setting than the readers saw was just masturbation. For me its making sure the Storyteller and players grok the setting and tone. Detailing who is who can be dangerous because who really uses NPC perfectly how they are written? Places area good idea and power struggles either in the open or behind doors are neccessary if you have some sort of living campaign that will be updated. Even updating can be dangerous! be careful that the new additions to a setting dont knock previous character goals or creation ideas, creating a crime fighting costumed group that works well with the law is great making them outlaws next game with a power play the players have no say in can be real bad and put players off. These are just some examples I can think of off the top of my head.
I'll concur that literary conventions may not always work. yet what I also want to get at is also too much setting can become a put off. For example a player tried to spark my interest in Runequest, it made me feel to painted into a corner. Yes I could have broken cannon and played the way I wanted, yet I can do that with a home brew setting and remember the names of the cults and gods. So why buy Runequest and its 70million books and eras when I can get the joy out of making my own.
Some settings work well with just the main book and never a mention of a splat book, others force you to buy them to learn more about the world/system/fancy character/equipment section/monster manual. While I wont dispute that these books are a good idea and add often worthy and useful rules, I am speaking about the main rulebooks missing half the stuff/fluff/rules and forcing you to buy extra books for a complete game.
While I'm in Ashland, I've been auditing creative writing courses at the local college. In the first semester, all three of my eight page pieces were related to Dead and Back - and can be acquired from my . The impression most people gave is that they found the stories a bit too heavy on the jargon and felt I was just dragging them through the world, and would be lost if they hadn't read them all in order. My characters tend to be author avatars/audience stand ins to be dragged through a big setting. Not the greatest way to run a story.
On the other hand, an RPG is about the players, and focusing on the internal conflicts and interactions of main characters would not be all that playable. Meanwhile they do need a place to explore.
We can all give Rifts a lot of Flak, but with fifty or so books in the series, covering everything from mystical England to pseudo wild west, to Russian cyborg war bands there absolutely has to be something that will appeal to any gamer out there. Varied settings and decent visual illustration of it has won out over a thirty year old set of mechanics. Most of these books are taking place in separate areas, so you are free to pick and choose what you want to combine, or keep them apart.
Games of imagination are never truly done. Yet tomorrow we shall start another one.
My issue however for a massive design on a setting is it robs individuality while giving everything for the game. For example my favorite game system was Deadlands the Weird West (Original rules), buy the end they had a lot of books from the great maze to back east. Yet they were very careful to only detail it in a way that some of it can be used and the rest was design by the storyteller. I guess what I am trying to say is be careful with cannon too much and it becomes a hamstring to creating you're own stories especially if the cannon constantly changes.
In a setting I am currently working on its up to the players and the Storyteller to create the city in which many of the games will be played. I also will create a "Capital City"where big moves and shakes will happen that if the Storyteller likes them, he can steal them or have some of the repercussions travel through. Also if popular enough I will detail cities during major events such as World war 2 etc. Overall I like flexibility in a game system and the ability to consume it in bite sized chunks often in any order.
The other thing the Deadlands setting taught me is to make the setting enjoyable and have stuff especially set apart for the player and the truth for the Storyteller.
But what are the most popular settings? D&D which has nailed down nearly every conceivable variable. Star Wars, which can totally paint you in a corner with time lines and cannon. The biggest, longest running games have tons of background material.
I've done a bit of collaborative world design and found that some think it's the only way to approach world creation, while others just think it's me being lazy and not coming up with a setting. I've heard both extremes.
The strength of a detailed setting is that it anchors players who don't have their own stories to tell, or that want to tell that story. It also gives a common point of conversion between groups when discussing their games. "My group killed a dragon!" means something when a dragon is established. If it's just storytelling, the accomplishment is the emotional connection the game makes with the players and that doesn't transmit well.
It depends on what the play group wants. Some want cannon, some reject cannon. Rifts' strength is that it provides endless stat blocks for the players to encounter. The total narrative of the world is just flavor. But if I say "My group beat 50 skull bots on foot" (or whatever they're called) then the next group knows what we did and I can share some of the experience with them.
I can't do the same with Diaspora. There's no telling what I can run into. Some players don't care about that some do.
Hm. A bunch of interesting points here. I do want to throw in one additional thing, which is that sometimes what seems like setting 'bloat' can have another purpose, namely to motivate players to want to play the game. This can be done well (the detailed history of Glorantha perhaps) or not so well (the sometimes awful fiction in some White Wolf books).
Modularity was mentioned. My impression is that modularity is often assumed in settlingless game systems... a GM might pick and choose which monsters from the manual actually exist in the world (or will ever be encountered, which is the same thing). But an explicit 'The Setting is at everyone's discretion too' box up front, or something similar that outlines how either the GM or the players in a more collaborative way can simply decide that X and Y don't exist in the setting maybe should be included more often. That sort of thing shouldn't be needed, but often is just because some players will read the books as sacrosanct canon.
What's my rule of thumb? I guess I think that if a player borrows a book or gets hold of a pdf and has a week to read it, then the material should be such that they are not at a disadvantage to someone who has had longer to read the material. That means that yes, maybe there are additional details and bits and bobs that you might be able to fossick away at, but the really key stuff should be kept more or less in one place and should be readable in one lazy afternoon or a few snatchers in evenings. The old ADnD Dragonlance setting book is probably a good example of what I mean. It's meaty enough to seem interesting, but not so overwhelming that you'll be left reeling from the impact of factoids. So, I guess 100-200 pages of core setting tops with everything else marked Optional?
Of course, I've broken that rule myself a few times, so I what do I know