With music theory, it's possible to compose a piece that sounds good without ever needing to sit down at an instrument to hear it.

Possible, certainly. Common? Certainly not.

I challenge you to find me a notable composer who writes purely by music theory, and never listens to his composition until it is "perfect"...

Then perhaps the reason why this approach does not work is simply due to the current state of knowledge about game design.

Perhaps. But then again, perhaps it doesn't work because it's a needlessly inflexible approach which offers no mechanism for dealing with uncertainty.

Even if someone did come along and invent a Grand Unified Theory Of Fun, that only addresses issues with the design that crop up during the development. It doesn't even begin to touch upon other aspects of uncertainty that might crop up during development. Developers get sick, or have babies or change jobs leaving you short-staffed, causing deadlines to be missed. Recessions hit, causing investors to get nervous and cut funding. Similar competitive products may beat you to market, forcing you to either differentiate yourself or risk losing market share. Technology changes, perhaps requiring you to support new hardware and features, or perhaps offering you the potential to do something you wanted to do, but ruled out as it was unfeasible at the time of design.

There's a whole load of things that can change during a game's development, even if you take changes to the core design out of the equation.

The problem with people who have ideas for videogames who don't know anything about writing code is that they generally don't actually have ideas for videogames, at least not the right kind of ideas. I think in order to even understand what game design is you have to have implemented games from scratch.

Say you have a design document that specifies in elaborate detail a story, genre, art style, and the game mechanics for an RPG. Nobody cares about your story, genre, or art style; people who like fiction and stories can come up with their own ideas for this kind of stuff that suits their own tastes better than what you wrote that suits your tastes. The game mechanics you specify can either be novel or not-novel. Nobody cares if they are not novel. And if you are specifying novel game mechanics, why should anyone believe that they are fun? In this world, the easiest kind of game to write is a game that is not fun to play. The hardest kind of game to write is a game that is fun to play, but fun-ness can never be captured in anything but an implementation.

This is why the following is true:

The people who tend to be better "idea guys" than other people don't need to be encouranged to learn real skills, they've allready learned them.

Those people aren't called idea guys, they're called programmers, artists, game designers, etc.

Almost all of us started out as "Idea guys", i was an idea guy when i was 10, when i was 11 i started programming to turn those ideas into playable games and on the way i picked up some basic art skills and got pretty decent at proper game design. (I still struggle a bit with level design, especially for puzzle style games (Getting a fun basic mechanic is easier than getting a smooth, challenging learning curve imo, its quite difficult to judge the relative difficulty of puzzles you know the solution to)

This is precisely what happened in music with the development of music theory. By learning about music theory (which is a standard part of music school) one learns about all of the guidelines about how to compose music that sounds good. One learns about all of the patterns in harmony, voice leading, chord progressions, structures, and the like. By using a set of rules which consist of patterns that have been time tested, this takes away a huge amount of guess work in the process of composing a piece. With music theory, it's possible to compose a piece that sounds good without ever needing to sit down at an instrument to hear it. When the same thing happens to game design, it will take away the need for all of that guess work, trial and error, and iteration that happens in the studio. It will be possible to put together a detailed design document of a game that works great before even needing to step foot in the production studio. This process of developing game design theory is already happening, with the help of people such as those at Extra Credits, and Errant Signal.

This is a question of preference.

The best musics I have ever heard are never one-composer musics.

Can we get over this music analogy by the way? Half of the bulk of the thread is about refining this analogy that is set for what reason exactly?

It's just lazy/coward/deferring not to try out the ideas from the beginning.