I can tell you with certainty that very early and unpolished prototypes of Minecraft were fun, and that the potential was obvious. (Though of course no one could predict just how huge Minecraft became.)
Prototypes are your absolute best indicators to determine if a project is worth investing into to reach completion or not. Because you find your fun factor at the prototype level.
You'll get at most better visuals, sounds and more balance in the game with polish. If your mechanics aren't fun to begin with, polishing them will do absolutely no good.
I doubt it too.
> Prototypes are not good indicators that something could be fun.
This is what I disagree with. The prototype won't be as fun as the final product (because otherwise, why bother polishing in the first place?).
Take Infiniminer, an obvious source of inspiration for Notch. It's only got the very minimum of features; it's kind of a Minecraft prototype. It's also got the features that originally got me into Minecraft: building arbitrary structures in a large virtual environment. Minecraft is better fleshed out, of course, but the core of the game is there.
There's going to be a point where you've got the irreducible kernel of a game concept. That is the prototype. Remove items, crafting, mobs, infinite world sizes, lighting, and textures from Minecraft, and you've still got the kernel: a virtual building kit. It's an awesome idea, fun on its own, and expandable into a "real" game with goals, achievements, etc. It's a great indicator that a fleshed-out version of the game may be fun.
Now, you can't build something smaller than a kernel of the concept, call it the prototype, and make a judgement based on that.
It was a clone with aspirations of becoming something greater.
Prototypes are not good indicators that something could be fun.