The Blender Game Engine is at least as good as, if not better than all the other options currently available to beginners.
All the tools are integrated, and there's no paywall to get the "real" features. You can run your scenes from day one, and iterate quickly. Yes, it takes some time to learn, and you'll need to know Python to do anything more complex, but it's all there for you to use.
Some points I need to correct:
Unity uses DirectX where Blender uses OpenGL, this gives unity better frame rate even if directx is a bit harder to use.
I very much doubt that, but even if it were true, it would only be true on windows, and the overall fps delta would probably be negligible; OpenGL is the standard on all other platforms, and its performance is on par with directx (when used properly).
Python is a better language overall, and enables faster development, which is far more important than general execution speed.
To quote myself from this thread:
I think it basically reduces down to "the speed of light vs the speed of sound" argument: The speed of sound is ~880991x slower than the speed of light, but this is only obvious in specific circumstances, and it's only then that it actually matters.
Overall, I would recommend to focus on something that doesn't require "the latest and greatest in game technology", because those are projects that you could conceivably complete by yourself, or with a small team - Everything else is basically banging your head against the wall, until you finally realize that you were too ambitious, and that game development is generally difficult, even when making relatively simple games.
And look at some of the most successful indie games in history: How many of them are 2D games?
More than a few ... Think about it, and you'll understand what really matters (hint: it's not the technology).
In response to "the BGE is changing into a prototype-only environment" comment: I'm pretty sure that Ton made those comments just to "rattle the developers" a bit, in order to encourage discussion, and set more direct goals. In truth, no one really knows what's going to happen, but in general, the community wants to keep the engine in one form or another (there's no way to just "remove" open source software - it will always be around). Worst case scenario: The BGE is deprecated out of the mainline trunk code, and there's a version lock for a specific BGE build, which everyone will still be able to use. Best case scenario: The new prototype features are strong enough to serve as a better alternative, and everyone can simply use those, shipping their games with blender itself as a dependency.