Python is used very heavily in scientific computing and I use Python regularly at work. Even if not for games it is a really nice language for scripting your way out of problems.
Python is also widely used in the Django web application framework.
I much prefer the syntax of Python to Lua. Especially when it comes to OOP.
I recommend Python because it is in my opinion a great language to learn.
I suppose I should point out, I am not dumping on Python as a language; obviously there are realms where it is very effective. I am talking solely in the domain of game development; or even more specifically new programmers learning game development.
There are python bindings to SFML as well: http://www.python-sfml.org/ . There is also pyglet. For 3D there is Python-Ogre. Pygame runs on Android. As far as PyGame being slow, this thread talks about some things that can be done to improve performance.
Bindings are just not the same as a native game engine. They are generally second class citizens when it comes to support, you often have no documentation, you generally lag behind the main branch by a couple of versions. Plus it adds another layer for bugs to be introduced and for performance to be sapped. I KNOW that was the case with PyOgre when I checked it out a few years ago. Frankly if you didn't know the native language ( in that case C++ ), could luck puzzling out the API. When it comes to new programmers ( and this is an overly broad statement )... language bindings are generally a bad thing. There are exceptions, but they are just that, exceptions.
I hadn't realized that the PyGame for Android subset existed. In terms of performance, as I said earlier, you don't want new developers to have to think about optimization at all ( within reason of course ). That there are ways to make it perform better are little consolation for a new dev.
I respect your opinion and I am not going to downvote your comment, but I still think you are incorrect (I disagree about Python not going anywhere, that there is just Panda3D and PyGame, that it is nonexistant for Mobile, that it easier to learn as that is subjective, and I also disagree that Lua has far more support in the game development world but I concede that I have no sources to back that up... although the term FAR more seems a bit extreme? Here is a list of games that use python though...) 
.
Well, that's exactly how opinions should work, I'm glad to see you get the up/down vote concept ( people... you aren't supposed to down vote because you disagree with something... ) and your rational response is easily deserving of my upvote. In terms of "not going anywhere", well frankly Python really does seem to have stalled. The lack of backward compatibility between Python 3 and 2 seemed to REALLY hurt the languages development. Lua suffers this to a smaller degree between versions 5.1 and 5.2, but to nowhere near the same degree.
In terms of my FAR comment, it's easiest just to look at development tools out there that support Lua... you get lots of mobile toolkits ( LOVE, Corona, Gideros, Moai, Dreemcheest ), 3D Game Engine ( Havok/Vision, Gameplay3D, CryEngine, Polycode ) and those are just the ones that come to mind without a Google search. This of course doesn't include bindings. So while there are a dozen or so games written in Python ( many on that list seemed to be server side only too ), there are literally thousands written in Lua including some pretty big named titles, like the upcoming Shadowrun Returns. I think FAR is actually a reasonable fair adjective to use.
