Originality

This was asked here about a year ago and one of the mods (can't remember which) said that a beginner should go for original work because any clone that he makes wont be as good as the real one. But an experienced programmer should go for the clones (unless of coarse he has a really great original idea).
I think that the original work is always more rewarding than making a clone.

I think a far more important aspect of any game project that a beginner works on is Percent Complete. If this is for your portfolio to try to get a job in game development, it's more important to show that you can finish a project.

It's for PyWeek. Briefly - a competition for writing a game in a week in python. Started today at 00:00...The theme the game should try to follow is "The only way is up". So important stuff here is - finish on time, working game, some originality.

But I was aiming at something more general :)
Mikle

I think something completely original is not necessarily good. There should be identifiable characteristics so the end-user feels at home with your software. Even Half-life, despite its rave reviews, was not original. It was based on the Quake II engine and amounted to little more than a first-person shooter, from a mechanical standpoint. As a practical matter, the way it presented a story — entirely in first-person — was revolutionary at the time, so everyone talks about it.

Start with something familiar, and make it yours by giving it a shot in the arm. That's where originality really shines.

The point of contests like this one is to see how creative you can be in a limited amount of time. If you just clone an existing game, then it's purely a technical challenge and not very interesting. That being said, a week isn't a lot of time so it's fine if you take an existing idea or theme and add your own flavor.

In general, go for originality, but make sure you don't shoot too high and design something you can't finish.