First, play our latest pyweek entry: http://www.pyweek.org/e/mamba/
The game went a lot smoother than pyweek 12's Nine Tales. I think we hit on a much better scale for this effort, given the time available, and, as a puzzle game, we were able to get the basic mechanics reasonably set early enough that we were never under pressure to have a game to submit, and so could do a fair amount of polishing. This also meant we actually spent a lot of time actually playing the game, so we could work off a few rough edges, which was good for the game.
We also spread things out a bit better than in pyweek 12. We avoided the major stumbling block of lots of central state people had to touch which we had in Nine Tales (the infamous game.json), which meant we had fewer merge issues than previously, and, by not using yet another version control system, benefited from the lessons learnt about mercurial last pyweek.
Unlike pyweek 11, where I ended up accidentally writing a level building tool, I more deliberately decided to write a level editor this time. While it's not perfect, and only learnt a few important tricks on the Saturday, the effort was, I think, well spent, as it allowed us to rope in spare people for extra level design effort, and made the addition of uploadable network levels a lot more feasible.
There are several issues. Most notable, our within a week widget stack, although inspired and influenced by past experience, is kludgy and lacks proper support for several important features. We should probably have a good look at the our last couple of pyweek entries and cut out a basic widget library from them, which can save us time in the future.
While not the best of our pyweek entries, I think we ended up with a fairly solid game, and it can fairly easily be extended by adding new levels, and, if we polish up the multiple tileset support, it would be quite easy to reskin levels.