A talk I gave for Warwick Game Design in March 2010. You will have to imagine my impassioned delivery of (and elaboration on) the material because unfortunately it wasn't recorded.
It takes a somewhat opposing view to Chris Hecker's Please Finish Your Game rant from GDC 2010. Namely, that while there are many games which should be finished, there are many more which should not.
In fact, I believe that the most effective way to create a great game is to first create a large number of games which are decidedly not great. Only by creating a large number of games will you be able to identify which ones actually have that "magic spark" of something with true potential.