Disclaimer as a friend of the author - but this is slick looking! I also like how I don't need to pay the New York Times to access Graphs.

I enjoy the basic feature a lot but I'd love to see a secondary challenge of identifying what precipitated the highest spike or drop if you get the first answer right. I think this would challenge the user in a different way - OK you can spot this trend - but how much do you really know it?

I think you could use a moving average and standard deviation to accomplish this in code - numpy has some native methods.

Great execution, keep it up!