The nature of algorithmic feeds like Reddit inherently leads to a survivorship bias: although users may recognize certain types of posts that appear on the front page, there are many more which follow the same patterns but fail.
Although visualizing basketball shots has been done before, this time we have access to an order of magnitude more public data to do some really cool stuff.
In general, it takes little additional effort to make something unique with ggplot2, and the effort is well worth it.
Using CPUs instead of GPUs for deep learning training in the cloud is cheaper because of the massive cost differential afforded by preemptible instances.